Looks like AMD got caught with its hand in the cookie jar again. First it changes usage from TDP to ACP to hide its power usage, then it doesn't tell the whole story about the ACP.
On the other hand, it does look like clever marketing to those who don't know better and/or are too lazy to investigate. To people buying their computers at Best Buy, they probably won't know and wouldn't understand if you tried to explain it to them.
---------------
Evil lurks in the databanks as it lurked in the streets of yesteryear. But it was never the streets that were evil.
I did a little snooping on the net, and it looks like AMD purposely deflate its own TDP rating, for marketing purposes.
For Phenom 9600, it has a TDP of 95W, which is identical to Q6600 G0 revision. So theoretically speaking, it should consume the same power as Q6600, right?
Here's TechReport's review on Phenom.
Doesn't look too "identical" to me. So either AMD is grossly understating its TDP, or Intel is grossly overstating its TDP. Which one does it seem more likely?
(1) Intel publishes balls to the wall max.
(2) AMD publishes based on anticipated workloads.
(3) This has been known for some time, and wasn't a problem before.
(4) It's going into the holiday season, the writer needed a story, but didn't want to work that hard.
Am I right? Am I wrong? Should I have just asked for bets??
--------------- The worst part of my Vista 64 experience is having to listen to all of the individuals who apparently feel it's their 'Grand Mission In Life' to tell me about all of the things that (supposedly) don't work, when it *does* work.
In the Desktop arena Intel only lists 65W, 75W, 95W, 105W and 130W. Looks to soon be joined by 150W for the Xeon QX9775 for the Skulltrail motherboard.
Intel has yet to come out with any processor listed below 65W DTP. I'm sure if they wanted too they could come out with a 45W DTP especially for any of the new Penryn based Dual Cores.
(1) Intel publishes balls to the wall max.
(2) AMD publishes based on anticipated workloads.
(3) This has been known for some time, and wasn't a problem before.
(4) It's going into the holiday season, the writer needed a story, but didn't want to work that hard.
Am I right? Am I wrong? Should I have just asked for bets??
I rarely side with AMD, but to a certain extent I must hear.
The writers assumptions were way off, even when looking just at Intel.
The E2140 and the E6750 have the same TDP.
However, the power they use at stock and clocked to the same speed is quite a bit different.
The E2xxx series uses far less power due to less cache.
Clearly you cant relate TDP or the Avg power consumption to closely.
For Intel, the TDP is clearly far more for an entire class of processors and even the highest of the high's max TDP even under extreme OC's.
The TDP for the Latest AMD chips appear to be closer to the max for the each individual chips.
AMD lies, its mistake or someones faults finding? The find is interesting, but writer should have gone all the way to prove or disprove this theory (like getting the cpu's and testing them!), now its juts pointing out few figure changes without understanding why it was done and immediately publicly call "AMD lies!".
Did they asked AMD to clarify? No.
Did they tested and saw for themselves? No.
Do they know ACP is tested in the wrong way and this article proves it? No.
Its unprofessional.
To quote discussion (by Sahrin) in DT:
"DT: There is no other feasible explanation of why a 20 W TDP increase would be accompanied by no increase in ACP.
Bull. AMD has historically defined TDP's for entire models in ONE number. This is why all parts of type X will be released at 95W, or 105, or 135W - when there's no way in heck the lowest-end part has the same power usage as the highest-end part. Every part gets released in the same envelope, because AMD plays the power game conservatively.
Maybe, they increased the TDP figure to raise the power envelope? Maybe it's to make room in the spec for FUTURE parts, that haven't been released? This would mean that the TDP would change - but the ACP wouldn't because the the ACP is based on ACTUAL POWER PERFORMANCE - not a design spec, like TDP. TDP is the maximum envelope in which any chip may run - ACP is the power that the typical user can expect to see. The only difference between AMD and Intel on this score is that Intel was already telling this lie.
Furthermore - if you can't conceive of a change in maximum power not affecting the average, then you've never sat in a statistics class before.
I won't deny it's possible DT may be right here - but 10 seconds of testing would verify this information. It's so circumstantial and anecdotal as to make even the most ignorant of deductive reasoning squirm.
This is FUD, from Dailytech itself. Congratulations, DT - you're today's Elmer.
So if you're a corporate customer planning to buy many thousands of these chips, your thinking is to disregard AMD's material specifically designed for this purpose, instead opting to do validation yourself?
No, I'm going to look for hard information before coming to a misinformed conclusion. First I'll apply fundamental understanding of the concepts (AMD's definition of TDP v. AMD's definition of ACP v. Intel's definition of TDP v. a middle school-level understanding of statistics) and then go out a see if others (those that get these documents and publicize their opinions of them (READ: Hardware Review Sites) have had the opportunity to test the chips against the information (like when AMD claimed they were seeing 40% performance improvement with K10 - it was unclear to what exactly they were comparing K8, so DT (or someone, I believe it was you guys) took an engineering sample and tested their claims - and it turned out they were wrong, or that the claim was being misinterpereted in AMD's favor by the community at large) to see if there is a discrepancy.
Circumstantial would indicate this is some corner case: it's not. Anecdotal would indicate there is some testing going on period. There is not.
Circumstantial = derived from circumstance. You have no basis in fact (that is, no empirical evidence, just a document from AMD that you are inductively using as a basis for your conclusion; without the circumstance of the document you would have no case - imperfect usage, I grant you, but not fundamentally incorrect).
Anecdotal = based on personal observation, case study reports, or random investigations rather than systematic scientific evaluation. That exactly describes the point I'm trying to make - that DailyTech is using a piece of evidence that is questionable, a synthesis that is questionable, and no additional evidence beyond it. It's like reading Fudzilla, only with a better layout.
These are white papers produced for corporate clients. If there is a large mathematical / logical mistake, that's not ancedotal or circumstantial.
If there's anything other than exactly what DT's implying, then it's absolutely anecdotal and circumstantial.
When DailyTech manages to misspell Ballmer's first name on the front page and someone raises issue, that's not circumstantial or anecdotal at all. It's in error, just like these whitepapers.
It is - but I don't think it merits the full force and fury of a front-page DT article alleging that a (relatively) upstanding company is "lying" to its customers. Certainly, when DT makes a spelling error there are no front-page headlines saying "Kubicki is t3h stooopid" and then tying the misspelling inductively to that rather poorly formed conclusion.
---------------
E6600 (B3)-HD4850 512MB GDDR3-2GB DDR2-800
"You figured it out. All new CPU's are nothing but overclocked Pentium 1's with a few bells and whistles added, ask any ol timer whose been around."
Bull. AMD has historically defined TDP's for entire models in ONE number. This is why all parts of type X will be released at 95W, or 105, or 135W - when there's no way in heck the lowest-end part has the same power usage as the highest-end part. Every part gets released in the same envelope, because AMD plays the power game conservatively.
Far too conservatively seeing that Intel's faster processors absolutely dominate all AMD processors in power use.
Maybe, they increased the TDP figure to raise the power envelope? Maybe it's to make room in the spec for FUTURE parts, that haven't been released? This would mean that the TDP would change - but the ACP wouldn't because the the ACP is based on ACTUAL POWER PERFORMANCE - not a design spec, like TDP. TDP is the maximum envelope in which any chip may run - ACP is the power that the typical user can expect to see. The only difference between AMD and Intel on this score is that Intel was already telling this lie.
The ACP almost as useful as the 400w PSU recommendation from my graphics card. I wouldn't put a heat sink rated to 40W on a processor that uses an average of 40W; I would find the maximum thermal dissipation from the CPU and use an appropriately rated HSF.
Message edited by jbj190 on 12-10-2007 at 11:44:59 PM
That FX-74 must be uber-powerful to consume that much power
Correct me if I am wrong but isn't the FX-64 a chip for the Quad FX set up? And it consumes 2x the power as the Q6600 idle and load and is not faster.
As for the Phenom being rated at 95w, with AMDs new ACP(love how they decided to change from TDP to ACP) it is made to make them look better even though they are not as efficient as a Q6600 yet alone a QX9650(idle and load)
I think Hector decided on ACP. Its funny. I miss the AMD that had a good product cuz what they said was true instead of being lies.
I just hate how many lies have come and more and more. Lies = baddddd.....
First it changes usage from TDP to ACP to hide its power usage, then it doesn't tell the whole story about the ACP.
The find is interesting, but writer should have gone all the way to prove or disprove this theory (like getting the cpu's and testing them!), now its juts pointing out few figure changes without understanding why it was done and immediately publicly call "AMD lies!".
like getting the cpu's and testing them!)
