New Nvidia 411.63 drivers hinder performance

[TheRealDeal]

Anand used press driver 411.53 so betting the rest used that for there Nvidia cards.

 

[XeNoNF50]

I posted my highest fire strike ultra score with my GTX 1070 earlier today (and 2nd highest of anyone who has my CPU & GPU combination of i7-7820x & GTX 1070) with the latest drivers fwiw.

 

[David Bisset]

Even if you can prove it's slower across the board it'll just be dismissed as a bug. To be fair, when a big new release is coming out is the most likely time for there to be regressions so it doesn't have to be nefarious, though of course rather convenient if it does happen!

 

[Apone70]

bug or not a 200mhz drop in boost is crazy tho, as soon as i reinstall the 399.24 and rebooted it was back to 2100mhz

 

[Panos]

loads of 1080's above your score in TimeSpy Where are you 3 months of tuning Vega64 scores?

I wrote Firestrike not timespy.

 

[Smffy]

Someone should prove it and sue them but even if they did di it what would we do? And how would we prove that?

Trading standards do very little why they and Ofcom are useless. Any fines levied are peanuts compared to the money gained by thier legal scams.

Pretty sure ofcom won’t be regulating GPUs unless they start containing phones or posting letters

 

[StarShock]

Slightly off topic, but my theory on the fine wine argument is simply that amd tend to have better specific support for their cards after eol. So I don't think Nvidia actively gimp older cards. But I think they do stop optimising for a specific card sooner than amd do.

That's kind of all been thrown out of whack this last gen though, just because of how long Pascal has been around. What with it being 2 years old, but until last week, still current gen. So all the "Vega vs Pascal 1 year on" things don't show any significant change.

All of that being said I'm not sure if I'd trust Nvidia not to be pulling every trick they can think of to salvage this lacklustre launch

 

[Nasher]

We've been seeing the rx480 and 580 leaving the 1060 behind and overtake the 980 in some tests though :/

 

[KillBoY_UK]

To be honest, i think they have slowly being doing that for a while now.

this old chestnut

No change in clocks or performance here with my GTX 1080 ti

 

[Marine-RX179]

Anand used press driver 411.53 so betting the rest used that for there Nvidia cards.

So you are saying that the 2080 reviews that been reporting being roughly 8~10% faster than the 1080ti is more like they have the 1080ti has been slowed down to put the 2080 in better light because of this new driver?

 

[Gregster]

Maybe FineWine is broken on Vega

You can't use that as a fair result. You need to find a decent Vega card, downvolt it, overclock it and then compare it to the 1080 (no OC or fiddling allowed) and then retest and it will be fair.

 

[TNA]

You can't use that as a fair result. You need to find a decent Vega card, downvolt it, overclock it and then compare it to the 1080 (no OC or fiddling allowed) and then retest and it will be fair.

lol

 

[Marine-RX179]

Maybe FineWine is broken on Vega

I got a feeling we might not see how well the Vega age until a much later date (probably not until after the 7nm Vega release) and look back, but that's discussion for another date.

Thought this thread is about trying to work out if Nvidia is trying to make Pascal vinegar, not AMD making Vega wine...

It would be great if that video author would make a similar video, testing the Pascal cards new drivers vs old may be in a year's time.

 

[TheRealDeal]

So you are saying that the 2080 reviews that been reporting being roughly 8~10% faster than the 1080ti is more like they have the 1080ti has been slowed down to put the 2080 in better light because of this new driver?

Was not me, i was just confirming what the press was using driver wise. This would need a lot of testing to confirm.

 

[TheRealDeal]

You can't use that as a fair result. You need to find a decent Vega card, downvolt it, overclock it and then compare it to the 1080 (no OC or fiddling allowed) and then retest and it will be fair.

Tbf this is overclockers and Vega 64 is handicapped with that stock cooler a good deal. It is definitely a tweakers card. It's AMD's fault but i can see why people are frustrated at seeing a stock Vega in game tests as it's throttling and not showing the full picture. Any how back to Nvidia. People should get testing as there is plenty on here with 10 series cards.

 

[dragonslayer]

To everyone saying Nvidia wouldn't gimp their cards remember when Linus " proved" apple didnt gimp their phones and actually INCREASED performance over time yet apple openly admits 2 weeks later that they do in fact gimp their phones

 

[hyperseven]

I wrote Firestrike not timespy.

you really didn't

 

[Kirby Wurm]

It's not necessarily lowering the performance per clock, just dropping it pretty significantly for some reason. Strange.

 

[Gregster]

Actually when I had the GTX1080 Armor (watercooled at 2164/2190 core), I saw 5% performance drop on the September 2016 drivers and another 3% on the March 2017 drivers.
That is why even today my scores in the benchmarks in this forum for the GTX1080 aren't beaten two years later. Only the TimeSpy lost the lead but due to the fact someone used the 11Gbps VRAM version.


@JediFragger because things are relevant, go to the FS Standard and look at my Vega 64 graphics score with 18.8.1 and compare it even with Kaap's 1850 clocked Vega 64.
And since you are there, look at the higher GTX1080, which last time I checked was still mine.

Looking at that Firestrike score of yours, you clearly had a golden chip and should be proud of that score. Looking at the Timespy, your score got beaten a lot but saying performance was gimped with later drivers is daft, purely because why would NVidia gimp performance to hinder their own performance against Vega? That is a daft statement.

 

[ubersonic]

you really didn't

He did, I think maybe you jumped from one paragraph to the next without realising the content/context was different?