Athlon 64 and the "tainted benchmark"

Have you read that athlon 64 benchmark? What is it with people and wanting to TAINT THE F****** RESULTS!?!? Okay..Why must they manipulate with the systems in ways, when they're trying to show something, Why don't they show them in their best light? F*** no...they're showing off tainted BULLSHIT.

My problems with the article:

Athlon XP running PC2700 DDR memory. Though they did not specify 'barton', I would imagine it's an original XP core(266Mhz FSB) since they had lowered the clock of the device to 1.2Ghz.

Another problem is the fact that they had lowered that clock. There is no Athlon XP at 1.2Ghz. No matter what one may say, the result there is that they had tampered with the clocks/speed of the hardware and therefore the results should be null and void(they should've used 1.33Ghz Athlon XP).

The Pentium 4 running on PC2100 DDR memory. This is absurd! We all know very well that running anything less than either dual channel PC2100 DDR memory or PC1066 RDRAM severely cripples the pentium 4 in terms of what it can support. That's horrible that they did that!

But go figure..this is what happens when they want to taint reviews... [Name, email address supplied]

----> Hi,

If one takes a good look at the article, the reasoning behind the specific choice of the systems would be quite clear. The authors intended to compare the AXP and the A64 on a clock for clock basis, running 32bit software. As such, they needed equally clocked cpu's and their Hammer sample was running 1.2ghz, thus they needed an AthlonXP running 1.2ghz (9*133mhz). If they would run the proposed comparison instead, the AthlonXP would have a 133mhz clock speed advantage and any results derived would be skewed. Certainly if you only work with percentage points since 'real' benchmark numbers aren't allowed by AMD (at least not before the NDA expires or using only engineering samples).

As for the choice of the Palomino instead of the Barton and the 266 FSB speed: the Barton is simply not available yet and there's no need for a higher bus speed at 1.2ghz. Also, you have to keep in mind that the Hammer will never be released at 1.2ghz; this is simply an engineering sample, so what if there never was an AthlonXP @1.2ghz? It doesn't mean jack schitt (http://users.skynet.be/TailwinD/jackschitt.swf)

This speed issue also rings true for the Pentium 4, while pairing PC2100 with a Pentium 4 will lower the performance somewhat, this is nearly not as severe as the writer would let you believe. The Pentium at this clock speed (2-2.2ghz) just doesn't need as much bandwidth as a 3.06ghz one with HT enabled. Also, almost all Intel systems released about a year ago (around the release time of these cpu's into the mainstream) carried PC2100. As such, this is the performance that most users will see using these processors. My guess is the authors used some standard Dell box they had lying around the office. Said writer should be happy they didn't choose to run a clock for clock comparison with the P4 (which would have been trashed by both Athlon's - even with PC1066 RIMMs).

I do have some comments of my own however. The benchmarks were performed with a Sledgehammer (1MB L2 cache), so it should be compared against a Xeon and not the plain jane P4. The Clawhammer will only have 256kb L2 cache for all we know. And since when does the INQUIRER sign a user letter, instead of the usual "Name, email address supplied”? [When we make a mistake, Ed.]

Email name, address supplied

---> Lettre Troisieme Whoever wrote in criticizing c't's benchmarks is rather missing the point. Even if c't's test don't reflect the most modern level of hardware available, each of their configurations is balanced to demonstrate a valid level of performance. Had the P4 been running on i845 SDRAM, that'd be one thing, but many P4 systems today DO sell with DDR2100.

Furthermore, when you consider that the CPU c't is testing is a prototype, its obvious that the point of their article is to give an indication of what Hammer performance might look like, NOT a look at the final product itself. We have no guarantee that the CPU in their hands represents final silicon, even final silicon running at a lower clockspeed. Its entirely possible that while the CPU is fully functional, it could have key functions disabled in hardware for compatibility testing or any number of other reasons. The bottom line is, when you are working with prototype hardware, you never know what you've really got.

One last point. There absolutely WAS an AthlonXP-equivalent CPU made at 1.2 GHz. We called it the Athlon MP back then, but it was a Palomino-core chip running at 1.2 GHz--and appeared months before AthlonXP launched. I'd suggest the writer of your first letter brush up on his benchmarking and CPU-history before he writes in again, as spewing such voluminous idiocies ultimately serves only to leave other readers with a clear view of his stupidity.

Name, email address supplied µ [This correspondence is now at a close, Ed.]