Conclusions from 180 series stability poll (so far)

[TomCooksey]

I've just swapped out my 7300 LE with a 6600 LE card instead and my system hasn't frozen yet. So, I've taken everyone's results (thanks!) and it _seems_ there's a bit of correlation between the graphics core and the stability of your system.

Stable cores are:

• MCP61P
• NV43
• G72M
• G71
• G86
• G84
• G80
• G94
• G92a (?)

Unstable cores are:

• G72
• G71M
• G84GL
• G92b

There are two anomalies to the pattern:

1) The 8200 has been reported as both stable and unstable - however it's a motherboard chip and the actual core number isn't known (so I guess it's possible it changes from chipset to chipset).

2) The G92b is reported as unstable, however the 9800 GT (which may be a G92a or G92b) is reported as stable. I guess the stable chip might be a G92a whereas the G92b is unstable

[Windseeker]

thank you very much for your great efforts. you made what nvidia devs still missed. hope they at least use these results for further investigation. for myself, i will use your list to raise my chances getting a stable system, because I think about upgrading the graphics board.

[mfb]

They probably already know all of this by bugreports, which include all the information they need.

For what it's worth:
My 9600M GT (G96-core) is unstable.

[zbiggy]

Originally Posted by TomCooksey

1) The 8200 has been reported as both stable and unstable - however it's a motherboard chip and the actual core number isn't known (so I guess it's possible it changes from chipset to chipset).

The actual core number is known from lspci:
00:12.0 PCI bridge: nVidia Corporation MCP78S [GeForce 8200] PCI Express Bridge (rev a1)

8200/8300 is the same chipset. 8300 is just overclocked version. 8100 is stripped down gfx core which is different a little.

There might be some bigger differences in gfx core between this chipset on AMD and Intel mainboards. For Intel is Geforce 9300/9400 integrated gpu, for AMD 8300/8200/8100. But these are still differences in terms of numbers of: MHz, shaders, bus wide bits, memory size etc.

I heard Intel Geforces have MCP79 or MCP7A codename.

[TomCooksey]

Looking at newer reports, it seems that there's more cores which can be both stable and unstable - so my conclusion doesn't hold at all anymore.

Ah well.

[quetzyg]

Originally Posted by mfb

They probably already know all of this by bugreports, which include all the information they need.

For what it's worth:
My 9600M GT (G96-core) is unstable.

I've got that chipset on my laptop, and it never gave me problems.

[marchetto]

What I would try to say is that you can't have any kind of conclusion like: "this core works" "this core doesn't work".

G96 of mfb is not stable.

Stability is not just a graphic core matter, not only. You're trying to simplify the problem, but I think this is a really wrong way. If problem is complex you have to analyze in every part, not saying ok this group works this other not.

You're giving false information these were my 2 cents, we are on a forum for discuss about things like that, if you open a thread with some (false in my opinion) informations, I've to say that.
Nvidia driver development doesn't need of false informations, has is own troubles, it doesn't need more and doesn't need false positive feedback.
Nvidia driver is on worst level I've ever experienced after years of heavy use, so the situation is not good, and I think needs of serious solving thread, where you say which is your problem and you try to solve one-by-one.

So, if this on this thread wants to be a serious analysis, I think it's on a wrong way, if you want to chat and poll about nvidia driver that's another level of discussion, but I don't think you can present it as a serious technical analysis.

At all, there's ONE interesting thing in that poll, and I think it's the only interesting number, 30% says they're working on an UNstable nvidia driver.

[castor]

I disagree. With large enough material this can have some statistical relevance (not with the current amount of responses though).
Of course, like with all statistics, the burden of interpretation is on the reader

But instead of statements like "this works, this doesn't", I'd like to see the actual figures per CPU (Gxy 40% stable, 60% unstable).

[marchetto]

A stat result of 60%-40% success-unsuccess is not "this work" "this doesn't work" that's really different.
Beside this, we can discuss about how reliable is a stat about a 9500gt where 10-15 users answer the poll. Maybe today they're stable, tomorrow they're not.

So there's a final point, and is that you can easy realize that any stats in this scenario will never be a (nowhere near) reliable stat.