Nvidia misses GTX470/480 launch targets

[Mungri]

You can't include adding volts if we measure overclocking.

Of course you can. How many people overclock CPUs by adjusting voltages? When overclocking, voltage is important for stability. Overclocking reviews on graphics cards which dont include voltage adjustment are completely ridiculous ever since MSI afterburner was released. If its so easy to increase the voltages when overclocking graphics cards these days then why bother ignoring it when writing reviews on overclocking potential of graphics cards?

I dont care for what any review says about overclocking if they only used stock voltage. They can also feel free to increase volts on the GTX 400s as well if they like, but those things wont have much headroom due to their temperatures.

Heres a good review which includes voltage adjustment for the 5870:

http://www.bit-tech.net/hardware/gra...tweak-review/9

[GoldenTiger]

We will be the judge of that, thanks


Thanks for the information though, I know you are eager to share information but always expect criticism because our standards are very high.


No ruckus caused here, only information sharing.

No problem at all... and yes, I was just trying to share info since not many have them yet, is all . Anyway, thanks for the info (I wasn't aware of those extra settings even existing), and hopefully that clarifies things some. Criticism's fine, and I'm glad to see it done in a constructive way here rather than outright flames like at some other forums.

[LordEC911]

You can't include adding volts if we measure overclocking. If we do that its going to be impossible to get any real comparison done because we open a whole new can of variables. All the oc with fermi so far have been with stock volts so if we want to do a more objective comparison, we should compare stock volts OC with the 5870. If you look at any review, you don't see them adding additional volts to a card do you?

The toxic editions of the 5870s cards which have better than average cooling, higher volts still only hits a 975. In addition they are significantly more expensive than their plan jane counterparts.

If you look at reviews, 1000mhz OC on the 5870 are rare.

You do realize stock voltage is variable...

[Sn@ke:~]

ok lets say 1000 mhz is rare..although on here there is more then a few ppl to hit 1000+...but for argument sake..its rare..

however u say 975mhz is only 50mhz overclock?? based off what? the stock original 5870? which is 850mhz...that looks to be a 125mhz over stock..which is still 100mhz more then what that review used as an "overclocked" 5870

now if they used at least a 950 as an overclocked version id say thats fair to say its overclocked but 25mhz isn't going to do much of anything as can be seen by the review....but we all know that 950-1000 vs 850 is quite a bit larger increase...at least from what ive seen floating around here.

None the less the 480gtx at stock beated the 5870 and the gtx480 can OC as much as the 5870. It is a newer card with no revision driver for it and yet no voltage modification was done, it can also be cooled better on AIR and even with WC and there is 32 core disabled via bios while the remaining core are not wotking at full capacity.

To be honest i don't see why i would even think of arguying these fact.

[Chickenfeed]

You do realize stock voltage is variable...

Yes this is a problem with 5870s given they can vary from 1.15 to 1.172 ( newest ATI reference bios ) Generally the vendors who shipped cards with lower stock voltages clock nicer but this isn't always true. An ASUS I used (1.15) did worse than my own card (1.168)

@ Damha

If I am not mistaken with current drivers and 5 series hardware there are only 3 AA settings (multi-sample AA, adaptive multi-sample AA, super-sample AA) I know the older Catalysts allowed picking standard and quality (drop down box) but I do believe the driver team mentioned this setting is now automatic based on application ( in otherwords they are making the decision at a driver level now not the user ) because they felt it was a confusing setting.

[kemo]

Fan on auto will do

That is the Max i got to pass 5 min of Furmark xtreme with fan at Auto and V on 1.35 , with fan @ 100% i could pass it at 1GHZ but with Auto = fail




92C with fan @ 62%

[Mungri]

That is the Max i got to pass 5 min of Furmark xtreme with fan at Auto and V on 1.35 , with fan @ 100% i could pass it at 1GHZ but with Auto = fail




92C with fan @ 62%

Upgrade the paste to MX3 and then try again. You will be surprised . Also you really shouldnt need 1.35v for 960 Mhz.

[kemo]

Upgrade the paste to MX3 and then try again. You will be surprised . Also you really shouldnt need 1.35v for 960 Mhz.

Like many said before , benching is something and 24/7 is something else , i can bench 15 minute Far cry 2 with fan @ 60% and 1GHZ with 1.32V and temps never get past 75C , and for the MX3 i dont really care much about temps and i game at stock clocks anyway

[GoldenTiger]

Like many said before , benching is something and 24/7 is something else , i can bench 15 minute Far cry 2 with fan @ 60% and 1GHZ with 1.32V and temps never get past 75C , and for the MX3 i dont really care much about temps and i game at stock clocks anyway

Definitely, I don't personally care if you can hobble a card to run at an insane speed for a short session: I care about what I can actually use it at in games, not just force it through a short 5-minute synthetic and pray it doesn't crash .

Here's a quick Unigine bench... anyone got same-setting 5870/5970 ones they can show?

[tajoh111]

Of course you can. How many people overclock CPUs by adjusting voltages? When overclocking, voltage is important for stability. Overclocking reviews on graphics cards which dont include voltage adjustment are completely ridiculous ever since MSI afterburner was released. If its so easy to increase the voltages when overclocking graphics cards these days then why bother ignoring it when writing reviews on overclocking potential of graphics cards?

I dont care for what any review says about overclocking if they only used stock voltage. They can also feel free to increase volts on the GTX 400s as well if they like, but those things wont have much headroom due to their temperatures.

Heres a good review which includes voltage adjustment for the 5870:

http://www.bit-tech.net/hardware/gra...tweak-review/9

The difference is volt changes in the GPU are not as easy to do. With some models that have the software added its easy(but these tend to be more expensive) but for the most part people don't want to do a bios videocard flash, so they can use other models software.

When you are talking about changing thermal paste on stock cooling, your not even talking about regular overclocking anymore because that will void the warranty with alot of companies. As see from the people in this thread, these 1000 mhz overclocks come at a cost. 1.35 volts is a tremendous jump in volts, in addition 100% fan is not pleasant and is only useful for benching.

Videocard overclocking has typically entailed using the stock softsoft, move up the clock and thats it. For benching you might move up the fan profile. However, it typically doesn't involve adding volts and changing thermal past or doing volt mods.

The gtx 470 are also decent clockers. Just look at golden tigers results, 20% overclock, likely stock volts.

[Chickenfeed]

I get 37 average with 7 min on my 5870 at those settings. I find it interesting that the 5 series drop so low ( especially in 2.0 ) It will be a more telling story once we see an actual game using uniengine.

Any chance you could run Vantage on the extreme preset with that OC (before/after)? Reviews didn't show much in the way of 470 overclocked numbers in much of anything.

[Eastcoasthandle]

I get 37 average with 7 min on my 5870 at those settings. I find it interesting that the 5 series drop so low ( especially in 2.0 ) It will be a more telling story once we see an actual game using uniengine.

Any chance you could run Vantage on the extreme preset with that OC? Reviews didn't show much in the way of 470 overclocked numbers in much of anything.

I think he was using satire when he posted that

[GoldenTiger]

The gtx 470 are also decent clockers. Just look at golden tigers results, 20% overclock, likely stock volts.

Yep, stock volts .

I get 37 average with 7 min on my 5870 at those settings. I find it interesting that the 5 series drop so low ( especially in 2.0 ) It will be a more telling story once we see an actual game using uniengine.

Any chance you could run Vantage on the extreme preset with that OC? Reviews didn't show much in the way of 470 overclocked numbers in much of anything.

I can install Vantage and do so... I'll post back soon. Test was with 197.41 drivers, FYI.

I think he was using satire when he posted that

?

[justin.kerr]

won't be home for a couple days... I have this one saved on photobucket, I know it is a ways off on settings, all stock on the 5870's, when I get home I can post same settings, with overclocks, cards are 1050, 1280 mem and have been using them for a while with no issues gaming. so settings seem good.

[GoldenTiger]

3dmark Vantage Extreme preset (197.41 drivers, 720c/1440s/3520m), GPU physx DISABLED in the Nvidia control panel:



http://service.futuremark.com/compare?3dmv=2092505

[Chickenfeed]

Heres a Vantage run on with a 5870 :


(excuse my dual copy paste, paint fail lol )
@ Justin

How many 5870s are you running? I find it odd that the min is that much higher... unless its 3 or 4 of them it doesn't reflect what I remember seeing in the Heaven 2.0 thread

@GT

Thanks for doing that. Obviously the first feature test doesn't work yet with G100 but what I find interesting is how much better they do on the cloth and particle simulatiom (Feature 4/5). I'd like to know how it is coded ( I'm guessing compute shader? ) Really shows GF100 strengths in physics calculations. Its almost like ATI / Nvidia have a role reversal... back in the day Nvidia gpus were texture monsters where ATI tended to be stronger compute wise and now the tables have turned hehe. Personally though I feel the days of texture heavy designs are behind us ( I'll be shocked if ATI / Nvidias next cards are as texture heavy - ie tmus , fillrate ect )

[InCredible]

None the less the 480gtx at stock beated the 5870 and the gtx480 can OC as much as the 5870. It is a newer card with no revision driver for it and yet no voltage modification was done, it can also be cooled better on AIR and even with WC and there is 32 core disabled via bios while the remaining core are not wotking at full capacity.

To be honest i don't see why i would even think of arguying these fact.

for one the gtx doesn't overclock as high
for two the OC 5870 was hardly that..why clock the 480 100mhz over but the 5870 only 25 when it can easily do 100mhz over with stock clocks

as far as it being beat and arguing the points...i mean we know the 480 is faster..thats for sure..all i was getting at is the over clock is hardly an overclock for the 5870..thats all i was getting at..

but your right u can quit arguing any points for the 480..it is faster then the 5870 while consuming 70+ watts or more doing so..we understand this much

[justin.kerr]

@ Justin

How many 5870s are you running? I find it odd that the min is that much higher... unless its 3 or 4 of them it doesn't reflect what I remember seeing in the Heaven 2.0 thread

3 at stock on that run

[trinibwoy]

Apologies if this is a repost, more overclocking results - 825/1100 GTX 480 vs 975/1300 HD 5870.

http://alienbabeltech.com/main/?p=16997

[eric66]

hmmm seems fermi arch is very responsive to oc btw love how review sites ignore 5970 lol

[japamd]

Definitely, I don't personally care if you can hobble a card to run at an insane speed for a short session: I care about what I can actually use it at in games, not just force it through a short 5-minute synthetic and pray it doesn't crash .

Here's a quick Unigine bench... anyone got same-setting 5870/5970 ones they can show?

Thanks for the heads up about 470, you got a nice overclocker there, congrats

Here's one with 5970 OC 1000/5000



The min fps is quick hiccup, always at same place.

Apologies if this is a repost, more overclocking results - 825/1100 GTX 480 vs 975/1300 HD 5870.

http://alienbabeltech.com/main/?p=16997

I think they pushed memory too far, the results are very bad for that much gpu clock.

[Chickenfeed]

Apologies if this is a repost, more overclocking results - 825/1100 GTX 480 vs 975/1300 HD 5870.

http://alienbabeltech.com/main/?p=16997

The mention of asking Nvidia about the TDP is interesting but they are in fact right. TDP is not max power draw like people seem to think. That said though obviously Nvidia took a liberty testing as they did as it turned out to be misleading ( ie AMDs TDP claim is more grounded in reality ) It was a PR move if you ask me. 250watt looks better marketing wise than the more realistic 300watt when comparing to the 5970/295.

It is still impressive how well it scales with clock speeds. I could only imagine what how this would do with a node shrink and the full core amount... I'm sure it may even trade blows with the 5970. Thats said though, can Nvidia out a shrinked fully operational Fermi based consumer gpu before AMD refreshes their line ( even if it means holding the PR crown for a short while )

@japamd

Yeah most people are seeing similar dips like that... Im really curious how justin managed 31 with 3 5870s... i wouldnt think a 3rd would magically make the min go from sub 10 to 30 but perhaps? I do notice he didn't use af in that test so just perhaps... I'm interested to see now if AF somehow nukes performance in Unigine on 5 series hardware..

[GoldenTiger]

Thanks for the heads up about 470, you got a nice overclocker there, congrats

Here's one with 5970 OC 1000/5000

[.IMG]http://i42.tinypic.com/152c4t0.jpg[/IMG]

The min fps is quick hiccup, always at same place.

Yeah, my card had that too but rerunning the bench a second time fixed it . Thanks for the result, was curious as to how it stacked up.

@Chickenfeed... Yeah, if they put out a shrink with full cores it would be nuts, these things seem to scale really well.

[eric66]

whats the avg overclock for 470 ?

[InCredible]

Apologies if this is a repost, more overclocking results - 825/1100 GTX 480 vs 975/1300 HD 5870.

http://alienbabeltech.com/main/?p=16997

much better oc testing