Deneb is exposed

[Cookie Monster]

http://img98.imageshack.us/img98/2238/phen24zi1.jpg
http://img227.imageshack.us/img227/817/phen2ly2.jpg
http://img147.imageshack.us/img147/3362/phen23iy6.jpg
http://img147.imageshack.us/img147/5827/phen22gy8.jpg

Good god the 5 digit naming scheme was false.
Looks like AMD is back in the game.

[Raqia]

Any prices? Looks like more a match for Penryn and if it's ~$300 for the 940, great value compared to i7.

[pjbliverpool]

Yep, at 3Ghz and some nice core improvements this could be a decent contender for the top end Penryns. And baring in mind that a a 3Ghz Penryn comes in at $999 at the moment (I think) we could be talking some serous value.

I don't see this doing much against Nehalem performance wise though so it wil need to be priced accordingly.

[Albuquerque]

Any prices? Looks like more a match for Penryn and if it's ~$300 for the 940, great value compared to i7.

If you can make an argument for quad core on your workstation or server, then a ~$300 quad core 3Ghz Deneb compared to the ~$320 quad SMT-core 2.66Ghz i7 (http://www.newegg.com/Product/Product.aspx?Item=N82E16819115202) is a losing argument. They'll have to do better than that...

And sure, we could go down the road of "well motherboards are more expensive", to which I'd say -- for the massive performance increase you'd see on well-threaded applications (ie, those which would perform well on a quad anyway) then the cost will make up for it.

A more valid comparison would be Intel's quad core Penryn (aka Yorkfield) line that someone already mentioned, and that's more than likely where Deneb will be received best. Hopefully the core improvements on Deneb get them up to parity with Intel's Yorkfield line, which means the clock speeds plus the dollar value will look good to those who don't want to bump to the Core i7 platform just yet.

[Skrying]

Yep, it's a solid competitor to the Penryn based Intel products. Sadly it won't go anywhere when compared to Core i7. If AMD can be price competitive then they'll have a better-than-they-were-before product and not much else to be quite honest. I'm putting together a AMD based system this holiday season but only because they offer extremely compelling solutions at the low end, awesome motherboard + CPU combo for $100 is a great thing but anything over $250 combined and Intel just becomes much better very quickly. Hopefully Deneb will off set that for AMD. Competition is good!

[ShaidarHaran]

If you can make an argument for quad core on your workstation or server, then a ~$300 quad core 3Ghz Deneb compared to the ~$320 quad SMT-core 2.66Ghz i7 (http://www.newegg.com/Product/Product.aspx?Item=N82E16819115202) is a losing argument. They'll have to do better than that...

And sure, we could go down the road of "well motherboards are more expensive", to which I'd say -- for the massive performance increase you'd see on well-threaded applications (ie, those which would perform well on a quad anyway) then the cost will make up for it.


Let's not forget that dreadfully expensive DDR3 memory, or even (shudder) FB-DIMMs.

Intel's high-end platform is *significantly* more expensive than AMD's. It's not fair to brush this fact aside for performance reasons when the performance increase is not linear.

[Raqia]

If you can make an argument for quad core on your workstation or server, then a ~$300 quad core 3Ghz Deneb compared to the ~$320 quad SMT-core 2.66Ghz i7 (http://www.newegg.com/Product/Product.aspx?Item=N82E16819115202) is a losing argument. They'll have to do better than that...

And sure, we could go down the road of "well motherboards are more expensive", to which I'd say -- for the massive performance increase you'd see on well-threaded applications (ie, those which would perform well on a quad anyway) then the cost will make up for it.

A more valid comparison would be Intel's quad core Penryn (aka Yorkfield) line that someone already mentioned, and that's more than likely where Deneb will be received best. Hopefully the core improvements on Deneb get them up to parity with Intel's Yorkfield line, which means the clock speeds plus the dollar value will look good to those who don't want to bump to the Core i7 platform just yet.

Don't forget power consumption either. If I were in the market for a fleet of workstations at a large corporation, the savings from Deneb might be enormous next to the comparatively hot i7. I'll reserve my final judgement until I see some final head to head efficiency benches though.

Those are some strange cache figures for the 8XX series which each have 6MB and 4 cores. Are they disabling defective 2MB blocks there?

[rpg.314]

They have triple cores with 7.5 MB total cache when quad cores have 8MB total cache. It seems that they are disabling a core and it's private L2 cache, but L3 is same as for quads. IIRC, in agena triple cores, L3 was less than that for corresponding quad cores.

[Albuquerque]

Let's not forget that dreadfully expensive DDR3 memory, or even (shudder) FB-DIMMs.

Intel's high-end platform is *significantly* more expensive than AMD's. It's not fair to brush this fact aside for performance reasons when the performance increase is not linear.

Really, is it? What of AMD"s ranks is anything close to an entry-level Core i7 920 2.66Ghz? Please, do me a favor and point us in the right direction. Best I can see is a dual-socket Phenom X4 and two processors. So that's $170 each for two 2.6Ghz processors. Then you'll need a dual socket AM2 board -- good god, tell me about expensive. And you weren't joking about FB-DIMMS, which by the way, Intel doesn't need on the i7 platform.

Let me know how that works out for your argument.

Don't forget power consumption either. If I were in the market for a fleet of workstations at a large corporation, the savings from Deneb might be enormous next to the comparatively hot i7. I'll reserve my final judgement until I see some final head to head efficiency benches though.
Yeah, I'll point you to my statement above -- when they can reach parity with performance, you can tell me how much more efficient they are with power. Until then, the "parity" parts on Intel's side of the fence are still doing better.

[ShaidarHaran]

Really, is it?

Compared to DDR2, of course it is. $70 (http://www.newegg.com/Product/Product.aspx?Item=N82E16820231166) buys 4GB of 1GHz+ DDR2. What does that get you in DDR3? 2GB 1333 with some weak timings? Ha!

What of AMD"s ranks is anything close to an entry-level Core i7 920 2.66Ghz? Please, do me a favor and point us in the right direction.

I'd love to do that, but last time I checked Phenom II X4 benchmarks were few and far between...

Best I can see is a dual-socket Phenom X4 and two processors.

Phenoms (or Phenom IIs) are not SMP-capable, nor are Core i7s.

Yeah, I'll point you to my statement above -- when they can reach parity with performance, you can tell me how much more efficient they are with power. Until then, the "parity" parts on Intel's side of the fence are still doing better.

AMD still has the upper hand in terms of performance per watt (for servers) when you consider the entire platform. Intel will no doubt close the gap when Nehalem-based Xeons are released next year, but that's next year.

[Raqia]

They have triple cores with 7.5 MB total cache when quad cores have 8MB total cache. It seems that they are disabling a core and it's private L2 cache, but L3 is same as for quads. IIRC, in agena triple cores, L3 was less than that for corresponding quad cores.

The 8XX's are quad cores though. I would imagine that they're disabling a bank of L3 cache to get that 6MB figure since it'd prolly wreak havok to have no L2.

[rpg.314]

The 8XX's are quad cores though. I would imagine that they're disabling a bank of L3 cache to get that 6MB figure since it'd prolly wreak havok to have no L2.

That's interesting. Their earlier triple cores were simply quad cores with one core disabled to increase yields. This time they appear to going for disabling in the cache as well. It seems totally pointless to me to tapeout a new chip just to have 4mb L3 cache instead of usual 6M for 9xx.

IIRC, this sort of "disabling-to-increase-yields" is a new thing for CPU's. They have so far used speed binning.

[Raqia]

That's interesting. Their earlier triple cores were simply quad cores with one core disabled to increase yields. This time they appear to going for disabling in the cache as well. It seems totally pointless to me to tapeout a new chip just to have 4mb L3 cache instead of usual 6M for 9xx.

IIRC, this sort of "disabling-to-increase-yields" is a new thing for CPU's. They have so far used speed binning.

The fact that AMD has independent power planes for each of the CPU cores seems to indicate that they have more control over each of the units on their die. It probably wasn't too much harder to implement fuses that totally disable defective cores or banks of cache.

[rpg.314]

The fact that AMD has independent power planes for each of the CPU cores.....

I think nehalem has it too.

[Blazkowicz]

Disabling cache is very common.
AMD only does it for the lowest end (lowest sempron and lowest dual core), did it before (my xp2400+ was a barton with half L2 cache) ; and more glaringly the entire Intel line up has lots of disabled cache chips, from the celeron up to the relatively expensive Q9400.

(X3 do have 2MB L3 btw, you might mistake L3 size with the 1.5MB total L2 size)

regarding motherboard prices : the gap is huge at moment with an entry price of 40€ for a mobo that takes phenom and 260€ for one that takes i7!
granted, the cheap stuff with two DIMMs and two sata don't inspire confidence but there's plenty choice in the 60-90€ range. You can get a mobo + a phenom for the price of an i7 mobo.

[Raqia]

I think nehalem has it too.

Right, I just mean more control than they did in the past; hence defective core and cache shenanigans. It certainly wouldn't hurt Intel to do something similar.

[bowman]

There will be Xeon dual-cores based on the Nehalem architecture in 09, for 'low power' applications - surely they're not going to manufacture a different dual-core die for the LGA1366 socket, it's probably the GPU-style fusing off two cores tactic in action there, too.

[Albuquerque]

Compared to DDR2, of course it is. $70 (http://www.newegg.com/Product/Product.aspx?Item=N82E16820231166) buys 4GB of 1GHz+ DDR2. What does that get you in DDR3? 2GB 1333 with some weak timings? Ha!
I wasn't speaking to DDR3; I'm quire aware of the price difference there. I'm speaking to the entire AMD system that you're going to build that's going to perform like a Core i7 920 system... By the time you find an AMD system that can match it, you'll have far surpassed the pricetag of the i7 setup.

Phenoms (or Phenom IIs) are not SMP-capable, nor are Core i7s.
Again, in order for an AMD rig to perform like an i7 setup, you will need a multiple socket system. I made this plainly clear in my original reply. Core i7 doesn't need multiple sockets to keep AMD at bay...

[ShaidarHaran]

I wasn't speaking to DDR3; I'm quire aware of the price difference there. I'm speaking to the entire AMD system that you're going to build that's going to perform like a Core i7 920 system... By the time you find an AMD system that can match it, you'll have far surpassed the pricetag of the i7 setup.

Ludicrous assertion on face value, as you haven't defined your workload. Are we talking servers/HTPC/workstation or consumer use?

Again, in order for an AMD rig to perform like an i7 setup, you will need a multiple socket system. I made this plainly clear in my original reply. Core i7 doesn't need multiple sockets to keep AMD at bay...

In *some* workloads, sure. You know this is not true for consumer workloads, though. There's not a game out there that will run significantly better on i7 than Phenom II, at least not at similar price points.

[Albuquerque]

Ludicrous assertion on face value, as you haven't defined your workload. Are we talking servers/HTPC/workstation or consumer use?
Yet again, back to my original statement that you seem to love to stretch and construe to your illogical argument, any case that can be successfully argued for a quadcore will likely be threaded sufficiently to benefit from the i7. It doesn't matter what "market" you're targetting...

In *some* workloads, sure. You know this is not true for consumer workloads, though. There's not a game out there that will run significantly better on i7 than Phenom II, at least not at similar price points.
And yet again, there's no argument for >2 cores, then why are we even talking about Phenoms again?

[ShaidarHaran]

Yet again, back to my original statement that you seem to love to stretch and construe to your illogical argument, any case that can be successfully argued for a quadcore will likely be threaded sufficiently to benefit from the i7. It doesn't matter what "market" you're targetting...

If a workload isn't well-threaded it won't benefit from an architecture which targets enhanced performance in well-threaded environments. Games are not well-threaded. The majority of consumer software is not well-threaded.

If you're going to sit there and pretend like consumer workloads don't matter, I'm not going to continue this discussion.

And yet again, there's no argument for >2 cores, then why are we even talking about Phenoms again?

You're arguing in my favor now, as the same argument can be applied to i7. So why spend more money for an i7 if it won't deliver more performance than a Phenom II?

[Albuquerque]

Shaidar, now you're purposefully forgetting things to advance your inane argument.

This is my quote, verbatim:If you can make an argument for quad core on your workstation or server, then a ~$300 quad core 3Ghz Deneb compared to the ~$320 quad SMT-core 2.66Ghz i7 is a losing argument. They'll have to do better than that...

And sure, we could go down the road of "well motherboards are more expensive", to which I'd say -- for the massive performance increase you'd see on well-threaded applications (ie, those which would perform well on a quad anyway) then the cost will make up for it.

A more valid comparison would be Intel's quad core Penryn (aka Yorkfield) line that someone already mentioned, and that's more than likely where Deneb will be received best. Hopefully the core improvements on Deneb get them up to parity with Intel's Yorkfield line, which means the clock speeds plus the dollar value will look good to those who don't want to bump to the Core i7 platform just yet.
You directly quoted the first two sections of my reply (which I've happily underlined for you) which specifically state that we're talking about operations that scale to quad processors, and after quoting me in that statement, you made this statement:

Let's not forget that dreadfully expensive DDR3 memory, or even (shudder) FB-DIMMs.

Intel's high-end platform is *significantly* more expensive than AMD's. It's not fair to brush this fact aside for performance reasons when the performance increase is not linear.

I state properly threaded, you QUOTE me saying properly threaded, and then go on this tangent about cost. If you want to have an argument about cost but WITHOUT proper threading, then don't start your argument by quoting me WHERE I SAID WE'RE TALKING ABOUT PROPERLY THREADED APPLICATIONS.

Do you get it yet? Being purposefully dense like this makes you just look stupid.

[pjbliverpool]

In *some* workloads, sure. You know this is not true for consumer workloads, though. There's not a game out there that will run significantly better on i7 than Phenom II, at least not at similar price points.

That may be true for GPU limuted situations (and hence would hold over to Core2, Phenom 1 and probably even the AthlonX2), but in none GPU limited situations (i.e. high end SLI) Nehalem has astounding gaming performance besting a dual core Penryn at the same clock speed by anything from 40% to 200%! And that includes older, none threaded games as well (well FEAR at least).

http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/1

[tongue_of_colicab]

But there is only a very small amount of people that can / are willing to buy a high end sli/crossfire setup. I suppose you could say that falls unders shaidars ''some'' workloads. If 99%+ of the pc's sold doesnt have high end sli its not that interresting to have a 40 to 20% increase. Its probably not something most of the pc buyers are going to use as a argument to buy or not to buy a cpu.

[pjbliverpool]

But there is only a very small amount of people that can / are willing to buy a high end sli/crossfire setup. I suppose you could say that falls unders shaidars ''some'' workloads. If 99%+ of the pc's sold doesnt have high end sli its not that interresting to have a 40 to 20% increase. Its probably not something most of the pc buyers are going to use as a argument to buy or not to buy a cpu.

Yep I can agree with that. Most gamers don't need anything more than a moderate Core2 Duo.

I was just pointing out that Nehalem has FAR more gaming power than Penryn and hence Phenom 2 under its hood, even if that isn't allowed to show in most cases.

Its still nice to know though that if a games gonna need it, you have it in spades :smile:

[tongue_of_colicab]

I suppose. OTOH by the time games might use all that power wont your cpu be outdated anyway? Now I havnt really kept track of game performance as I hardly game on my pc so I might be very wrong on this but I have a q6600, but how many games really do make proper use of this kind of cpu's? are there already engines that can make good use of multicore on pc?

[pjbliverpool]

I suppose. OTOH by the time games might use all that power wont your cpu be outdated anyway? Now I havnt really kept track of game performance as I hardly game on my pc so I might be very wrong on this but I have a q6600, but how many games really do make proper use of this kind of cpu's? are there already engines that can make good use of multicore on pc?

A small handfull do, but they are becoming more common pretty quickly. Farcry 2 is a recent exceleent example of a game than benefits significantly from more than 2 cores.

You also have to consider how long your going to keep your CPU. I generally keep a CPU for at least 2 years and by then, GPU's with the power of 3 GTX 280's in SLI will be available for a reasonable price. I might appreciate having bought a Nehalem 18 months ago compared to something slower if I were to upgrade to a GPU like that.

[Jawed]

Also, Windows 7's/D3D11's multithreading drivers (which will also support multi-threading of earlier hardware - in theory) may well show up the value of a better processor architecture.

Jawed

[ShaidarHaran]

That may be true for GPU limuted situations (and hence would hold over to Core2, Phenom 1 and probably even the AthlonX2), but in none GPU limited situations (i.e. high end SLI) Nehalem has astounding gaming performance besting a dual core Penryn at the same clock speed by anything from 40% to 200%! And that includes older, none threaded games as well (well FEAR at least).

http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/1

I question Guru3d's results, as no other multi-GPU scaling investigation on i7 has shown anywhere near such increases.

[ShaidarHaran]

Shaidar, now you're purposefully forgetting things to advance your inane argument.

This is my quote, verbatim:
You directly quoted the first two sections of my reply (which I've happily underlined for you) which specifically state that we're talking about operations that scale to quad processors, and after quoting me in that statement, you made this statement:



I state properly threaded, you QUOTE me saying properly threaded, and then go on this tangent about cost. If you want to have an argument about cost but WITHOUT proper threading, then don't start your argument by quoting me WHERE I SAID WE'RE TALKING ABOUT PROPERLY THREADED APPLICATIONS.

Do you get it yet? Being purposefully dense like this makes you just look stupid.

Was a mistake on my part, nothing intentional. I simply forgot your premise and focused on the body of your text.

[Raqia]

This will definitely end up being a very popular enthusiast part; from the hardocp review:

http://enthusiast.hardocp.com/article.html?art=MTU4NSwxLCxoZW50aHVzaWFzdA==

it seems that the i7 chugs power and doesn't OC particularly well atleast compared to the claims being made for Phenom II.

OTOH i7 systems are cheaper than I thought they'd be; a base model was selling for 999 w/ 23" monitor and 4GB at Dell (for one of their day long sales), so partners seem to have pretty steep discounts on all the relevant components.

[Raqia]

Grossly unscientific platform oriented benches for Dragon over Spider (heh):

http://en.expreview.com/2008/11/18/amd-dragon-platforms-plan-and-performance-exposed.html#more-1307

I don't think they're even using the same graphics cards in those gaming scores.

On the CPU tests, we're seeing what a 15% boost in clocks speed from 2.6 Ghz Phenom to the 3.0 Ghz Phenom 2 gets us: 40% scaling in POVRay, Quicktime encoding, and OfficeFX (whatever that is) scores seem pretty promising.

[Gubbi]

http://enthusiast.hardocp.com/article.html?art=MTU4NSwxLCxoZW50aHVzaWFzdA==

it seems that the i7 chugs power and doesn't OC particularly well atleast compared to the claims being made for Phenom II.


You get a 30% increase in cpu frequence for a 15% increase in platform (ie the entire machine) power usage under load. Looks pretty damn good to me.

Cheers

[pjbliverpool]

I question Guru3d's results, as no other multi-GPU scaling investigation on i7 has shown anywhere near such increases.

Are you sure? Toms seems to agree with Guru's results. You have to take into account the resultions being used of course, lower resolutions give bigger differences because they remove more of the GPU limitation.

http://www.tomshardware.co.uk/core-i7-gaming,review-31407-8.html

I haven't seen any other SLI tests with i7 so I can't speak for anything that might contradict those results. On the face of it, single GPU tests and even the lower end SLI / X-fire might seem to contradict these results but in those cases it could simply be that the Penryn being compared too is already able to saturate the GPU setup and hence Nehalem isn't able to show its extra power. Even when game details are switched to low, this could still be the case as although the load on the GPU's is decreasing, some game settings will also decrease the load on the CPU.

[pjbliverpool]

This will definitely end up being a very popular enthusiast part; from the hardocp review:

http://enthusiast.hardocp.com/article.html?art=MTU4NSwxLCxoZW50aHVzaWFzdA==

it seems that the i7 chugs power and doesn't OC particularly well atleast compared to the claims being made for Phenom II.

Doesn't overclock well?? :???: They got 3.8 Ghz out of of a 2.66Ghz CPU on air cooling. Looks pretty damn good to me. What are we expecting out of Phenom II?

[pjbliverpool]

This has just occured to me so its probably just me being slow, but has anyone noticed that the new Phenom model numbers match up to Nehalem?

Could this be a return to the old AthlonXP naming scheme of trying to identify their processors against intels from a performance point of view?

[ShaidarHaran]

This has just occured to me so its probably just me being slow, but has anyone noticed that the new Phenom model numbers match up to Nehalem?

Could this be a return to the old AthlonXP naming scheme of trying to identify their processors against intels from a performance point of view?

Yep yep, I noticed that too. Although they don't have anything to match up with the 965 EE model. I wonder if that's intentional...

The other thing to note is that Intel's returned to the naming scheme they used for Presler/Cedar Mill, albeit shifted back 400MHz to accomodate the lower clocks of Nehalem.

[pjbliverpool]

Yep yep, I noticed that too. Although they don't have anything to match up with the 965 EE model. I wonder if that's intentional...

My thoughts exactly. In other words they're positioning the 3Ghz Phenom II against the 2.93Ghz i7 and the 2.8Ghz Phenom II against the 2.66Ghz i7.

If thats the case then my guess is its a marketing tactic. AMD likley won't be as fast at those speeds but it could be close enough for AMD to claim the same model number is justifiable. AMD will then simply undercut Intel on price which should give us a good idea of what kind of prices to expect for Phenom II. Nehalem minus $20-50 sounds about right.

[Raqia]

Doesn't overclock well?? :???: They got 3.8 Ghz out of of a 2.66Ghz CPU on air cooling. Looks pretty damn good to me. What are we expecting out of Phenom II?

Again, I'm saying this in comparison to rumors about the Phenom II OC'ing effortlessly past 4 ghz on air; I realize I may be forced to eat my words when the product actually comes out, but I am just totally speculating based on internet jawjacking on this point.

[Albuquerque]

Call me skeptical about 4Ghz clocks, especially on air, more especially given the 'stock' voltage.

And you can call me double-skeptical on the Phenom II being anywhere close to an i7 product -- be it single or multi threaded.

[ShaidarHaran]

I doubt anyone with a modicum of overclocking knowledge would expect a Phenom II to o/c to 4GHz on air with stock volts... Even Penryn can't do that, regardless of the cooling used or the chip bin.

[pjbliverpool]

Again, I'm saying this in comparison to rumors about the Phenom II OC'ing effortlessly past 4 ghz on air; I realize I may be forced to eat my words when the product actually comes out, but I am just totally speculating based on internet jawjacking on this point.

But even if thats true, your talking a 1Ghz overclock (assuming the rumors refer to the top end part). The Nehalem is already proven to get 1.133Ghz on air. (and in percentage terms thats much higher)

So while it would be nice, and indeed impressive to see a 3Ghz Phenom overclock to 4Ghz on air, I wouldn't say its particularly impressive in comparison to Nehalem.

[ShaidarHaran]

But even if thats true, your talking a 1Ghz overclock (assuming the rumors refer to the top end part). The Nehalem is already proven to get 1.133Ghz on air. (and in percentage terms thats much higher)

So while it would be nice, and indeed impressive to see a 3Ghz Phenom overclock to 4Ghz on air, I wouldn't say its particularly impressive in comparison to Nehalem.

True, but we need to frame this in context. Compared to the original Phenom parts it's a MASSIVE improvement, and very welcome.

[Raqia]

But even if thats true, your talking a 1Ghz overclock (assuming the rumors refer to the top end part). The Nehalem is already proven to get 1.133Ghz on air. (and in percentage terms thats much higher)

So while it would be nice, and indeed impressive to see a 3Ghz Phenom overclock to 4Ghz on air, I wouldn't say its particularly impressive in comparison to Nehalem.

Lower clocked Phenom II's will probably have more proportional head room than the highest end parts; the same holds for i7's as well. I wouldn't be too surprised if the 3.2 ghz i7 model doesn't OC very far past 3.8 ghz. Also, Kyle from HardOCP said in his video that the overvolted 3.8 OC on the 920 wasn't stable, I'd really just grant it 3.5 ghz (see 3:00 in his vid. on http://enthusiast.hardocp.com/article.html?art=MTU4NSwxLCxoZW50aHVzaWFzdA==)

i7 OCs are still impressive; it's just that if rumors are correct I'm thinking more along the lines of a stable 4.2 ghz OC from 2.6 ghz on the Phenom II. It might sound be a pipe dream but there are some persistent rumors floating around in presentation slides and word of mouth from motherboard engineers. The strongest evidence might be this positive report about AMD's 45nm process over Intel's: http://www.eetimes.com/news/design/showArticle.jhtml?articleID=212002243

Anyway, I realize banking on a big OC for an unreleased part is stretching things a bit thin, but it's good to see AMD better keeping pace with Intel and leaving enthusiasts some headroom and value. It would be a nice surprise if a $250 Phenom II OC'd can perform on par with a much higher priced 2.93 ghz i7.

[Thorburn]

In *some* workloads, sure. You know this is not true for consumer workloads, though. There's not a game out there that will run significantly better on i7 than Phenom II, at least not at similar price points.

I can't speak for Phenom II, but as someone who spends his days defining, automating and testing application and game benchmarks I would have to disagree with you there if you look at the current Phenom.

[Thorburn]

Again, I'm saying this in comparison to rumors about the Phenom II OC'ing effortlessly past 4 ghz on air; I realize I may be forced to eat my words when the product actually comes out, but I am just totally speculating based on internet jawjacking on this point.

I'd be inclined to agree, same as 4GHz on Bloomfield is apparently 'easy'.

I'm not saying it isn't possible, hell on a test rig with a loud cooler and a rather flexible definition of 'stable' its not that tricky at all, but theres a damn good reason my home system is running 3.6GHz.

Sites like Fudzilla throw numbers from rumors around with gospel and everyone listens to the idiots.

[ShaidarHaran]

I can't speak for Phenom II, but as someone who spends his days defining, automating and testing application and game benchmarks I would have to disagree with you there if you look at the current Phenom.

Well that's fine and dandy but we aren't discussing the Phenom, we're discussing the Phenom II.

[ShaidarHaran]

I'd be inclined to agree, same as 4GHz on Bloomfield is apparently 'easy'.

I'm not saying it isn't possible, hell on a test rig with a loud cooler and a rather flexible definition of 'stable' its not that tricky at all, but theres a damn good reason my home system is running 3.6GHz.

Sites like Fudzilla throw numbers from rumors around with gospel and everyone listens to the idiots.

Well-said. High overclocks are always glorified but rarely obtained. I also run my CPU at a mere 3.6GHz for stability, power, and heat reasons, despite being capable of much higher clocks (tested up to 4.2).

[Thorburn]

Well that's fine and dandy but we aren't discussing the Phenom, we're discussing the Phenom II.

True, for some reason I read the original statement as CPUs don't affect games. Will be interesting to find out how Phenom II fairs.

[Cookie Monster]

One thing ive noticed is that the AM3 revisions are all rated at a TDP of 95W. This allows room for higher clocked offerings with a TDP of ~130W.

Im guessing that the rebirth of the FX series could be well within the realm of possibility.

[Raqia]

Sites like Fudzilla throw numbers from rumors around with gospel and everyone listens to the idiots.

It is surprising though that AMD itself is touting OC figures on its BE parts, provided that those are real slides about the Phenom II that are presented to customers that we're seeing. If they are, I'd figure they'd be somewhat conservative with their figures or risk losing significant credibility. (Not that I'd put it past them or any other company...)

As for the lack of official parts clocked at higher speeds, I think they're waiting to bin enough high clock parts before announcing a new FX line. This would be in line with rumors of a 4.0 ghz FX part from earlier in the year. Ofcourse, this is all best taken with a grain of salt.

Edit: I hope this isn't like the BS Barcelona 3Ghz demo: http://www.pcper.com/comments.php?nid=6455 The 1.9v figure they cite for the 5 and 6 ghz OCs is a bit ugly tbh. The reasoning in that article for why AMD isn't releasing faster parts seems sound; they really can't afford a price war at this point.

[Cookie Monster]

Techreport (http://www.techreport.com/discussions.x/15927)

AMD overclocks 45nm Phenoms to 4GHz and beyond
by Cyril Kowaliski — 1:47 PM on November 20, 2008

According to AMD's latest roadmaps, 45nm Phenom II processors are just a few weeks away from launch—they'll materialize at the Consumer Electronic Show in early January. Not a bad time to start gauging the overclocking potential of these chips, right?

AMD took care of that itself at an event in Austin this morning, where it showcased four overclocked Phenom II systems. TR editor-in-chief Scott Wasson was on the scene, and while AMD didn't allow attendees to take pictures, he jotted down a few details.

The slowest system—cooled with a heatsink and fan—managed to reach just under 4GHz with a 1.55V core voltage. With liquid cooling, AMD successfully pushed a 45nm Phenom II in another machine just over the 4GHz mark. That required kicking up the CPU voltage to 1.6V, however. For the other two systems, AMD took out the big guns. One was strapped to a phase-change cooler and reached the mid-4GHz range at 1.7V, while the fastest system managed to break the 6GHz barrier using liquid nitrogen (which kept the core temperature down to a chilly -185°C).

Considering the latest leaked roadmaps suggest the fastest Phenom II launch CPU will run at 3GHz, those are pretty decent overclocks. We should note that AMD was using Crysis to test stability, however, and that only stresses a couple of cores at best.

Link 2 (http://www.pcper.com/comments.php?nid=6455)

The original Barcelona is a 65 nm part and really the first of its architecture. Since its release and the subsequent shuffling around of design teams, AMD has refocused their energy on delivering a design that would not only have a better per clock performance than the older part, but allow the design to scale in clockspeed on AMD's new 45 nm process. So far what we have seen shows that AMD has done very well. While the 3.0 GHz Phenom II will likely not perform anywhere near that of the new 3.2 GHz Intel i7, it will show very well against the older Intel Core 2 quads. Not only that, but we will see how AMD scales clockspeed in the future, as well as how they handle DDR-3 support early next year. With AMD's constant process improvement program, and tweaks to the Phenom II design that we typically see in the manufacturing stage, AMD could theoretically have a 3.4 to 3.6 GHz processor at moderate TDP levels (think 125 watts) that would more adequately perform against the mighty i7 within the next 8 months.

[Thorburn]

1.55V for 4GHz? On 45nm?

Chip won't last long like that.

[INKster]

1.55V for 4GHz? On 45nm?

Chip won't last long like that.

My feelings exactly.
I thought 1.55v was already pushing it..., back in the Athlon 64 "Newcastle"/"Winchester"/"Venice" (130nm/90nm) days, let alone now in a "supposedly" cutting edge 45nm process.

Even a typical 65nm Intel Core 2 Quad doesn't need that much juice to reach 4GHz, let alone a newer 45nm variant of the same chip(s).

[ShaidarHaran]

My feelings exactly.
I thought 1.55v was already pushing it..., back in the Athlon 64 "Newcastle"/"Winchester"/"Venice" (130nm/90nm) days, let alone now in a "supposedly" cutting edge 45nm process.

Yes, it does seem high. I had to put 1.57-1.59V into my old Opteron 175 to hit 2.6-2.7GHz on (chilled) air.

Even a typical 65nm Intel Core 2 Quad doesn't need that much juice to reach 4GHz, let alone a newer 45nm variant of the same chip(s).

Come again? 4GHz 65nm C2Qs are hardly "typical", at least on air ;)

[INKster]

Come again? 4GHz 65nm C2Qs are hardly "typical", at least on air ;)

The AMD demo had both air and water-cooled variants of the 4-to-4.5GHz Phenom II setups (although there are no pictures of either of them yet).
The "juice" (vcore) comparison is perfectly valid, IMHO.

[Skrying]

A 4Ghz Quad core is rare, these results are pretty good in that light. It's a decent amount of voltage but nothing shocking.

[Raqia]

These temps look promising:

"On air, the Phenom II X4 was able to hit around 4GHz at 1.55V with 32ºC temps. With water cooling at 1.6V the new Phenom II X4 broke 4GHz with a 39ºC core temp and in a Vapochill setup we saw 4.4+GHz. The final icing on the cake, so to speak, was the LN2 setup where the Phenom II ran stable at speeds well in excess of 5GHz and booted at speeds in excess of 6GHz."

http://hothardware.com/News/AMD-Sneak-Peeks-Phenom-II-Overclocks-To-5GHz/

[Accord1999]

These temps look promising:

I doubt the temperature readings are accurate.

[Raqia]

I doubt the temperature readings are accurate.

They look do pretty surprising; then again AMD supposedly reduced leakage a lot with this node.

[Moloch]

I doubt the temperature readings are accurate.
I would be willing to bet money it isn't, unless the system was in a fridge :wink:

[Mariner]

I would be willing to bet money it isn't, unless the system was in a fridge :wink:

Alternatively, it could be that they were using some fans with serious airflow. This would, of course, also indicate some serious noise.

[Thorburn]

I'd suggest the temperatures are meaningless, how were they measured?

[Dave Baumann]

Chip won't last long like that.
Even realtively minor process variantions can make changes to voltage/lifetimes - do you know the characteristics of the process being used here? :)

[aaronspink]

Even realtively minor process variantions can make changes to voltage/lifetimes - do you know the characteristics of the process being used here? :)

I doubt they characterized or designed the process to run much higher than 1.3 volts. at 1.5 and above EM will be a major factor, etc. if you upgrade every 6 months you might be ok, but beyond that, nope.

Dave its physics, at the geometries we are talking about 1.5 and above will have issues.

[Tchock]

I doubt they characterized or designed the process to run much higher than 1.3 volts. at 1.5 and above EM will be a major factor, etc. if you upgrade every 6 months you might be ok, but beyond that, nope.

Dave its physics, at the geometries we are talking about 1.5 and above will have issues.

The 75W ACP ("95W" TDP) Shanghai chip already has a default vCore of 1.35V.
As with the Phenoms there might be minor variations, but I've never seen a Shanghai CPU-Z so there you go.

So what you'd think the 105W ACP ("125W" TDP) 2.8Ghz SE part will have as a core voltage? 1.4V?

Surely that's not safe for servers already, ey? :wink:




The temps are more accurate than not, based on the reasoning that Crysis does not tax 4 cores at 100% (the temps were with Crysis benchmarks; 33 and 38 degrees for 1.55 air and 1.6 water respectively) However, with 1.35V as a reference, 1.5V-1.55 would likely give you ~150W, which the Cooler Master heatsink used would dissipate with a 18 Degrees Celsius delta compared to ambient temp (Tested by Frostytech on LGA775, but it's about the same).

[dizietsma]

Looks like for punters the new AMD 45nm process will be a lot better than previous AMD offerings against Intel.

[Thorburn]

The 75W ACP ("95W" TDP) Shanghai chip already has a default vCore of 1.35V.

Didn't realise that, gotta admit I've used a lot of Phenoms but been a while since I looked at anything AMD beyond the benchmark numbers.

[Skrying]

Well, personally I'm building an AMD based system. I don't know if it was the confirmed Phenom II support for the mATX motherboard I was looking at, or the $50 cheaper motherboard with equal features, or the $30 cheaper processor with equal performance that did me in... Now, if I was wanting to spend another $200 I'd probably have went with Intel but to be quite frank I much much rather have a GeForce GTX 260 or HD4870 1Gb and a Athlon X2 5600+ than a Core 2 Duo E5200 and HD4850. Toss in the fact that down the line my upgrade path is going to be very similar (higher end Core 2 Duo vs Phenom II) and it wasn't a hard choice other than getting over the "oh noes I won't be teh uber 1337" stigma around AMD processors right now.

[ShaidarHaran]

Well, personally I'm building an AMD based system. I don't know if it was the confirmed Phenom II support for the mATX motherboard I was looking at, or the $50 cheaper motherboard with equal features, or the $30 cheaper processor with equal performance that did me in... Now, if I was wanting to spend another $200 I'd probably have went with Intel but to be quite frank I much much rather have a GeForce GTX 260 or HD4870 1Gb and a Athlon X2 5600+ than a Core 2 Duo E5200 and HD4850. Toss in the fact that down the line my upgrade path is going to be very similar (higher end Core 2 Duo vs Phenom II) and it wasn't a hard choice other than getting over the "oh noes I won't be teh uber 1337" stigma around AMD processors right now.

:yep2:

Sounds quite reasonable. I've had difficulty recommending Phenom to any my clients and confidants, but Phenom II is a whole different ballgame.

Unless the only game you play is MS Flight Sim, you're better off getting a faster graphics card as you've said.

[pjbliverpool]

Yeah, if you want a cheap, but still decently powerful GPU AMD is definatly the only way to go right now.

Most of us as enthusiasts think £150 for a CPU is rock bottom so all we see is Intel but most regular people would prefer to spend <£100.

I was looking for something in that pric range for a friend the other day and by far the best bargain I could find was the X2 5600 at about £47. Even after throwing in a new mobo and 2GB of RAM it was still coming in about the same as the cheapest decent Core2 CPU.

And an X2 5600 is still more than enough to handle any game out there as long as you have a decent GPU.

[rpg.314]

Is Phenom 2 competitive with low end i7? This pricing leak (http://www.fudzilla.com/index.php?option=com_content&task=view&id=10665&Itemid=1) suggests so.

[Blazkowicz]

Toss in the fact that down the line my upgrade path is going to be very similar (higher end Core 2 Duo vs Phenom II) and it wasn't a hard choice other than getting over the "oh noes I won't be teh uber 1337" stigma around AMD processors right now.

exactly. I was more radical and went with the AMD single core (very low power, cheap and much faster than the athlon XP it replaced), there's a lot of upgrade options. such as a cheap 3Ghz K10.5 with 2x1MB L2 coming mid 2009, or a triple core.

[Skrying]

Yeah, if you want a cheap, but still decently powerful GPU AMD is definatly the only way to go right now.

Most of us as enthusiasts think £150 for a CPU is rock bottom so all we see is Intel but most regular people would prefer to spend <£100.

I was looking for something in that pric range for a friend the other day and by far the best bargain I could find was the X2 5600 at about £47. Even after throwing in a new mobo and 2GB of RAM it was still coming in about the same as the cheapest decent Core2 CPU.

And an X2 5600 is still more than enough to handle any game out there as long as you have a decent GPU.

I think you should go look at actual benchmarks. You seem to think the performance differences are greater than they actually are. Your type of thinking is what causes someone to buy a E8400 and HD4850 instead of a E7300 and GTX260. If you actually care about your gaming experience and not about how you'll look on online forums then the choice is easy.

I had a Core 2 Duo E6300 from basically day one of the Core 2 launch. If I hadn't of sold that system it would still have given 99.9% of the gaming experience a new Core 2 Duo would. But again, when I can spend $150 on motherboard, CPU, memory vs $270 for a similar performing Intel solution. Only when you spend that extra $200 for that extra .1% does Intel make sense, but that's just wasteful, I rather buy 4 new games instead and actually enjoy my hardware.

Call me crazy but I laugh at those who buy Intel + HD4850, it's just not the right gaming first choice.

[pjbliverpool]

I think you should go look at actual benchmarks. You seem to think the performance differences are greater than they actually are. Your type of thinking is what causes someone to buy a E8400 and HD4850 instead of a E7300 and GTX260. If you actually care about your gaming experience and not about how you'll look on online forums then the choice is easy.

I had a Core 2 Duo E6300 from basically day one of the Core 2 launch. If I hadn't of sold that system it would still have given 99.9% of the gaming experience a new Core 2 Duo would. But again, when I can spend $150 on motherboard, CPU, memory vs $270 for a similar performing Intel solution. Only when you spend that extra $200 for that extra .1% does Intel make sense, but that's just wasteful, I rather buy 4 new games instead and actually enjoy my hardware.

Call me crazy but I laugh at those who buy Intel + HD4850, it's just not the right gaming first choice.

I'm not sure where that came from, I was agreeing with you! :???:

[Accord1999]

But again, when I can spend $150 on motherboard, CPU, memory vs $270 for a similar performing Intel solution.
It doesn't take $270 to get an equivalent Intel system, not when the E5200 is cheaper than a 5600+ (and faster, overclocks better and uses less power), unless you think an Intel MB costs $120 more than an AMD one.

[Skrying]

It doesn't take $270 to get an equivalent Intel system, not when the E5200 is cheaper than a 5600+ (and faster, overclocks better and uses less power), unless you think an Intel MB costs $120 more than an AMD one.

The 5600+ gets put into combo deals, the E5200 does not. Using combo deals on Newegg is the absolute best way to save a lot of money. Without them my current build would be about $900 instead of $700.

[Raqia]

Some gaming results for the Shanghai Opteron:

Good performance on UT3:

http://www.gamestar.de/hardware/tests/prozessor/1951279/amd_phenom_ii_x4_p3.html

but disappointing results for COD4, even losing to the original Phenom:

http://www.gamestar.de/hardware/tests/prozessor/1951279/amd_phenom_ii_x4_p4.html

Caches are paramount to gaming so Phenom II sees massive improvements in UT3. Seem like there are some platform issues to resolve when it comes to gaming from the latter test. The PCIe interface on that mobo is probably not so ideal for the GTX280, nor is the slow ECC server ram in general. Reminds me of what happened to the original K8; while the Opteron wasn't so hot in gaming benches, the Athlon 64 was great.

[Blazkowicz]

I don't remember that actually, they sold the opteron as "Athlon FX 51".

[Moloch]

Some gaming results for the Shanghai Opteron:

Good performance on UT3:

http://www.gamestar.de/hardware/tests/prozessor/1951279/amd_phenom_ii_x4_p3.html

but disappointing results for COD4, even losing to the original Phenom:

http://www.gamestar.de/hardware/tests/prozessor/1951279/amd_phenom_ii_x4_p4.html

Caches are paramount to gaming so Phenom II sees massive improvements in UT3. Seem like there are some platform issues to resolve when it comes to gaming from the latter test. The PCIe interface on that mobo is probably not so ideal for the GTX280, nor is the slow ECC server ram in general. Reminds me of what happened to the original K8; while the Opteron wasn't so hot in gaming benches, the Athlon 64 was great.
You dont need to use ECC ram with S939 opterons (not sure about 940) so performance was identical to S939 A64s assuming equal cache...so your statement about Opteron not doing well in gaming is totally inaccurate.

[Raqia]

I was referring to initial platform bottlenecks for the first tests of the Opteron's desktop performance, not so much the chip itself and later iterations of its platform.

However, I definitely misremembered these benchmarks:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=1818

turns out Opteron was more than competent at gaming from the get go so my bad.

[Cookie Monster]

These are the supposed launch dates of Deneb/Propus

Link (http://www.dailytech.com/article.aspx?newsid=13571)

January 2009:

* Phenom II X4 940 at 3GHz, 8MB total cache using the AM2+ socket
* Phenom II X4 920 at 2.8GHz with 8MB of total cache on the AM2+ socket

February 2009:

* Phenom II X4 910 2.6GHz with 8MB total cache on the AM3 socket
* Phenom II X4 810 Phenom II X4 810 2.6GHz with 6MB total cache on the AM3 socket
* Phenom II X4 805 Phenom II X4 805 2.5GHz with 6MB total cache on the AM3 socket
* Phenom II X3 720 Phenom II X3 720 2.8GHz with 7.5MB total cache on the AM3 socket
* Phenom II X3 710 Phenom II X3 710 2.6GHz with 7.5MB total cache on the AM3 socket

April 2009:

* Phenom II X4 945 3.0GHz with 8MB total cache on the AM3 socket
* Athlon X4 615 2.7GHz with 2MB total cache on the AM3 socket
* Athlon X4 605 2.5GHz with 2MB total cache on the AM3 socket
* Athlon X3 420 Athlon X3 420 2.8GHz with 1.5MB total cache on the AM3 socket
* Athlon X3 410 Athlon X3 410 2.6GHz with 1.5MB total cache on the AM3 socket

June 2009:

* Athlon X2 240 2.8GHz with 2MB total cache on the AM3 socket
* Athlon X2 235 Athlon X2 235 2.7GHz with 2MB total cache on the AM3 socket


There also a rumour of the fastest Deneb being priced lower than the Core i7 920.

[digitalwanderer]

Would it make anyone feel better if I told them my Phenom II is running at 3.6Ghz without me trying hard at all yet....and its idle temps are like 22c? :|

[Thorburn]

Would it make anyone feel better if I told them my Phenom II is running at 3.6Ghz without me trying hard at all yet....and its idle temps are like 22c? :|

Not really :)

[digitalwanderer]

Well then I guess I'll just keep it to myself then. :razz:

[bearmoo]

Well then I guess I'll just keep it to myself then. :razz:

Lucky bastard :eek:

[Tchock]

Propus die size is out?

12.3 x 11.5 (mm) = 141.5 (mm^2)


Looks like this will be the first QC to reach $100USD. :O

[Thorburn]

Propus die size is out?

12.3 x 11.5 (mm) = 141.5 (mm^2)


Looks like this will be the first QC to reach $100USD. :O

But how will it perform with no L3 cache?

The Phenom X3's have been pretty cheap, but for the most part dual-core Core 2 parts have been faster.

[pjbliverpool]

But how will it perform with no L3 cache?

The Phenom X3's have been pretty cheap, but for the most part dual-core Core 2 parts have been faster.

Except in the only game where >2 cores matters... GTA IV.

[Thorburn]

Except in the only game where >2 cores matters... GTA IV.

My testing hasn't been extensive on GTA4 (I've done zero AMD testing), but I will admit it does scale very well with core count.

That said what I've seen of Phenom X4 results suggest C2D isn't far behind. I need to do my own testing before I can draw real conclusions though (I trust no one :wink:)

[pjbliverpool]

My testing hasn't been extensive on GTA4 (I've done zero AMD testing), but I will admit it does scale very well with core count.

That said what I've seen of Phenom X4 results suggest C2D isn't far behind. I need to do my own testing before I can draw real conclusions though (I trust no one :wink:)

Good! I look forward to your results!

[digitalwanderer]

Ooooh, wait'll you see how the deneb does with it! ;)

[V3]

I am seeing Phenom II starting to be available on some online retail, where are the reviews ?

[digitalwanderer]

Patiently waiting for the NDA to expire. :razz:

[DJ12]

Would it make anyone feel better if I told them my Phenom II is running at 3.6Ghz without me trying hard at all yet....and its idle temps are like 22c? :|
Are you using ACC and overdrive?

I've read a little about ACC and it makes a massive difference to the speeds you can get.

[digitalwanderer]

I'd tell you, but I think I'd be contractually obligated to kill you after so it's probably best I don't for now.

[DJ12]

Fair enough, I don't want to die :lol:

I didn't think ACC would be under NDA though, it's supported on already released AM2 Phenoms.

Or maybe they've got something better already....

[V3]

It's still NDA ? Well I found the PII 940 for the same price as i7 920. So I am curious at the performance.

[rpg.314]

When does the NDA lift then? Or is that under NDA too? :)

[digitalwanderer]

That's under NDA too, ask me again on Friday...

[Accord1999]

I am seeing Phenom II starting to be available on some online retail, where are the reviews ?
Here's a review in Greek:

http://babelfish.yahoo.com/translate_url?doit=done&tt=url&intl=1&fr=bf-home&trurl=http%3A%2F%2Fwww.hwbox.gr%2Fshowthread.php%3 Ft%3D3189&lp=el_en&btnTrUrl=Translate

The same site also compared Phenom II vs Yorkfield vs Kentsfield vs Core i7 at 3.7GHz and Phenom I vs Phenom II at 3GHz.

http://babelfish.yahoo.com/translate_url?doit=done&tt=url&intl=1&fr=bf-home&trurl=http%3A%2F%2Fwww.hwbox.gr%2Fshowthread.php%3 Ft%3D3253%26garpg%3D32&lp=el_en&btnTrUrl=Translate

[pjbliverpool]

Seems like a decent step forward and seems to hang in pretty well with the Penryns, albeit when its running at a faster clock speed. Still, if they are pricing the 3Ghz Phenom2 against the 2.8Ghz Penryn, its a decent alternative.

i7 slaughters all as usual! The Phenom II 940 is simply no competition for an i7 920.

[Skrying]

No one expected the Phenom II to reach Core i7 levels of performance, at least no one who isn't biased as hell and semi-intelligent. There was never a realistic hint that it could reach that level. The fact now however is that it is very competitive with the Core 2 lineup and has some advantages to those with AM2+ boards. But to be honest I think the line up will be in just as bad of a position come midway this year. The options right now are really rather horrible to be quite honest. AMD appears to have no plans but to really help those AM2+ users except for a very brief period and buying into an Intel system with intentions of maintaining your base system for two years is just... well yeah, not so hot.

Bad market right now.

[Brad Grenz]

I am hopeful my AM2+ mobo will support a Phenom II. Upgrading my CPU is my next planned move, but a full mobo replacement to go Intel will just be too expensive to manage. Asus has been really good about releasing BIOS updates for my M2N-SLI Deluxe to support everything up to the 9950 Phenom, so Ideally I'll be able spend around $200 and just swap CPUs. I've already got 4 GB of RAM, a Radeon 4850, lots of disk space, etc.

Actually, It's surprising how forward compatible the AM2/AM2+ platform has been. I don't think I've ever even considered a CPU upgrade without swapping motherboards before, and I've been building and upgrading my own systems since my first 486.

[Raqia]

Low power, low cost, OC's well, pretty good for games and certain applications. Guess the outcome isn't much of a surprise, you can't really expect it to do better clock for clock than Core 2 (not to mention i7) w/ only the memory and cache tweaks we heard about in the days leading up to launch. Competitive pricing is how AMD is playing things and this could prove dangerous since it's Intel's move now and they've had the performance lead for a while now.

There's seems to be about 5% of head room moving to the AM3 platform left from increasing memory controller as well as ram clocks from what AMD indicated. It's unclear where AMD is planning to move this year since Bulldozer is slated for '11. A high-k process in the middle of the year might boost clocks a bit, and Istanbul should offer quad channel memory and some more IPC enhancements but there's been no word of a desktop derivative.

[Ateo]

http://enthusiast.hardocp.com/article.html?art=MTYwNywxLCxoZW50aHVzaWFzdA==
http://anandtech.com/cpuchipsets/showdoc.aspx?i=3492
http://xbitlabs.com/articles/cpu/display/amd-phenom-ii-x4.html
http://www.guru3d.com/article/amd-phenom-ii-x4-920-and-940-review-test/

[pjbliverpool]

The Anand review is good (I haven't read the others yet).

Its very clear on which parts are competing with which. At the current price points its clear that Phenom II can compete well against the Penryn parts its targetted at. If you were interested in getting one of those parts (the Q9400 or the Q8200), then the Phenom is a viable alternative. Probably even the clear better alternative in the case of the 920.

There are two problems though. For only a little extra cash you can get better performance from the Q9550 and AMD has nothing to offer in that range. But even if your not willing to spend the extra cash, it will only take one relatively small price drop from Intel (which is rumored to occur) to bring the Phenom into competition with the Q9550 and Q9400 respectively. And then it once again becomes an unviable alternative unless your fixed to an upgrade path because of motherboard.

Its definatly a good step in the right direction, but AMD's new architecture can't come fast enough IMO.

Either that, or they need the capability to release faster clocked Phenom II's and be able to lower prices to remain in line with Intel performance. But i'm not sure either their technology, or their margins are capable of that.

[rpg.314]

Well, if they OC well, as has been said, then all they need is to wait a while, cherry pick them and release a higher clocked/BE part which can compete well.

Historically (in the good old Athlon days), AMD had a more efficient architecture and Intel compensated with it's process advantage by clocking higher. Today, it's the opposite and Intel has more efficient architecture. I guess, AMD will have to do very well in the next gen as well just to gain architectural parity.

[swaaye]

Coppermine was very competitive with K7/K75/Thunderbird even though AMD's was seen as the the "superior architecture". Look back at some old reviews and you'll see that the margin is very, very tight. It was just the first AMD chip to actually compete with Intel's best so people got really excited over it.

I actually think Coppermine was more efficient because it was cooler and basically as fast. It kept up mainly because its L2 cache was superb and the chipset->RAM performance was usually better than the VIA and AMD chipsets managed.

[Albuquerque]

Coppermine was very competitive with K7/K75/Thunderbird even though it was a "superior architecture". Look back at some old reviews and you'll see that the margin is very, very tight. It was just the first AMD chip to actually compete with Intel's best so people get really excited over it.

Out of fairness, it wasn't the first AMD chip that was competitive with Intel's best -- look at the AM 5x86 133/150 parts for 486 motherboards and how they competed with the (then-new) Intel P66 and P75 parts. Ever since then they kinda dropped out - mostly in floating point performance rather than anything else - but eventually came back like you said around the K7 days.

[swaaye]

Out of fairness, it wasn't the first AMD chip that was competitive with Intel's best -- look at the AM 5x86 133/150 parts for 486 motherboards and how they competed with the (then-new) Intel P66 and P75 parts. Ever since then they kinda dropped out - mostly in floating point performance rather than anything else - but eventually came back like you said around the K7 days.
True. I owned 5x86 and ran it at 160 for years. :) I didn't upgrade until Pentium II. That 5x86 was as good as a P75-P90 for the games that were around before 3D really kicked in.

But with Athlon it was more than just the CPU. AMD defined their own platform for the first time and it was actually of decent quality. AMD had been running on some very low quality platforms (usually Intel's leftovers) before Slot A / Socket A. Not that Socket A and Slot A didn't have problems initially, but it was still way better than most of the Super 7 hardware I've worked with.

[Albuquerque]

I never got into the Super7 stuff, but I remember a lot of chipset hell when AMD started the Slot A work. That was one of the primary reasons I was a Slot 1 user instead: too many stories and reviews about VIA and SIS (? I think ?) chipsets sucking unfathomable amounts of canal water even before you started overclocking...

The NForce line was great, but NF2 bit a bunch of people in the ass and made me skip that too. I wouldn't mind owning one of the upcoming SB800 setups though...

[swaaye]

Socket A came into its own once we had nForce and KT266A around.

I was a 440BX owner until about '01 when I went to a KT133A (Abit KT7A). That board was ok but had the 686B southbridge that gave plenty of people complete hell.