Over the weekend rumors popped up that nVidia was preparing to begin a full recall on all Kepler based products. The rumor cited a leak from TSMC stating that Kepler “chips may be suffering from serious performance degradation over long periods of heavy load”. This sounds almost exactly like what came out of “sources” at TSMC during the “bumpgate” issue that happened when the solder bumps in certain GPUs were failing due to degradation over multiple heat cycles. However, does this rumor have any truth to it? Or are we seeing a return to the days of Guerilla marketing like we did in the early 2000s.Although there are always issues with devices as complex as a GPU companies like nVidia and AMD work very hard to make sure they are working properly before they push them out the door. We know from the issue with the solder bumps that nVidia and TSMC changed the materials that were used in future products to prevent this from happening again. So where is this coming from?
One possibility (as we noted) is a guerilla marketing campaign by someone that has an interest in slowing nVidia sales. This type of tactic used to be very common as companies would seed rumors, comments and more around the internet to different journalists to see what would stick. In the Fermi days is was the use of a mock up for a demo when they showed off the card. With AMD’s launch of the 5 Series GPUs we got reports that their memory bus was having performance issues and that nVidia’s GPUS were still a better value. It was pretty ugly to be honest.
This could be what we are seeing here, but there is more to the story than first meets the eye. nVidia is having a problem keeping the GT 104 based cards on the shelves. This could indicate a poor yield in the manufacturing process at TSMC. Fermi had some of those issues when it was launched so we would not be surprised to see that happening again here. Another item that might be causing the lack of stock could also be that other parts of the BOM (Build of Materials) are in short supply. We have seen this before with HDDs when certain parts that made up the motors were not available after flooding in the areas that they are produced; so there are multiple reasons why the stock of these parts in short supply with manufacturing flaws only one (and a very unlikely one).
Where the story gets interesting is when people have been trying to link this issue with the recent recall of the GTX 670 SuperClock from EVGA. If you do not know about this here is the background.
EVGA sent out a message to Kyle Bennett at [H]ard|OCP and let him know that they (EVGA) were recalling a batch of their GTX 670 SuperClock Graphics cards (Part Number: 02G-P4-2672-KR). They said that the issue was with “an early batch of GTX 670 Superclock cards (P/N: 02G-P4-2672-KR) that were not properly screened during QA/QC procedure.” They were asking for owners of this card that were having any problems to return them and they would replace them with an upgraded card; the GTX 670 FTW (P/N: 02G-P4-2678-KR).
Now we contacted EVGA about this since the new rumor popped up and have a statement from them, but before we go there let’s think about this. If there was an issue with the GPU… If there was a massive hardware defect in all Kepler based GPUs that caused “serious performance degradation over long periods of heavy load “, why on earth would EVGA give them an upgraded card? The 670 FTW has a higher clockspeed than the SC (1006MHz Vs . 976MHz). You would think that they would want to avoid that extra load on the GPU if there were issues. No even before we received the word from EVGA we had to count this out as having anything to do with the rumors about Flaws in Kepler.
Still because we want to be thorough we did ask EVGA and they replied: “EVGA has isolated this problem to an early batch of GTX 670 Superclock cards (P/N: 02G-P4-2672-KR) that were not properly screened during QA/QC procedure. This particular Superclock card is the only affected model.”
They also went on to state that the reason for the recall on the GTX 670 SuperClock (P/N: 02G-P4-2672-KR) had nothing to do with the GPU.EVGA is not aware of any hardware issues with Kepler.
We have also reached out to nVidia, but at the time of this writing have not heard back from them directly. All we have is the single comment that has been reposted online “There is no truth to this” from Bryan Del Rizzo.
We are starting to lean toward this being a bad rumor and one that might have a limited impact on Kepler. The timing is interesting as they hit right when supplies were drying up; this could be an indication that the rumor is intended to get people to buy another GPU (one that is available) instead of waiting for more stock to reach the stores. We are going to have to call BS on this one until we hear something more concrete than what we currently have from the “leak” at TSMC.
** Update 5/21/2012-19:02 EST **
Bryan Del Rizzo has replied from nVidia with the following comment "There’s no truth to the rumor." It is looking even more like this is a false one to either hurt sales or drive traffic...
Discuss this in our Forum
