The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
I don't think there is a K10. The Barcelona design isn't K10, anyway. Barcelona looks an awful lot like a further continuation of K7/K8, really.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2939
AMD had worked on dramatic successors to the K8, rumored to be K9 and K10, but both appeared to be scrapped or at least focus was shifted away from them in favor of a more evolutionary take on the K8 architecture.

http://www.realworldtech.com/page.cfm?ArticleID=RWT051607033728&p=1
Barcelona is the first major architectural alteration to the K8, since it debuted in 2003.
 
Last edited by a moderator:
I don't see too much of a problem with just 2Ghz processors coming out at the start for servers as AMD have promised that higher performing ones will come out the quarter afterwards and this still fits in well with what they have been saying all along, first servers and then desktops.

The only people who will be unhappy are server folk who wanted the top speed processors at the start, but I would imagine there are a lot more who are happy to wait ( server people being fairly conservative ) or get the low power versions of the K10 when they are released.

The importatant thing is for the more speedy K10's to come out in Q4, be at least 2.4GHz and not be delayed again.

I can't remember exactly where I saw it (I believe it was Anand's piece) but AMD is said to believe that Barcelona doesn't truly "come alive" until around 2.8GHz.....So although a lower frequency part will be competitive.....it isn't expected (given what we currently know) to give Intel much trouble until they can effectively raise clocks up to 2.8GHz or higher...
 
I can't remember exactly where I saw it (I believe it was Anand's piece) but AMD is said to believe that Barcelona doesn't truly "come alive" until around 2.8GHz.....So although a lower frequency part will be competitive.....it isn't expected (given what we currently know) to give Intel much trouble until they can effectively raise clocks up to 2.8GHz or higher...

Man, I read that and the ghost of P4 walked up my spine. I get nervous when silicon is described as only *really* shining at some faster clock speed than can be hit right now.
 
Re: 2.0GHz... the rumours suggest that there will be at least one more stepping between the server and desktop CPU releases. So the initial "Barcelona" release will be capped at 2.0GHz, but the desktop version of the chip ("Phenom"?) will hit 2.8.

We'll see.
 
So, if you can't provide better performance and you can't provide a viable cost vs benefit for your product versus the competitor -- then what do you have?

Well if you have hundreds of servers running AMD socket motherboard then it might well be a cost efficient option to upgrade just the cpu to one which is faster than what you have currently and also uses a lot less power, rather than having to go to Intel which changes the whole caboodle ..all because you cannot wait 2 months to see what comes out the door in Q4 ? No it's if AMD cannot get up to large speeds in the future that people will jump ship and that is still undecided.

As for this "only shines at high speeds " business, that sounds like non linear performance gains which I think is poppycock. It can be none linear in a negative fashion due to other limitations but it cannot be non linear in a positive fashion.I think the Prescott proved this, as it got faster it's limitations in pipeline length became less of a problem but it cannot have been said to have "shone" :)
 
The delay bug has taken hold at AMD/ATI. I read this as clock speeds are the issue and the ramp and yields are not going nearly as well as hoped for either Barcelona or Phenom. I am not so sure AMD survives this if Barcelona and Phenom aren't able to ramp at competitive clocks on time.

http://www.digitimes.com/mobos/a20070702PD201.html

Mobo makers say AMD Phenom delayed until 1Q08, but AMD says launch remains set for 2H07 (update)

Monica Chen, Taipei; Joseph Tsai, DIGITIMES [Monday 2 July 2007]

AMD's Phenom processors were originally planned to begin test-production between September and October, and to start shipping in November this year. However, AMD may now be planning to postpone Phenom's launch date to the first quarter of 2008, according to sources at motherboard makers.

AMD intends to modify its process technology to increase the yield rate and frequencies of Phenom processors in order to improve the competitiveness of the CPUs against Intel's Penryn family, said the sources. AMD has already notified several motherboard makers that Phenom will now begin test-production in November or December, and will launch in first quarter of 2008, noted the sources.

However, AMD has responded to this report by stating that the company's official launch schedule for Phenom remains the second half of 2007 and this schedule has not changed. The company added that it has not contacted motherboard makers concerning a delay to the schedule.

With Phenom set to not appear in 2007, several motherboard makers have made plans to decrease shipments of socket AM2+ motherboards and to increase those of Intel's platform, added the sources.
 
It's starting to get to a point between who do you believe more, a web site or AMD.

"AMD intends to modify its process technology to increase the yield rate and frequencies of Phenom processors in order to improve the competitiveness of the CPUs against Intel's Penryn family, said the sources"

Has the 40% beating of Intel disappeared into thin air or was it never there to begin with ?

Is the K10 going to be another HD 2900XT ?
 
I don't think the "important" thing is clockspeed here, I think the real importance is AMD being able to clearly define what makes them a viable option.

2Ghz wouldn't be an issue if they could outperform Intel's offerings. Chances are, however, that they cannot. 2Ghz wouldn't be an issue if they could provide a better price-to-performance ratio than Intel. But again, chances are that they cannot.

So, if you can't provide better performance and you can't provide a viable cost vs benefit for your product versus the competitor -- then what do you have?

Can barcelona at 2ghz even outperform current k8s, in absolute performance, performance per dollar, or performance per watt?

Oh, and maybe the "coming alive" at 2.6ghz is either amd's way of saying "this is the speed we need to hit before you can expect us to compete with intel" or maybe it's got some kind of internal multiplier limitation and they can't clock the hypertransport bus full speed until it hits that speed. Yeah, probably just marketing bs.
 
It's starting to get to a point between who do you believe more, a web site or AMD.

"AMD intends to modify its process technology to increase the yield rate and frequencies of Phenom processors in order to improve the competitiveness of the CPUs against Intel's Penryn family, said the sources"

Has the 40% beating of Intel disappeared into thin air or was it never there to begin with ?

Is the K10 going to be another HD 2900XT ?

I am giving more weight to external sources outside of both AMD and ATI based upon AMD/ATI's recent delivered products and schedule compared to the internal AMD hype. AMD management has broken it's word for 3 quarters in a row now and ATI is well over 18 months. The Q2 2007 CC is going to be interesting.

Remember when Hector's favorite line at the CC was "We did what we said we were going to do".
 
Can barcelona at 2ghz even outperform current k8s, in absolute performance, performance per dollar, or performance per watt?

Per socket performance on anything decently threaded should be better than K8, since Barcelona has twice the cores and clocks more than half as high.

Performance per watt should be better or at least no worse on those same workloads, in part because AMD has to fit four cores into the same sockets as K8 due to it being a drop-in replacement.

AMD seems to be using the power angle, and it is likely helped by the fact that Barcelona's restricted clock keeps power low at the launch speeds.
 
Per socket performance on anything decently threaded should be better than K8, since Barcelona has twice the cores and clocks more than half as high.

Performance per watt should be better or at least no worse on those same workloads, in part because AMD has to fit four cores into the same sockets as K8 due to it being a drop-in replacement.

AMD seems to be using the power angle, and it is likely helped by the fact that Barcelona's restricted clock keeps power low at the launch speeds.

Can't really argue with any of that. And further, AMD still holds the overall performance crown when it comes to x86-64 last I checked. This is something Intel is working towards with Penryn (register space pressure was, I believe, the generally agreed-upon reason why the C2D cores don't run as quickly under x86-64). I know that our enterprise has brought in several AMD-based servers specifically for huge SQL boxes just because it runs better on AMD's 64-bit option than either our Itanium or Xeon boxes.

As for someone upgrading processors on existing servers -- I'd take that with a grain of salt. I'm sure there are businesses out there who outright purchase modern servers, but I'm going to wager that's not the majority. If I were a betting man, I'd put money on the majority of business that use large quantities of modern servers will be leasing them, which means you won't be replacing processors.

Of course, I have no numeric backing to my claim, which is why it's all just my opinion. But for the big names in my city whose IT shops I speak with on a semi-regular basis, nobody seems to be outright purchasing servers.
 
As for someone upgrading processors on existing servers -- I'd take that with a grain of salt. I'm sure there are businesses out there who outright purchase modern servers, but I'm going to wager that's not the majority. If I were a betting man, I'd put money on the majority of business that use large quantities of modern servers will be leasing them, which means you won't be replacing processors.

The drop-in replacement angle is probably more pertinent for AMD customers such as OEMs and system providers that don't want to have their system designs and inventory rendered obsolete by Barcelona.

Being able to plug a Barcelona in with only a bios update and knowing the chip will not exceed thermal design parameters made for Opterons would ease the transition.

(I think some large HPC machines have contracts that provide for CPU upgrades, though those are a minority share of AMD's target market.)
 
This is something Intel is working towards with Penryn (register space pressure was, I believe, the generally agreed-upon reason why the C2D cores don't run as quickly under x86-64). I know that our enterprise has brought in several AMD-based servers specifically for huge SQL boxes just because it runs better on AMD's 64-bit option than either our Itanium or Xeon boxes..

Also some low-level features like macro-ops fusion (If I remember this well) are disabled in 64 bit mode.
 
The drop-in replacement angle is probably more pertinent for AMD customers such as OEMs and system providers that don't want to have their system designs and inventory rendered obsolete by Barcelona.
Ah yes, that makes perfect sense when you put it that way.
 
How much is the deal with any company? I'd imagine they make more money off of microsoft than nintendo.

Honestly, I'd probably have to disagree here....If you remember, Microsoft gave NVIDIA a hefty fee for the GPU in the original Xbox. For Xbox 360, Microsoft opted to be extremely bullish with pricing....offering a much smaller amount for the GPU...

The GPU in the Wii is clearly cheaper with lower margins....but the volume is HUGE given the console's success....

I'm not positive Nintendo's revenue contributions are higher than Microsoft's...although I'm convinced it is anything but a landslide situation in Microsoft's favor...
 
Honestly, I'd probably have to disagree here....If you remember, Microsoft gave NVIDIA a hefty fee for the GPU in the original Xbox. For Xbox 360, Microsoft opted to be extremely bullish with pricing....offering a much smaller amount for the GPU...

The GPU in the Wii is clearly cheaper with lower margins....but the volume is HUGE given the console's success....

I'm not positive Nintendo's revenue contributions are higher than Microsoft's...although I'm convinced it is anything but a landslide situation in Microsoft's favor...

Microsoft has sold more 360's in total though, and aren't far behind nintendo in their current sales rate.
 
Status
Not open for further replies.
Back
Top