AMD's Rivas downplays Phenom TLB performance hit

— 2:31 PM on December 11, 2007

ChannelWeb has taken some time to sit down with AMD Computing Product Group Executive VP Mario Rivas for an interview about AMD's quad-core processor launch and, more specifically, the TLB erratum that made headlines last week. In the interview, Rivas admits that the quad-core Opteron and Phenom rollout could have gone better, although he suggests AMD's stumbles can be explained by the fact that K10 is "a very complex, maybe the most complex x86 processor ever designed." However, Rivas then attempts to downplay the performance hit induced by the BIOS fix on Phenom processors:

Phenom is shipping now, and it's got the workaround in place. To what level does that workaround affect performance on Phenom specifically, and will future editions of Phenom be corrected at the design level in the same manner that Barcelona will be?

I'll answer the second part of your question first. Yes, it will be corrected. When we launched the devices rated at a particular speed, that already includes the degradation. So that's why you saw those lower clock-speed parts first, with future speeds to come. So you don't have to go and say, 'Well, jeez, if you have a Phenom running at 2.0 GHz, then I have to take 5 to 20 percent of performance off it.' No, I'm already doing that. And you have to remember, too, that the Phenom parts that we launched were really targeted at the mainstream computer user and not the enthusiast guys. So the mainstream computer user really isn't going to notice any impact. It all depends on the workload, obviously. But to touch the workloads they're running, they aren't going to really notice this.

This statement is ambiguous, but as far as we can tell, erroneous. The Phenoms that were benchmarked at the product's introduction didn't have the fix applied, and neither did the Phenom-based systems at AMD's press event prior to the Phenom launch. Our own benchmarks of the TLB erratum patch (which are mentioned earlier in the interview) suggest an average performance difference of 13.9% with the TLB fix applied. The differences with with Mozilla Firefox and picCOLOR are 57.1% and 29.9%, respectively—well outside of the 5-20% range Rivas mentions.

Rivas also asserts that existing Phenoms aren't really targeted at enthusiasts, even though AMD's own website bills the Spider platform (which includes Phenoms, 790FX chipsets, and Radeon HD 3800 graphics) as enthusiast hardware. When confronted with this fact by ChannelWeb, Rivas explains, "when you look at Spider, and what we're trying to tell the world, is that performance is not limited to the processor. We're looking at a platform and the fact that we have, the technical term is 'kick ass graphics.'"

Moving on, Rivas discusses the schedule for parts without the bug. He mentions that AMD will receive samples in January, but that those samples must go through a "very rigorous verification process" then be sent to customers for further validation before they can hit the market. AMD's Desktop Product Marketing Manager, Michael Saucier, quoted a "mid to late Q1" schedule for availability when he talked to us last week. (Thanks to TR reader Doug for the tip.)

Update: ChannelWeb has updated the interview to remove the erroneous statements in Rivas' aforementioned reply. Sentences between "Yes, it will be corrected" and "You have to remember . . ." are gone, and ChannelWeb has added an asterisk that says, "Updated Dec. 11 at 6:40 PM EST to remove a factual misstatement."

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.