NV's answer to the 7900 Series' problem (OC'ing in general)

[lopri]

I'm noticing extremely aggressive throttling with the new Forceware 91.31. I don't even know what kind of logic they use for throttling. For instance I OC'ed my 6200TC from 350MHz to 390MHz, and temp goes barely over 44C. Still, according RivaTuna monitoring, the clock jumps to 390MHz upon 3D application launch, but soon (very soon), it clocks back down to 351MHz. No warning or notice any of sort, of course.

On the other hand, this card is good up to 450MHz with previous Forcewares. Is NV being more aggressive to hunt down OC'ing? Could someone else verify my finding?

If this is the case, BIOS flashing will be almost required for "stable" (lol) OC. Because no matter what you do (i.e. V-mod), the card will throttle down to its stock setting for no apprarent reason. I'd like other people's experiences.

Discuss.

 

[BassBomb]

are you saying it downclocks when still running the 3d program?

 

[lopri]

Yes. Not from its stock clock, but from whatever you set (higher than the stock, of course)

 

[dug777]

sneaky buggers they are if true it is

 

[Lord Banshee]

my StockOverclock evga 7800GT-CO does not throttle, but i do think my 3dmark06 score went up lol 4276. Max temp was 77*C

 

[nitromullet]

Originally posted by: Lord Banshee
my StockOverclock evga 7800GT-CO does not throttle, but i do think my 3dmark06 score went up lol 4276. Max temp was 77*C

It wouldn't because it's OC'ed via the BIOS. lopri is saying that the new driver throttles the card back to stock (BIOS) settings.

 

[Lord Banshee]

But Nvidia's issue was not people the overclock their 7900's it was the stock overclock version comming from evga, XFX, and others. Why would nvidia try to fix a problem to ones own fault, ie overclocking to ranges where artifacts appear and un-safe for cards, that is the person's problem no Nvidia's.

At least that is what i thought when the OP said "NV's answer to the 7900 Series' problem"

 

[beggerking]

Interesting... so it would be necessary to flash the bios to maintain overclock?

 

[nitromullet]

Originally posted by: Lord Banshee
But Nvidia's issue was not people the overclock their 7900's it was the stock overclock version comming from evga, XFX, and others. Why would nvidia try to fix a problem to ones own fault, ie overclocking to ranges where artifacts appear and un-safe for cards, that is the person's problem no Nvidia's.

At least that is what i thought when the OP said "NV's answer to the 7900 Series' problem"

From what I read in HOCP's article about the 7900GT issues, eVGA was claiming that most of the returns were from people who had OC'ed their card beyond the factory OC. If NV was concerned with the likes of eVGA or XFX OC'ing their parts too high, they would simply disallow them from doing it contractually. I'm quite certain that as the supplier of the chips, that are allowed to dictate some terms.

One reason they would try to fix it this way as opposed to limited the OC in the control panel is simple:

You would have a bunch of people on message boards talking about their monster OC's, but their board partners wouldn't have to deal with as many returns. Sounds like a win-win for everyone except the consumer.

 

[Lord Banshee]

Hmm i was under the impression that many many stock overclocked video cards were getting artifacts and thats what Nvidia was interested in.

link to story
http://www.xbitlabs.com/news/video/display/20060526165136.html

 

[nitromullet]

If what lopri discovered proves to be true, it would prevent this:

EVGA has observed some users pushing the card to an unreachable amazing clock speeds and causing the GPU and/or memory to fail permanently and this has gotten our attention.

http://enthusiast.hardocp.com/article.html?art=MTA2OSwyLCxoZW50aHVzaWFzdA==
Obviously, that doesn't mean that there isn't an issue with factory OC'ed cards, but at least eVGA, BFG, and XFX bite the bullet on their own OC's. If the consumber OC's their card and fries it they simply return it.

Anyway, the point isn't really the motive, the reason, or whatever, but rather the existence of the excessive throttling. I'm going to check this out later on today, as my 7950GX2 easily clocked to 550/700 using the "get optimum clocks" feature in the NV CP. I'll check this out with rivatuner when I get home.

 

[A554SS1N]

Bah, this is bad news...

 

[skooma]

Interesting. I'll check it out on my 7800gtx but does riva tuner conflict with the registry hack for coolbits at all?

 

[lopri]

It's certainly understandable from NV/partners' point of view. But for users who look for most bang/bucks, this wouldn't be too exciting. Basically the only way to tell throttling is either via a monitoring prog in the background, or unexpectedly low benchmark scores. Things will get complicated w/SLI, because, in a case of low performance, a user should judge whether it's because of thermal-throttling or the SLI isn't working properly.

 

[beggerking]

quote:

--------------------------------------------------------------------------------
EVGA has observed some users pushing the card to an unreachable amazing clock speeds and causing the GPU and/or memory to fail permanently and this has gotten our attention.
--------------------------------------------------------------------------------

I failed to see how overclocking would cause permanent damage to hardware...shouldn't there be things in hardware to prevent this from happening?

 

[nitromullet]

Originally posted by: beggerking
quote:

--------------------------------------------------------------------------------
EVGA has observed some users pushing the card to an unreachable amazing clock speeds and causing the GPU and/or memory to fail permanently and this has gotten our attention.
--------------------------------------------------------------------------------

I failed to see how overclocking would cause permanent damage to hardware...shouldn't there be things in hardware to prevent this from happening?

That prevention would be the throttling that we're talking about.

 

[skooma]

Well, I just installed riva tuner and ran 3d06 with 91.28 drivers. Does it save a text log somewhere or the entire graph?

stock- 450/1200
oc'd- 500/1300

Riva reported 540 peak then 450 then back to 275 (2d). I'm not sure why it reports 450 at all. Maybe it is throttling back to 450 and then going back up to 540 for the next test?

Let me try some Oblivion.

 

[beggerking]

lol. good point . haha. hardware isn't my strong area.

 

[Pens1566]

Originally posted by: nitromullet
If what lopri discovered proves to be true, it would prevent this:

EVGA has observed some users pushing the card to an unreachable amazing clock speeds and causing the GPU and/or memory to fail permanently and this has gotten our attention.

http://enthusiast.hardocp.com/article.html?art=MTA2OSwyLCxoZW50aHVzaWFzdA==
Obviously, that doesn't mean that there isn't an issue with factory OC'ed cards, but at least eVGA, BFG, and XFX bite the bullet on their own OC's. If the consumber OC's their card and fries it they simply return it.

Anyway, the point isn't really the motive, the reason, or whatever, but rather the existence of the excessive throttling. I'm going to check this out later on today, as my 7950GX2 easily clocked to 550/700 using the "get optimum clocks" feature in the NV CP. I'll check this out with rivatuner when I get home.

That EVGA info in the article may have been true at first, but for the last month or so no one is oc'ing their evga cards at all. Even at the EVGA stock speeds there are plenty of cards dying within a day or so. Anyway, all signs point to a problem with the memory system and not gpu oc'ing as the problems go away when the memory is clocked below nvidia reference specs.

 

[Modular]

Well I have an awesome solution to this if it is true:

Don't run the new drivers...

DUH!

 

[nitromullet]

Originally posted by: Modular
Well I have an awesome solution to this if it is true:

Don't run the new drivers...

DUH!

Not a possibility with a GX2, and most likely any subsequent new cards. Besides, should you have to do that?

 

[Munky]

Originally posted by: Modular
Well I have an awesome solution to this if it is true:

Don't run the new drivers...

DUH!

<Peter Griffin> Well, gee, I never thought about it that way... Alright!

 

[Modular]

Originally posted by: nitromullet
Not a possibility with a GX2

Okies, so 12 people worldwide can't run their cards overclocked...

Besides, a Rivatuner update will fix this problem, so it's a non-issue.

 

[nitromullet]

I tested this with RivaTuner in Oblivion, HL2:EP1, and rthdribl, and I didn't see any drops in core or memory clocks. There were a few dips in the games for level loads, but with rthdribl the core stayed pegged the entire time and the temps got up to 86C.

 

[Elfear]

Originally posted by: nitromullet
I tested this with RivaTuner in Oblivion, HL2:EP1, and rthdribl, and I didn't see any drops in core or memory clocks. There were a few dips in the games for level loads, but with rthdribl the core stayed pegged the entire time and the temps got up to 86C.

Maybe an isolated case for the OP than???

Off topic- How do you like your 7950GX2 compared to your X1900XTX nitro?