THERE HAS BEEN a lot of talk recently about ATI’s ‘Grey Screen of Death’ and almost all of it is hysterically overblown. Let’s take a look at the problem in a bit more detail.
The short story is that a number of people have reported crashes with various ATI 5-series cards where the screen goes grey or striped in appearance. Some have gone to great lengths to quantify the problem, and a few threads have proposed solutions of varying effectiveness.
The GSoD in question is only on GDDR5 equipped 58xx and 57xx cards, and only on Windows 7. The problem first started when Microsoft put out an update to Win 7, although the exact patch and what it affected is not clear. One problem is that the GSoD problem is a specific problem, but there are other issues that may mimic the same fault, and striped screen crashes have been around for ages.
Talking to OEMs, AIBs and a few ATI people over the past couple of weeks, a few patterns became evident. First is that the number of affected cards is very small. The largest forum threads have hundreds of posts, but of those, only a few hundred cards seem to be affected.
As of a few weeks ago, there were 2 million-plus Evergreen (5xxx) cards shipped. If there are 1,000 GSoD affected cards out there, that means 0.05 percent are ‘bad’. If there are 10,000 cards – more than the number of posts about it in all the forums I’ve read combined – that means half a percent. To put things in perspective, normal cards have a return rate that is 10 time that number. Every AIB I talked to said the numbers were not high enough to be concerning, so they are, on average, very low.
We could not get a definitive answer from anyone, but piecing together bits, it looks like the problem has to do with GDDR5 changing power states or speeds. The good news is that this is not a card killing problem, it is just a crash problem. That might not be soothing to you if your card keeps crashing, but there is hope.
We have word that there are two fixes coming – one is a software hotfix, and the other is a GPU BIOS update. The patch should work by just downloading and installing it, and it will be rolled into the next driver update. SemiAccurate was not able to get a definitive date on this, or whether it will make the Catalyst 10.2 release, but it shouldn’t be long in coming. The BIOS fix will be loaded onto cards as soon as it is done, making all parts from that date ‘fixed’ in the firmware. There is no word if that BIOS will be released as a user upgrade.
The workaround that people have proposed, upping voltages, disabling power savings, or nuking HDMI, all seem to be rather hit or miss. This is likely because they do not address the GDDR5 state change problem directly, just accidentally. What works for someone else might not work for you and your situation.
Short story, everyone tells me there is a fix coming really soon. ATI, unlike some other companies, is not trying to bury this problem. The aggregate numbers of cards affected is extremely small, far under one percent from what we hear. ATI’s 5-series GSoD’s have been turned into a feeding frenzy, mostly by people without technical understanding of what is going on, and by a few with an axe to grind.
The problem is specific to certain 58xx and 57xx cards and involves GDDR5 timings or state changes. It was brought on mainly by a change that Microsoft made deep within Windows 7, and it should be totally fixable with the impending patches. While it is annoying, it will be history very soon.S|A
Latest posts by Charlie Demerjian (see all)
- You can now buy bare Snapdragon SoCs with 410E and 600E - Sep 28, 2016
- Spin Transfer Technologies talks about their ST-MRAMs - Sep 27, 2016
- ARM adds CMN-600 interconnect and DMC-620 memory controllers - Sep 27, 2016
- Globalfoundries 7nm process isn’t even close to the name - Sep 26, 2016
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016