AMD Responds to TechPowerUp Investigation, Issues Remedy

Friday, October 24 2008

Earlier this week, when evaluating samples of the new Radeon HD 4830 GPU-based graphics card, W1zzard, who also authors the GPU-Z diagnostic utility, had found an anomaly with the stream processor (SP) counts on samples sent by AMD. The GPU could access only 560 out of its 640 SPs available. This affected the card's performance significantly, and TechPowerUp's findings were validated by several other reviewers with similar samples.

AMD on its part quickly followed up the issue with its engineering department, and released a video BIOS update that fixes the issue and makes available all the 640 SPs. AMD will be circulating this BIOS to all its press contacts and add-in-board (AIB) partners, to make sure the issue is isolated and fixed. The updated BIOS file can be downloaded from here. AMD also asserts that the issue isn't hardware related, and that updating the BIOS resolves the issue completely.

posted by btarunr - 4:03 AM | Related News |

User comments

by JrRacinFan (October 25th - 4:10 AM) - Reply

Great to hear this! I was about to say there are going to be alot of people really P-O'ed about it.

by Fitseries3 (October 25th - 4:27 AM) - Reply

our reviewer W1zzard? more like TPU admin/owner.

glad to see we got this going.

by driver66 (October 25th - 4:34 AM) - Reply

IMO the only reason this was brought out was because of the TPU review.
It's called accountability!!! AMD tried to pull an nvidia and got caught. If it were never found out do you think that they would have fixed it? "The fix would have been a new 4835 with 700 SPs" with 640 actual!!!

Ponder :toast:

by JC316 (October 25th - 4:36 AM) - Reply

Way to go W1Z! Catch AMD maybe trying to pull the wool over our eyes.

by marsey99 (October 25th - 4:36 AM) - Reply

more like tinternet gpu god :)

worshipping aside does this mean we could mod the bios of these 4830's and turn them into 4870's?

if its only soft modded to disable the cores/pipes w/e its gotta be doable right?

by MarcusTaz (October 25th - 4:41 AM) - Reply

Doubt AMD was trying to pull a fast one or anything of the sort. The card was meant to beat out the 9800/8800 segment. As it was it would not have been much of a matchup with a defective neutered bios.

W1z great job for helping AMD/ATI solve an issue that could have gotten a lot worse!

by btarunr (October 25th - 4:42 AM) - Reply

by: fitseries3
our reviewer W1zzard? more like TPU admin/owner.

glad to see we got this going.

We know who he is, but when in news, in context of the incident, he's our reviewer, and GPU-Z author. It relates better to the guy at AMD/non-TPU member reading the news.

by p_o_s_pc (October 25th - 4:45 AM) - Reply

thats a fast response by amd but i like to see that.

by Fitseries3 (October 25th - 4:55 AM) - Reply

by: btarunr
We know who he is, but when in news, in context of the incident, he's our reviewer, and GPU-Z author. It relates better to the guy at AMD/non-TPU member reading the news.

oh come on... i was just messin with ya.

by jbunch07 (October 25th - 5:01 AM) - Reply

So this only makes me question if a bios release enabled 640 could another (modded) bios release even more?

by driver66 (October 25th - 5:02 AM) - Reply

by: jbunch07
So this only makes me question if a bios release enabled 640 could another (modded) bios release even more?

exactly.......see previous statement :roll:

by marsey99 (October 25th - 5:03 AM) - Reply

or mine :)

by Solaris17 (October 25th - 5:05 AM) - Reply

I dont think so....the rest of the SP's according to w1zz aparently hard locked kinda like laser locking not the same but same principle...however lowing the SP's is possible because you can tell the bios not to map them. this is post 6666

by newtekie1 (October 25th - 5:06 AM) - Reply

But did any of these cards make it to retail channels?

If ATi was trying to pull a fast one, I would think they would have at least sent out real cards to the reviewers so they reported good performance.

by jbunch07 (October 25th - 5:08 AM) - Reply

by: Solaris17
I dont think so....the rest of the SP's according to w1zz aparently hard locked kinda like laser locking not the same but same principle...however lowing the SP's is possible because you can tell the bios not to map them.

you just posted you 6,666th post. :eek:

But OT, that's a bummer, Oh well, im sure the 640sp will be plenty to fill the gap!

by Solaris17 (October 25th - 5:08 AM) - Reply

by: newtekie1
But did any of these cards make it to retail channels?

If ATi was trying to pull a fast one, I would think they would have at least sent out real cards to the reviewers so they reported good performance.

well you can be sure none will be made as AMD has issued the fix to its partners however if some batches were already shipped their is a chance of contamination unless the partners do a mass recall to check the cards.

by roofsniper (October 25th - 5:25 AM) - Reply

its just a bios issue whats the point of amd trying to give people less sps? they are on the card which means it costs amd the same. no matter what it hurts amd because the reviews are bad and the people get less performance than they expected and theres no point in that.

by MarcusTaz (October 25th - 5:40 AM) - Reply

by: roofsniper
its just a bios issue whats the point of amd trying to give people less sps? they are on the card which means it costs amd the same. no matter what it hurts amd because the reviews are bad and the people get less performance than they expected and theres no point in that.

Exactly, what is the point, it is pointless and a silly presumption...

by PCpraiser100 (October 25th - 6:02 AM) - Reply

It feels great to be part of a solution to AMD. Thanks for the notice!

by WarEagleAU (October 25th - 6:18 AM) - Reply

Nice to see AMD respond so fast. I highly doubt they were trying to dupe anyone. Everyone has issues and mistakes come up and not all are always caught.

by eidairaman1 (October 25th - 6:52 AM) - Reply

So Wizzard are you going to Re-review the boards since they are now fixed to see if there is any fluctuation in the performance analysis?

by theJesus (October 25th - 7:30 AM) - Reply

Like others have said, it's really nice to see a company respond to an issue like this so quickly and efficiently :)

@eidairaman1: I don't really see any reason to re-review the amd card any further than verifying the number of SP's enabled and maybe running a couple quick benches, since the other card had the full 640 SP's enabled and we have the performance results from that.

by eidairaman1 (October 25th - 7:38 AM) - Reply

well its on the clause just to make sure, i mean wizzard can run the cards and let us know the results, if they deviate from the originals then something else maybe amiss.

by theJesus (October 25th - 7:51 AM) - Reply

Right, that's why I said just verify the SP count and run a couple quick benches, but I think as long as those benches yield results on par with the other card then it would be relatively safe to assume that it's ok. I'm sure W1z is a pretty busy guy ;)

by ulrik (October 25th - 9:22 AM) - Reply

Question: Is there already a discussion going on about possible energy savings by switching off a shader block? Since in techpowerup and computerbase.de test the unpatched amd radeon 4830 (having 560 shader count) seems to require less power than the powercolor 4830 having 640 shaders.

Could somebody try this out, please? Maybe this could be a new strategy for 4850/70 energy savings/Powerplay? Another interesting way would be to try rivatuner shader block bit masking.What's the real difference in the old and new amd bios?

Post your own comment

Username:
Password:
Register Here!