4.23.2007

BARCELONA NOT FAST ENOUGH - AMD is DOOMED!

AMD today disclosed performance projections for it upcoming Barcelona Quad Core. The "projections" shows the new 65nm monolithic QC to have up to 50% FPU advantage and 20% advantage in Integer over Intel's highest performing QC's at the same frequency.

This raises a lot of suspicion. AMD used to be clear cut when it comes to boasting performance numbers. What you see today is some vague-lawsuit-evading announcement with qualifiers: "intel's highest performing QC's" and "at the same frequency". Either AMD extrapolated the Barcelona performance numbers to 3GHz or they extrapolated down intel's highest performing QC to whatever god-awful speed they manage to get Barcelona running. in any case, AMD can't seem to claim absolute performance. And if by chance of a miracle that AMD is telling the complete truth, their advantage will be short lived as Penryn has proven to be offering almost the same performance advantage over current QC's.

Given that we already know what speeds Barcelona will be offered, a 20% in INT will not be enough. Not with the clock advantage and overclockability of C2D and the upcoming 45nm speed demons from intel. At an indistinguishable performance advantage with over 3X cost per CPU, this is most certainly a lost cause for AMD.

In 2005 AMD said they have the performance advantage. In 2006 they said they had the performance-per-watt advantage. Today, in 2007, they're claiming to have the performance-per-watt-per-square-silicon advantage. I would like you to think about what that means. If you are smart enough to be able to understand and visualize his new metric from AMD, I'm sure you would come to the same conclusion as I have.

AMD is DOOMED.

16 comments:

Anonymous said...

...performance-per-watt-per-square-silicon advantage. I would like you to think about what that means.

brain... hurts...
what a sidlly metric... oh, god. what woul happen if amd also loses this one?

performance-per-watt-per-square-silicon-per-citizen-who-actually-cares.

Anonymous said...

per-square-silicon advantage

They really have lost it.

Anonymous said...

You are an asshole!

Unknown said...

nah... they goes back to having the best "performance".. using bogus benchmarks.

http://www.amd.com/us-en/Processors/
ProductInformation/0,,30_118_9485_13
041%5E13077%5E14820,00.html

if you click on it, which shows the setup of the machines, you'll notice something at the bottom line...

AMD MAKES NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE CONTENTS HEREOF AND ASSUMES NO RESPONSIBILITY FOR ANY INACCURACIES, ERRORS OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION.

this means that even if someone proved them wrong, AMD is not liable for the incorrect number.

what a great way to market your processor.

Anonymous said...

you're brain dead

Anonymous said...

hahahahhah AMD fanboi's only defense to their company's incompetence is to call people names for pointing out the truth. Another great and TRUTHFUL article by the real PHD roborat. Keep up the great work.

AMD BK Q2 08

Anonymous said...

"real PHD roborat"

You are all so stupid, Roborat is the stupidisc!!!

I will not write anything on this blog, because is a piece of shit and nobody understands nothing.

AMD will kill Intel!

Do you know what means 2,5GHz and 40*70% faster against 3ghz??

You are all soooo stupid, I just can't believe.

=) ha, ha ha,

"real PHD roborat"

he is an idiot, like all of you!!!

Unknown said...

Wow, the anons are out today :).

Anonymous said...

Stupid is the company that gives up a year's worth of market share over Christmas.

Randy Allen said...

Excellent post! AMD may well be as fast as they claim at the same clockspeed. If we add up the performance gains they believe to be true, 50% in FP, 20% in Integer. We arrive at ~35% more efficient than Clovertown, at the same frequency.

But, Penryn will be fighting Barcelona, not Clovertown alone. Intel has shown off a few benchmarks, I think it's fair to say that an average IPC increase of 10% is about accurate. So if we accept AMD's claims and valid and true, there we have it: Barcelona will be 25% faster at the same clockspeed than Penryn.

So a 2.5Ghz (the fastest at launch) Barecelona will be (roughly) as fast as a 3Ghz Penryn, and would be slightly faster than a 3Ghz Clovertown.

Will Barcelona close the gap, and ensure that AMD and Intel are about equal on performance in servers? Yes. Will it be 200% faster like Sharikou claims? Not a chance.

Roborat, Ph.D said...

Roborat is the "stupidisc"

can i claim copyright on this? good name of a volatile memory disc.

Anonymous said...

You forgot to mention the marketing plan - "future proofing platformance"

You know like all of those people that bought those 939 socket Athlons that were 64bit compatible - where would they be right now if those chips WEREN'T 64bit compatible? There's barely any SW that still runs 32bit today?!?!?

Think about it and now look at all of those people that fell hook line and sinker for the intial 4x4. Yes they, will get to reuse there HW if they wish and only throw away ~$1000 worth of chips and get to use AMD's latest tech with some restrictions on it thank to not quite complete compatibility. And they're probably want to upgrade memory too. If people bought this, thinking they were going to upgrade to Barcy in 1 year, well they got what they deserved - an alpha solution that is less than optimal even after they put the Barcy chips in. In all likelihood AMD will be releasing new boards and chipsets when K10 DT chips come out which will take care of a lot of issue and those enthusiats who bought the initial solution will end up buying an all new solution.

People who try to future-proof they're purchase generally end up paying for more than what they really need and then get a solution which they will end up throing out at the end of it's lifecycle instead of upgrading it. (BTW - the same thing hold for Intel chips/systems as well).

Best bet for the average consumer (not necessarily server though) is to buy a system with only the features you need now or can foresee needing in ~12 months time horizon. Anything else is "blue crystals".

Rant over...

Anonymous said...

New metric - silicon area x performance per watt x IPC per technology node.

As AMD chips are bigger than Intel's, they have greater value.

Intel's higher clocks can be ignored because it is not fair.

45nm products cannot be compared to 65nm products as that is also not fair.

Thus you can now only compare similar clocks on similar technology nodes, REGARDLESS of whether there products that are higher clocked and/or exist on more advanced tech nodes.

Oh and Intel needs to cut down on the amount of cache too as this is only a bandaid...but L3 cache can be counted as that is not a bandaid because AMD does it.

With all of these factors taken into consideration, it should be clear that AMD will retake the crown.

All kidding aside, I think K10 will be a good design, will likely outperform Intel's 65nm products slightly on INT and more on FP and will be roughly on par with initial 45nm Penryn. The real question is whether AMD will be able to scale clocks on K10, either on 65nm (which doesn't seem to be the case as there doesn't appear to be much clock improvements between now and Q2'08) or on 45nm which will also be difficult early on until AMD scales the gate oxide (highK/metal gate)

I also do believe AMD may fare better on the power end (for given tech node) due to the deisgn of Barcy - though this will be likely offset by Intels 45nm process (until AMD gets K10 to 45nm)

Unfortunately AMD desperately needs this to be superior to Intel's 65nm and early 45nm products to allow them to justify pricing these higher and getting back to profitability - otherwise they will simply just start to replace the K8 SKU's and price points and AMD will remain under financial pressure.

Anonymous said...

What's up to all, since I am in fact keen of reading this website'ѕ post
to be updated regularly. It includes plеаѕant informаtion.



Аlso visit my ωebpage empower network

Anonymous said...

Whеn Ι oгiginally lеft a comment
I appeaг tο haѵe clіcκed the -Nоtіfy
me when new commеnts aге added- сhеcκbох and noω ωhenever а comment іs addeԁ I gеt 4 еmails with thе same cοmment.
Ρerhaps therе is аn eаsy mеthod уou can
гemоvе me frοm that
sеrvіcе? Thanks a lot!

my wеb pagе: green coffee

Anonymous said...

Ι simply сoulԁn't leave your website before suggesting that I extremely loved the usual information a person provide to your visitors? Is gonna be again regularly in order to check up on new posts

Look at my homepage ... green coffee