AMD Exec Discusses Barcelona Debacle

A senior executive at Advanced Micro Devices said Monday that launching the first quad-core Opteron server chip to great fanfare in September was "a stupid decision" given a glitch that has halted distribution to all but a few customers, but insisted that it was "the right decision" at the time.

"Now if I was to do it all over again, I have to tell you, with the data I had at the time, I would make the same decision again. With the data I have now, clearly, that was a stupid decision. But with the data I had that day, it was the right decision," said Mario Rivas, head of AMD's Computing Products Group.

In an interview with ChannelWeb, Rivas said that AMD had stopped shipment of quad-core Opterons to all but a few customers due to a major glitch discovered on the chips. He claimed AMD's system builder channel was being treated no differently than its tier one partners in this regard, saying, "It's unfortunate that the problems we are causing are coming across like we don't care about the channel. ... But no, we are definitely committed [to the channel]."

Rivas described the chronology of the chipmaker's discovery through testing of the data-corrupting bug on its Opteron and Phenom chips that would require workarounds and ultimately a design correction.

"It was not until mid-November that it actually moved from just being an observation in a testing environment to being a more serious bug. We tried to do BIOS workarounds, we looked for board modifications, even in some instances, for some patches we could do that would not degrade performance," he said.

When AMD came to the decision that the glitch could "affect a real-world application," Rivas said the Sunnyvale, Calif.-based chipmaker quickly alerted customers.

"When we reached that point, it was a Friday and we started notifying the customers on Monday," he said.

AMD developed a workaround for a bug that causes data integrity problems related to the translation lookaside buffer (TLB) on quad-core Opteron and Phenom CPUs. But performance degradation caused by the workaround on the Opteron server-workstation processors was such that only a few customers decided to go ahead with orders.

Both the Opteron and the desktop Phenom chips can suffer 5 percent to 20 percent performance hits on some workloads as a result of the BIOS fix. But Rivas clarified a point about the quad-core Phenoms, which were launched in November. Those processors have been shipped with the workaround in place, and at the ramp-up AMD planned. Enthusiasts, however, have the option to disable the fix in the BIOS or in AMD's overclocking tool, OverDrive. AMD's quad-core Opteron, commonly known by its codename Barcelona, will get a design correction before it is made available in volume to the channel and tier ones, Rivas said. The next batch of Phenoms will also be partially redesigned.

According to Rivas, samples of the design-corrected Opterons will be available in January. He described a rigorous validation process with partners before volume shipments of the fixed chips would take place sometime in the first quarter. Asked if that kind of cautious approach ought to have been taken with the launch in the first place, Rivas made his statement about the launch being the "right decision" with the data at hand.

Rivas would not comment directly on a question about possible changes to personnel at AMD as a result of the Opteron situation, saying only, "That's up to the board of directors. We serve at their pleasure."

At no time, Rivas said, did AMD consider shipping Opterons with no warnings once the glitch was discovered.

"Once we determined that this problem could be found in a real situation, we said, 'We need to tell our partners.' At no point did we say, let's hide this and hope nobody finds out," he said.

Updated Dec. 11 at 6:40 PM EST