Phenom video leaks from Lake Tahoe

Hey I'm not saying that AMD hasn't been influential in many ways. I lived through it too. I liked my Athlon 64s, X2s, Athlon XPs and Duron. I do think they have benefited from underdog mind share a bit too much though. Intel has pretty much always had a product that was at least comparable to what AMD was pushing at various times.

My earlier post was a response to the thought that suddenly optimizations can save AMD's new architecture. I don't think that will happen, and I brought up that I believe AMD designs their chips to function as well as possible on Intel-optimized code. There is plenty of evidence to support this.

Why should companies specifically optimize for AMD anyway? It seems that, like NVIDIA, Intel pays attention to developers and sponsors them. AMD (and ATI) never seem to bother with such developer relations. AMD has never been able to market anything impressively.... Companies aren't going to develop intensively for them without reason or incentive. This is especially true when there is incentive from a bigger, more influential competitor whose products are in more of their customers' hands.
 
Last edited by a moderator:
Is that also why Intel went with x86-64 support in Core 2 long after AMD had established it as a standard with the A64?
You mean Prescott, right?

If you'll recall Intel didn't want to go to Core 2 in the beginning--it wanted everybody to move off of x86 and go to Itanium to get their "64-bits on the desktop." As well, I believe I read a few weeks ago of AMD proposing SSE5...and making plans to implement it in '09...?
Intel rather quickly discounted IA-64, and 64-bits in general, for the desktop very early on.
The big mistake was its plan to keep x86 limited to the low end to allow Itanium room to grow in the server and workstation markets.

AMD neatly slid in to that gap, and Microsoft in particular was happy that it did.

There is no overarching reason why anyone should care about SSE5 at present.

Basically, I think the influence of AMD on the direction of the markets since the introduction of the k7 in 1999 has been huge--so huge it is difficult to estimate, imo. Even Microsoft agreed with AMD that there were much better ways than Itanium to do "64-bits on the desktop" and thus Windows x64 was developed and marketed, but not unfortunately until after a lengthy delay designed to allow Intel to get on board the x86-64 train.
That is not what happened. Intel's official line, once it was clear from internal evaluations that Itanium did not supply enough performance with backwards compatibility, was that 64 bits was not needed for the desktop.
Even with the bloatware of Vista, it is still mostly correct.

64-bit Windows had everything to do with Microsoft trying to break into a high-end market 32-bit Windows was shut out of.
There is no corresponding "OS with FMADD support" market Microsoft is shut out of, so AMD has no further product gaps to exploit.


...

I've just snipped a whole lot about Intel's nefarious deeds that have absolutely nothing to do with Phenom not matching CPUs that have been out over a year.
 
Let me also ask you a logical question...if AMD really didn't think that Phenom could be "competitive" from a price-performance perspective, then why would AMD release Phenom in the first place? To short circuit you answering with "AMD has nothing else," let me just assure you that we don't really know how competitive Phenom is going to be with Core 2 from here on out.

So the "logic" is... The reviews are bad... but why would AMD make a bad product... so the reviews must be wrong. Brilliant. I am almost tempted to pull up some of your more juicy R600 posts, just to check on how those "give it time, Kevin" predictions turned out. Almost.

So to answer your question, "What is AMD going to do?" I can think of only one answer, and that is that AMD is going to continue to *compete* as it has always done. Look, if pre-K7, Intel didn't have enough clout to make AMD dry up and blow away, why on earth would you assume Intel has a better chance of succeeding today, when AMD is so much more competitive in every way than it was then?

Just a guess here, but maybe because AMD was not losing over $2 billion a year back then?
 
BTW, has anyone pointed out to Walt that many of those horrible evil reviews were configured and orchestrated by AMD?
 
mmmmm, pass the popcorn...

popcorn.gif
 
BTW, has anyone pointed out to Walt that many of those horrible evil reviews were configured and orchestrated by AMD?

I'm quite certain he's figured out that part already. The thing is, when God gave Moses the 10 commandments atop mount Sinai, he was also a tad bit ellyptical, so Moses didn't get it at first. It took him and his descendants time to figure out what God actually wanted and what He was giving to man. It's the same with Phenom...a couple generations from now, ppl will finally get the greatness of this CPU, and thus finally be able to comprehend what AMD is offering now. Coz AMD is kindof like God...kindof.
 
I'm sure that's what you think. Sure, let's just discount the fact that Core2's been shipping for a year and has undergone numerous supporting core logic changes, supporting system bus changes, fsb memory controller changes, core bios revisions, compiler revisions, and benchmark optimizations that play to Core 2's strengths---not even, as you say, counting process changes and refinements to subsequent revisions, or even clock speeds. Let's fantasize that all of those changes over the course of the last year just don't matter to Core 2's performance today.

Ah, you're talking about the 0-2% gain seen in some of the reviews of the 1333 FSB procs? Sorry Walt, even the first generation Core 2's would kick Phenom's ass thoroughly today. There has been no tangible improvement of the Core2 processor or its supporting platform since the launch last year.
 
This is an industry where stuff arrives, has a short life, and then is discarded in the face of a better, newer model. Is anyone here naive enough to buy stuff based on the marketing promise of "it will get better at some point in the future"? Especially so in the face of a competing product that gives you what you want now?

Why have "jam tomorrow" from AMD, when you can have "jam today" from Intel?

Phenom needs to get better, faster and cheaper, than it is. RD700 needs to get cheaper and have a properly paired southbridge today, not some time in March. RV670 needs to clock higher (but probably does with overdrive).

Honestly, it's such a shame that Spider is being pulled down by slow Phenom and a mismatched southbridge.
 
Here's the difference between now and when K8 launched:
K8 was FASTER than the competition FROM DAY ONE.

Since Phenom is slower than last gen's top bin part, which was already slower than the competition's parts several bins down from the top, they have a virtually insurmountable disadvantage to overcome. Given their financial situation, the lack of OEM support (where are the OEM design wins for Phenom/Spider?) and Intel's huge manufacturing advantage, I think Phenom is DOA.

I would love nothing more than for AMD to be competitive in all market segments, and even win back the performance crown, but let's not try to sugar-coat a turd here.
 
That's too bad, since AMD controlled most of the vague and unclear reviews.
Coincidence?

Well, what a brilliant stroke it was for AMD to "control" these reviews to the extent that all of the reviewers were supplied with handy engineering samples of pre-production 45nm Core 2 cpus. Your contention is, then, that it was AMD's idea to contrast the 65nm cpus they were launching today with the 45nm cpus from Intel which have yet to ship? Your position is that this was all AMD's doing?....;)


That should tell you something.
The product is days away from supposedly shipping in quantity, and the platform is in shambles.
It's a repeat of the Athlon days when Intel pressured motherboard makers to not make Slot A motherboards.
Only this time, the board makers would kill to make an AMD board, only AMD's failed execution has stopped them.

So your contention is that when AMD is ready to ship them that nobody will ship them? I mean, if they'd "kill" to ship the motherboards this week, I see no reason that attitude might dramatically change next week, do you?

There is no way in hell that a final product that is shipping for real money should be in that position.
Opteron and A64 had running boards and near full-speed processors months prior to launch.-/

Opteron was by some accounts a year late. By others it was two years late. And after launch, Opteron ramped up in MHz and yields--just like Core 2 did after it shipped at ~1.9GHz. Regardless of the lateness of Opteron, which could be debated from here to eternity, the delay did not prevent the chip from being a smash success when it shipped, did it? Some might say that Core 2 was 3-4 years "late"--which hasn't detracted from its commercial success in the slightest. When new cpus ship--and this is true for everybody including AMD--what is represented at that point in time is the beginning of the cpu's lifespan, as opposed to the end, right?

Phenom's status--as a product people pay real money for--is unacceptable, and the fact that AMD and partners are going through with it is indicative of desperation.

Let's not forget that the lousy chipset situation and motherboard issues really hurt AMD's gains with K7.

Well, then--what about 45nm Core 2 in terms of "status"--in terms of product people pay real money for? (I didn't know you could buy products with play money, but that's beside the point...;)) Regardless of that status, that didn't prevent users from using the results garnered from essentially the cherry-picked 45nm Core 2 samples Intel sent out to people expressly for the purpose of diluting Phenom's launch, did it?

About the K7's initial chipset/motherboard problems--don't forget that Intel was directly behind many of those problems, as we've known for quite some time.

No, Phenom's status is significantly worse.
There are questions about stability and functionality days prior to launch.
How often have either Intel or AMD ever admitted that there is a critical show-stopping crash bug that basically limits CPU clock speed to less than the top bin of chips from two process generations before?

Questions posed without answers are merely speculation--gossip, if you will. As far as your remarks about clockspeed are concerned, you are aware I'm sure that Phenom does not share the architecture of earlier AMD cpus, and therefore direct comparison with the clockspeeds of earlier architectures is likely to tell us absolutely nothing...?

Memory results on a number of programs (including AMD's) do not correctly detect the memory controllers.

Which seems to dramatically illustrate the point that Phenom is indeed a new architecture. Should we expect that utilities written for the P4, even if written by Intel, should always work with Core 2? I would never make such an assumption.

That being said, AMD's greatest success didn't come when its products ran AMD-optimized software best.
Remember K7 vs P3 and K8 vs. P4?
AMD made money by making processors that beat Intel at its own game.
They ran Intel code better than Intel did.

I guess that here your supposition must be that nobody ever optimized for AMD cpus, and that AMD did nothing to push its optimized compilers/optimizations out to anybody? If so, I'd have to disagree. It's well known that Intel optimized for the P4 in all these respects--and of course there is absolutely nothing wrong with that at all. It seems only logical, however, since the P4 and the A64 were and are entirely different x86 architectures, to expect that like the P4 benefited the A64 would benefit from AMD optimizations of various kinds. I think that is beyond argument. In some cases that I can clearly recall, P4 compilers placed flags designed to clearly disadvantage non-P4 x86 cpus like the A64; in other code, for instance some benchmark code, the degree of P4 optimization was so heavy and pronounced that A64 was automatically disadvantaged when running it.

While I might say that sometimes the A64 ran rings around the P4 when running certain types of generic x86 code, not even counting 64-bit code, I would never say that that the A64 could run highly optimized P4 code better than a P4, for obvious reasons. IE, there's a big difference between "x86 code" and "Intel code," as one need not necessarily be the same as the other at all.

Now, AMD is having a hard time making a chip that beats K8 in some situations.
I guess we should blame Sandra for being K8-centric.

Interesting observation, because I also noticed that in a few of these benchmarks the Phenom tested ran all over the Q66/6800's...;) So I guess what the Phenom is either slower than or faster than is highly dependent on the software being fed to Phenom, isn't it? Sandra, as I said, is mostly Intel-centric, and always has been. Efforts to better represent the differences between the P4 and the A64 inside Sandra always came long after the fact--with the OOB Sandra experience being very sympathetic to Intel architectures. Accordingly, Sandra is a program that I have installed, and then uninstalled, at least four times over the years--always being unhappy with how the software had difficulty in even correctly identifying the hardware I was running at any given time.

Just how long from now is "too late"?
Let's decide on a date, and we'll mark it on our calendars.
When that date comes to pass, we can all get together and then we can discuss things over tea.

For a successful product, the degree of lateness would seem not to matter--as I pointed out above with respect to both the A64 and Core 2. If Phenom proves itself unsuccessful in comparison with Core 2, it will not be because Phenom was late, it will be because Phenom was so inferior that no amount of price disparity could serve to make it attractive in volume. If Phenom otoh proves itself a successful competitor to Core 2, then the fact that Phenom was late out of the gate will simply not matter, because it will succeed on its merits as opposed to its calendar release date.

Phenom has been released in 4Q 2007, with a top bin half a GHz below the originally planned speed on a motherboard a revision back from what it should have been.
It has chips that cannot be trusted to overclock, because even if they did clock high, they'd freeze the system.

Which is part and parcel of all overclocking, isn't it? You can take a Core 2 cpu and if you overclock it enough you can "freeze" the system, too...;) This was especially true when Core 2's first shipped at ~1.9Ghz, but really, it is true of all cpus. Some of them can be overclocked to a great degree, and some of then cannot be, but the fact is that *none of them* has been validated by the manufacturer to successfully run in all situations at clocks higher than those at which they are sold.


I'm sure the one Phenom AMD manages to build in the next quarter will be there for you to buy.

Heh...;) Yup, right along with the "one" 45nm Core 2 Intel manages to build next quarter...;) That's a bit of a ridiculous comment, don't you think?


So AMD's latest and greatest can't even trump its old value lines?
Sounds fantastic.

Again, how does this statement square with the fact that AMD's latest and greatest easily walks all over a Q66/6800 when the two cpus are fed the appropriate benchmark software--and even when the Intel Q's are clocked half a GHz higher than the Phenom? The fact that Phenom is inconsistent in that regard seems to me entirely consistent with the fact that Phenom is a new architecture which has not yet been properly optimized for or supported by much of the benchmark software that pretends to be able to measure its performance potential. I think these inconsistencies will sort themselves out in time as the Phenom architecture becomes better known and supported.


I'm sure Phenom-optimized programs will come out.
Going by the current crop, the primary optimizations will be to downclock the chip to 2.3 GHz and only run on one core.
Either that or enable the hidden time machine to bring the Phenom release up 6 months, or do you really think (edit: not) beating a (edit: lower bin of) chip that is been out for over a year a staggering accomplishment?

Again, you are ignoring the benchmarks in which the Phenom easily bests the Q66/6800's as well as older Athlons. Sparse though such examples may be at present, they do indicate dramatically that the situation is nowhere near as consistent as you represent it.

I'm also somewhat unsure as to where that 30-50% factor is coming in.
Seriously, I haven't seen anything to indicate that a decent motherboard and processor from either platform has that large a disparity.

I've seen prices quoted for the top-end Core 2 cpus in excess of $1,000. It was to those that I was referring, the price of the motherboards at that point being moot.

The contest they're declaring over right now is Phenom versus 65nm Core2.
And given Intel's schedule for production, they're pretty much right.

I fail to see why they'd declare it "over" since Intel isn't yet shipping 45nm Core 2's at any level that could be considered comparable to the volume of 65nm Core 2's it is shipping, if Intel is shipping 45nm Core 2's in any volume at all...;) As is evident in this industry, a company's production plans for a product do not equal actual production of the product itself, as often even the best-laid plans of mice and men go awry...;) IE, it isn't wise for any company to do too much of counting its chickens before they hatch. I'm sure you know that Intel has wound up with egg on its face many times in the past for doing just that.

The basic thrust of what you seem to be saying here is that you regard Intel as a perfect corporation always manufacturing perfect products, perfectly on time according to its pre-announced schedules. Of course, I would energetically disagree with any such characterization of any hardware company, AMD and Intel included. OTOH, you don't seem to be able find anything approaching perfection when it comes to your assessment of AMD.
 
Walt, I'm not sure which alternate universe you live in where Core 2 only shipped @ 1.9GHz at launch, but in this one they shipped the X6800 @ 2.93GHz the same day as the rest of the original 6xxx family members.
 
WaltC is known for... lets call it... very positive outlook on AMD/ATi parts. Not surprising that he's trying desperately to make it look "good" instead of the steaming pile of bad that it is. In fact, I think he's at a point where he even buys into his bullshit.
 
WaltC is known for... lets call it... very positive outlook on AMD/ATi parts.

And I'm not? ;)

Not surprising that he's trying desperately to make it look "good" instead of the steaming pile of bad that it is. In fact, I think he's at a point where he even buys into his bullshit.

I'd love to believe that compiler optimizations & software recompilations will bring about measurable performance increases for Phenom yet somehow won't also raise the performance of C2Q processors, but I don't live in a fantasy world so I can't believe that. I don't really buy anything he's selling, and wish AMD would wake up and smell the coffee. When one of their most hardcore fans abandons them altogether, they need to take a hard look at their failure to execute.
 
I know it's probably embarrassing for AMD to admit it, but they really need this out now, badly:



If they have any hopes of riding the wave and sell CPU's at higher profit (maybe even GPU's -why not ?-, on the cheaper 750a models), the motherboard chipset also needs to entice prospective buyers with any new feature they can muster.
So what it's from another vendor ? It's not like it hurt them during the K7/K8 glory days, isn't it ?
 
Anyway, back in the real world.

After a year of hype, AMD fastest newly launched CPU that is slower than competitor's SLOWEST comparable model which has been on the market for 10 months, while consuming more power that the aforementioned. To someone who was really looking forward to what K10 can do, this f-ing sucks.

[FONT=arial, helvetica, sans serif][/FONT]
 
Oh, the memories...

The last minute and a half = pure gold.. If there is ever an investor lawsuit against AMD, it might crack Top 10 on the plaintiff's evidence list.
 
Last edited by a moderator:
Sometimes reality is a hard pill to swallow. You are NOT going to see vast optimizations for Phenom that will bring it on par with Intel's Core architecture no matter how long you wait. It's just slower, plain and simple. And that's before accounting for Intel's clock speed, power consumption and process advantages.
Intel does have a tremendous process node/fabrication advantage. Without much better IPC, AMD's PTP advantage won't outperform C2Q on the desktop.

Fuckin' SB600...it's all your fault.
To not primetime SB700 with the RD790 chipset was unfortunate...

The last minute and a half = pure gold..
Yeah, I remember that one. Certainly cringeworthy in retrospect.
 
Last edited by a moderator:
Back
Top