Nvidia Pascal Reviews [1080XP, 1080ti, 1080, 1070ti, 1070, 1060, 1050, and 1030]

[firstminion]

Good catch.
So AMD was not using the right driver for their presentation.
Although timeframe was a bit tight, so giving them benefit of doubt but maybe they should had pulled that segment.
If Nvidia marketing had wanted to be nasty (nasty being they know it is probably a genuine mistake) they could had gone with a public narrative AMD deliberately used old driver and did nothing to stop the wrong conclusion being published by Stardock (where it was inferred Nvidia altered this for performance gain, emphasis inferred).
Unless the issue is still with 368.25 and Dan has wrong info, should be easy for anyone to test though.
Must be at least 10 Nvidia owners who bought this game

Cheers

It's the same driver used for every 1080 launch review. How is it any way "wrong"? # Except that it makes every review "wrong"?

Reviewers can't check for image discrepancies, they do automated benchmarks: http://www.phoronix.com/scan.php?page=news_item&px=Why-Test-Automation

 

[Silent_Buddha]

Wow SLI is broken in launch drivers and it's not NVidia's fault.

Yeah I think Nvidia might be back sliding a bit on quality. Was watching DIstortion2 (very popular speed running streamer on Twitch, current world record holder for DS3) and his 970 recently died and took out his PCIE slot. So he had to replace both his graphics card and MB. He explained that's why he didn't stream for a bit. And then a lot of viewers started commenting on how their 970 has also died out of warranty and they had to replace it. One guy even mentions that his 970 caught on fire.

Which has me worried as I recently recommended a system with a 970 to a gaming friend of mine that is into PC gaming but knows nothing about PC hardware. Nice fast system. But unfortunately their computer locks up about once or twice a day while gaming with GPU unresponsive error. After watching that stream I'm dearly hoping it doesn't happen to her as I was the one that recommended the machine to her.

No problems so far with my Nvidia card though (knock on wood). Well unless I run Chrome. And then I get a system lock up if I have a game running and Chrome open after anywhere from 1-5 hours.

Regards,
SB

 

[CSI PC]

It's the same driver used for every 1080 launch review. How is it any way "wrong"? # Except that it makes every review "wrong"?

Reviewers can't check for image discrepancies, they do automated benchmarks: http://www.phoronix.com/scan.php?page=news_item&px=Why-Test-Automation

Because AMD are NOT reviewers and should be showing a retail 1080 setup vs their in-house development
Who is to say that the Day 1 retail drivers are not faster in general for games?

If they intend to compare to the competitors, then they should use retail cards with retail day 1 drivers when your comparing your own card to a competitors.
Also bear in mind this was a new card with pre-release drivers when reviewed, new drivers will happen and to be expected....

AMD screwed up and caused the mess that followed, and Stardock screwed up even more with their accusations straight off the bat.
Reviewers know and expect that something very new (before launch) whether a game or GPU may involve new patches/drivers.

Bear in mind AMD has had issues with new cards and drivers pre-release and still at time of review themselves in the recent past.
And why AMD cannot be defended as it is a very simple mistake to not check what are the retail drivers; still raises were they using a reviewer card and how why did they not have Day1 retail drivers (if the bug was resolved in them as mentioned by Oxide).

Cheers

Edit:
BTW That last point was summed up by Geforce Sean Pelletier (bit more stylish than one can expect from Roy Taylor) on twitter in response to Dan Baker twitter:

Also curious how AMD was running unreleased press drivers instead of the public launch driver for GTX 1080.
....
Weird. 368.19 was a press-only driver and not public. 368.25 was the first public driver. Wonder where they got that driver...

 

[Anarchist4000]

The simple answer to that would be that whoever set up the demo likely assembled everything with the newest drivers available at the time he or she created it. Any sales/marketing guy that values his career probably didn't procrastinate and create a demonstration for a major product launch less than a week in advance. Getting day 1 retail drivers prior to the launch is likely difficult.

 

[CSI PC]

The simple answer to that would be that whoever set up the demo likely assembled everything with the newest drivers available at the time he or she created it. Any sales/marketing guy that values his career probably didn't procrastinate and create a demonstration for a major product launch less than a week in advance. Getting day 1 retail drivers prior to the launch is likely difficult.

The demonstration-event was after the official 1080 launch when it was possible to buy it
And as I said they should had checked and at least pulled that segment if they did not have time to redo with the drivers that came out on the 25th.
When doing such a public international event, this type of project checks should be expected.

But applying your logic, AMD are still at fault because they are using NDA drivers given to reviewers, and possibly an NDA reviewer card.....
That is just as naughty, and even more of a reason they should be keeping an eye out for Day1 drivers.
Especially as they know from their own experience drivers will continue to evolve before-at review time and at time of release.
Whoever it was in AMD, they did screw up; made worst as they now it seems have publically shown they were using NDA non-public related drivers that AMD should not have (should had been kept internal within AMD and not presented if that is what they used).....
Cheers

 

[Anarchist4000]

The demonstration-event was after the official 1080 launch when it was possible to buy it
And as I said they should had checked and at least pulled that segment if they did not have time to redo with the drivers that came out on the 25th.
When doing such a public international event, this type of project checks should be expected.

But applying your logic, AMD are still at fault because they are using NDA drivers given to reviewers, and possibly an NDA reviewer card.....
That is just as naughty, and even more of a reason they should be keeping an eye out for Day1 drivers.
Cheers

Maybe, but if they have to pack it up and get it through customs likely weeks in advance, risking breaking the demo may be unwise. They also wouldn't want the numbers to not match up with what was in their presentations. Last minute adjustments for a presentation your boss is making might be pushing it. At the end of the day this wouldn't likely be an issue if Nvidia didn't have a buggy press driver.

 

[CSI PC]

Maybe, but if they have to pack it up and get it through customs likely weeks in advance, risking breaking the demo may be unwise. They also wouldn't want the numbers to not match up with what was in their presentations. Last minute adjustments for a presentation your boss is making might be pushing it. At the end of the day this wouldn't likely be an issue if Nvidia didn't have a buggy press driver.

Yeah AMD has never given buggy press drivers themselves
Seriously these are early drivers, both manufacturers continue to evolve their drivers up to Day1, and even afterwards a new GPU usually sees driver updates more frequent than usual.
If Nvidia had pulled the same stunt against AMD I would be just as critical and scathing of them.
Cheers

 

[trinibwoy]

Has anyone confirmed whether performance is different in the bugged drivers?

 

[lanek]

Has anyone confirmed whether performance is different in the bugged drivers?

I have not seen any reviewers do a proper comparaison test... If, theres a difference should not be by much anyway. ( well if the difference is only the snow ).

 

[CSI PC]

Well here is an alternative conspiracy theory.
How about Nvidia did deliberately create this bug specific to AoTS just to see if AMD would use NDA related 1080 setup (card+driver) rather than Day1 retail.
Very simple way to catch them out, and the odds of AMD using AoTS are pretty high.

Sort of like their own watermark as a gotcha
Yeah this is a tin foil hat type of post, and agree more likely was a genuine bug.
But as a conspiracy kinda amusing.
Cheers

 

[Anarchist4000]

Well here is an alternative conspiracy theory.
How about Nvidia did deliberately create this bug specific to AoTS just to see if AMD would use NDA related 1080 setup (card+driver) rather than Day1 retail.
Very simple way to catch them out, and the odds of AMD using AoTS are pretty high.

Sort of like their own watermark as a gotcha
Yeah this is a tin foil hat type of post, and agree more likely was a genuine bug.
But as a conspiracy kinda amusing.
Cheers

Or maybe AMD did it on purpose to show how 1080 runs hot compared to Polaris and all the snow melted!

 

[Lightman]

CSI PC I understand your points but in big businesses there is no place for improvisation when comes to public launches. That's why so many of the 'real-time' demos and 'running on actual hardware' is really replayed video or simulated to avoid any mishaps. There is no way someone doing all the benchmarking would be allowed to change driver very late in the process just because new driver was released few days earlier. This applies to AMD, Intel, nVidia, Microsoft and so on. There is a chain of command and the tester at the bottom, even if he could re-do all the benchmarks, has to first ask for permission someone above him and that person more than likely needs permission from higher up and so on. It can take days if not weeks to re-approve and in most cases the bottom guy will not even bother asking. No conspiracy needed here.

 

[Ryan Smith]

Has anyone confirmed whether performance is different in the bugged drivers?

Just checked it this morning on a GTX 1070. The average performance difference between 368.19 and 368.39 is 0.6%, well inside the margin of error.

 

[firstminion]

The demonstration-event was after the official 1080 launch when it was possible to buy it
And as I said they should had checked and at least pulled that segment if they did not have time to redo with the drivers that came out on the 25th.
When doing such a public international event, this type of project checks should be expected.

But applying your logic, AMD are still at fault because they are using NDA drivers given to reviewers, and possibly an NDA reviewer card.....
That is just as naughty, and even more of a reason they should be keeping an eye out for Day1 drivers.
Especially as they know from their own experience drivers will continue to evolve before-at review time and at time of release.
Whoever it was in AMD, they did screw up; made worst as they now it seems have publically shown they were using NDA non-public related drivers that AMD should not have (should had been kept internal within AMD and not presented if that is what they used).....
Cheers

1: NDA doesn't apply to other parties.
2: Why the driver is under NDA?
3: Why the hell the driver would be under NDA days after the product launch?

 

[Silent_Buddha]

Just checked it this morning on a GTX 1070. The average performance difference between 368.19 and 368.39 is 0.6%, well inside the margin of error.

Thanks for retesting Ryan, that's good to hear. I didn't want my cynical side to get any more cynical than it already is.

Regards,
SB

 

[Razor1]

http://videocardz.com/60973/nvidia-geforce-10-mobile-series-to-be-released-in-august

Notebook parts coming August

 

[CSI PC]

1: NDA doesn't apply to other parties.
2: Why the driver is under NDA?
3: Why the hell the driver would be under NDA days after the product launch?

1. True but AMD had a non-retail card, which they should not had as the only cards out there were under NDA.
They then went on to use that card publically without permission.
2. Because they are non-public drivers and never meant to be.
3. See point 2.

Cheers

 

[CSI PC]

CSI PC I understand your points but in big businesses there is no place for improvisation when comes to public launches. That's why so many of the 'real-time' demos and 'running on actual hardware' is really replayed video or simulated to avoid any mishaps. There is no way someone doing all the benchmarking would be allowed to change driver very late in the process just because new driver was released few days earlier. This applies to AMD, Intel, nVidia, Microsoft and so on. There is a chain of command and the tester at the bottom, even if he could re-do all the benchmarks, has to first ask for permission someone above him and that person more than likely needs permission from higher up and so on. It can take days if not weeks to re-approve and in most cases the bottom guy will not even bother asking. No conspiracy needed here.

I understand that, and I have been involved in a few international tech demos myself both to press and technical analysts.
Even up to the event we check everything, and if a segment is not right then it gets pulled, as long as it is not core to the presentation and event.
That is what I said quite awhile ago but lost in every post I am having to respond to, that 2 minute segment around AotS should had been pulled or heavily edited.
And yeah with live presentation involving tech demos, you ensure there are ways to pull them/skip a segment just in case they go awry.
There is a lot of responsibility to ensure that it not only works but the demo data presented is accurate to the press.

Their checklist to ensure their data was correct should had flagged the driver update from pre-launch to Day1 retail.
If they wanted to keep it in, then they should had a footnote clarifying they were using a pre-retail (driver+card) Nvidia product.
Cheers

 

[CSI PC]

Just checked it this morning on a GTX 1070. The average performance difference between 368.19 and 368.39 is 0.6%, well inside the margin of error.

I doubt that will kill the conspiracy theories though....
Still leaves my favourite one; They still could had done this to catch AMD using a NDA related 1080 and driver
Now where is my tinfoil hat.
Cheers

 

[DavidGraham]

The least AMD and Oxide should do is shut up about the visual corruption with the press drivers. And wait to see if it is present in the final version, instead they opted to make fools out of themselves, which lead them to look unprofessional, and made Oxide look like a pawn in AMD hands, surrendering to AMD's whims and releasing useless statements without checking facts first.