NVIDIA GeForce 378.78 Performance Driver Follow-Up @ [H]

I'd love to see 980 TI figures, Nvidia is notorious for lack of improvements on last gen cards. And because I have a 980ti!
 
I feel NVIDIA is being a little misleading here by comparing release performance to the new driver performance, but at least they're transparent about it in the included tables. When the optimized DX11 driver came out they were comparing performance from the prior driver release, if I'm not mistaken.
Thank you for these benchies and checkups on nvidia\amd.
Nvidia have a lot of catching up on Vulkan\DX12 to do, and they've done quite a bit but it's nice to see them actually keeping it up
Catching up with what? Performance delta? :rolleyes:
 
I'd love to see 980 TI figures, Nvidia is notorious for lack of improvements on last gen cards. And because I have a 980ti!

I unfortunately have a GTX970 (Which i absolutely hate, on my 2nd card and artifacts again + memory bug.)
I'd like to know how well it does, waiting for the e-tailer to go out of stock on all GTX970, RMA it again due to random artifacts and get a GTX1060 :)



I feel NVIDIA is being a little misleading here by comparing release performance to the new driver performance, but at least they're transparent about it in the included tables. When the optimized DX11 driver came out they were comparing performance from the prior driver release, if I'm not mistaken.

Catching up with what? Performance delta? :rolleyes:

Versus AMD in Vulkan\DX12 yes.
And versus themselfes in DX11 Yes
 
nv currently has the top 3 video cards on the market. It's nice to see them work on drivers for a bit of improvement. Driver dev for the most part has been really good on the nv side performance wise. I would say most driver releases are more about bug fixes than trying to improve performance numbers.
 
Disclaimer. Review is for DX11/12 games at 1440p.
Awaiting the DX12 4K driver review.

Driver improvements are for DX12 games @ 4K.
 
Last edited:
Disclaimer. Review is for DX11/12 games at 1440p.
Awaiting the DX12 4K driver review.

Driver improvements are for DX12 games @ 4K.
Yep, you are right. I can see in the small print that the resolution is set to 4K, max in game settings, and framerates are compared to driver 368.25.
 
I'd love to see 980 TI figures, Nvidia is notorious for lack of improvements on last gen cards. And because I have a 980ti!

This drivers performance update were specifically mentioned for Pascal GPUs only, hence only needing to re-test the 1080.
 
So did they say how they made this miraculous discovery of hidden performance that was untapped for so long?

Maybe they can fix their driver interface next, still from early 2000 design and locks up screen for a about a minute
 
Nice to see an overall driver timeline, but...was this really unexpected? These drivers were specifically for DX12 improvements and only 2 DX12 titles were tested, it appears.

For instance, it would have made more sense to test DX:MD in DX12 to see if there was any improvement there.

EDIT: Same with Rise of the Tomb Raider for that matter.
 
When you do the DX12/11 review of games with driver versions for Pascal... could you include the full line of Pascal boards, 1060, 1070, 1080, and 1080ti.

The reason is I wonder if the different number of shaders and other processors will benefit differently... and I own a 1070.
 
Nice to see an overall driver timeline, but...was this really unexpected? These drivers were specifically for DX12 improvements and only 2 DX12 titles were tested, it appears.

For instance, it would have made more sense to test DX:MD in DX12 to see if there was any improvement there.

EDIT: Same with Rise of the Tomb Raider for that matter.

It's still a new driver, and relevant to see if it has helped in other games performance, especially since it was touted as a performance driver.
 
So did they say how they made this miraculous discovery of hidden performance that was untapped for so long?
This isn't some conspiracy. GPU driver development is difficult, and they have had years to work on the DX11 driver. Direct3D 11 was released with Windows 7, in 2009. They've had 8 full years to optimize the D3D11 code path. DirectX 12 is still very new, and not many games use it. Also bear in mind that each game tends to do things differently, so a lot of the time, they'll have to analyze and optimize games on a per-title basis.
 
It's still a new driver, and relevant to see if it has helped in other games performance, especially since it was touted as a performance driver.

I think the issue is that both Tomb Raider and Dues Ex support DX12 while you ran the tests in DX11. Since this is a DX12 focused driver, it would have been nice to see those games tested in DX12 to see if Nvidia's claims of DX12 performance held true there too. Nvidia claims a massive 33% increase in Tomb Raider on DX12, seems like a missed opportunity not to test that claim.
 
I think the issue is that both Tomb Raider and Dues Ex support DX12 while you ran the tests in DX11. Since this is a DX12 focused driver, it would have been nice to see those games tested in DX12 to see if Nvidia's claims of DX12 performance held true there too. Nvidia claims a massive 33% increase in Tomb Raider on DX12, seems like a missed opportunity not to test that claim.
Did you read the whole article? They're doing a follow up covering all DX12 games.
 
At first i was a bit put off that this wasn't just a dx12 focus piece, as that was the PR Nvidia was highlighting in this driver release. Closer inspection points out that you are going to do a more in depth analysis of that in future editorials so I see where you are coming from. Most of your data has been based around a specific test bench and to recreate all of Nvidia's claims would require a massive undertaking of driver swapping under DX12 to verify their specific statements regarding dx12. However, Nvidia's blanket statements don't exactly make it clear where or when the performance gains will show, which is deceptive without argument. I'm all for driver improvement, as it is free performance that you are gaining after you have already committed to purchasing a product at face value. DX12 has been an ugly duckling for Nvidia so far this gen, and there has been little reason to use it other than w10 slowly forcing people to adopt it. Let us hope that Nvidia continues to improve in this area of performance since so many game studios are using custom engines that will continue to fragment the developer community and further the need for performance across an array of focuses.
 
Did you read the whole article? They're doing a follow up covering all DX12 games.

I did, but it seems very strange to say you're testing a performance driver but not actually focus on what the driver changed. It seems a waste of time to run tests the driver in no way claims to improve. A DX12 focused driver didn't help in DX11, not very informative.
 
I did, but it seems very strange to say you're testing a performance driver but not actually focus on what the driver changed. It seems a waste of time to run tests the driver in no way claims to improve. A DX12 focused driver didn't help in DX11, not very informative.
They explained that in detail:
The goal of this specific evaluation is simply to follow-up from our initial driver evaluation we performed back in February. We are taking the exact same games, the same setup, the same video card (GeForce GTX 1080) and just adding on this new driver to the graphs.

The goal is to simply find out if there are any performance differences at all with this new driver compared to the previous driver in the games we already tested and used in our driver evaluation. Those games are a mix of older games, DX11 games, and a couple of DX12 games.

The main focus on this is not DX12 today, but rather to see if this new "Performance Driver" has helped any at all in any situation of old game, DX11 game or the DX12 games we used previously. It is strictly that, a test to find out if this driver has improved since the last one in other ways than simply DX12 and to provide a follow-up to our first driver article.

We fully understand there are many DirectX 12 games or games with DirectX 12 render paths that need to be re-evaluated under this new driver. We are planning just that in the future. In the future, at some point, we are going to go back through all the games with a DirectX 12 API path and re-evaluate DX12 versus DX11 performance in them using this newer performance driver. That will be a very fun and comprehensive evaluation to finally put the nail in the coffin on whether DX12 provides a performance impact in those games now.

That will be an upcoming article, planned for the future, it’s on the board but an exact ETA we do not have just yet, but it will come in time. Therefore, look forward to that re-evaluation of DX12 API games with this (or newer) driver performance. For now, today’s evaluation is just to see where this driver stands in relation to what we’ve already tested in terms of driver over time in January as a quick follow-up, nothing more. The format is staying the same from the original driver evaluation, we are using the same graphs and performance we are just adding one more bar to the graph with the new driver at the top.
This was a quick follow up to the testing they already did on both AMD and NVIDIA's driver performance improvements over time. Redoing all of the testing in both DX11 and DX12 in all DX12 compatible titles is a much larger time investment, hence this quicker article.
 
They explained that in detail:This was a quick follow up to the testing they already did on both AMD and NVIDIA's driver performance improvements over time. Redoing all of the testing in both DX11 and DX12 in all DX12 compatible titles is a much larger time investment, hence this quicker article.
Sure, but there wasn't anything marketed by Nvidia that was tested, so it seems odd that the DX11 tests which weren't mentioned as part of the performance improvements were even bothered with first(or at all to be honest, my expectation for DX11 performance would be that it's next to nothing, lo and behold it was next to nothing).
 
I think the issue is that both Tomb Raider and Dues Ex support DX12 while you ran the tests in DX11. Since this is a DX12 focused driver, it would have been nice to see those games tested in DX12 to see if Nvidia's claims of DX12 performance held true there too. Nvidia claims a massive 33% increase in Tomb Raider on DX12, seems like a missed opportunity not to test that claim.

The explanation of the review setup is explained in the review, with an explanation that we will work on a DX12 evaluation. This review is a followup to the one we did in February, using the same games, same setup, to evaluate driver performance over time. It's all explained in the review. Reading the text is important.

I would argue DX11 performance is still important, and relevant to test. It was necessary to find out if DX11 performance was improved in any games. I believe this information will be useful to gamers.
 
Sure, but there wasn't anything marketed by Nvidia that was tested, so it seems odd that the DX11 tests which weren't mentioned as part of the performance improvements were even bothered with first(or at all to be honest, my expectation for DX11 performance would be that it's next to nothing, lo and behold it was next to nothing).

I did, but it seems very strange to say you're testing a performance driver but not actually focus on what the driver changed. It seems a waste of time to run tests the driver in no way claims to improve. A DX12 focused driver didn't help in DX11, not very informative.

Then I guess the article isn't meant for you, you can easily click close on the article and simply not read it if it is of no interest to you, there would also be no reason to post in this thread if you find no interest or point in the article.

IMO, DX11 performance is still important, and worthy of being tested, especially if a driver is touted as being a performance driver, as it was. It was only later that it was found out it was more specific to DX12, than anything.

Even finding out that it doesn't help in any DX11 games, is information worthy in of itself. Added to the other driver performance we tested it makes for a complete look at driver performance over time, which was the goal of the test, as explained in the review.
 
Then I guess the article isn't meant for you, you can easily click close on the article and simply not read it if it is of no interest to you, there would also be no reason to post in this thread if you find no interest or point in the article.

IMO, DX11 performance is still important, and worthy of being tested, especially if a driver is touted as being a performance driver, as it was. It was only later that it was found out it was more specific to DX12, than anything.

Even finding out that it doesn't help in any DX11 games, is information worthy in of itself. Added to the other driver performance we tested it makes for a complete look at driver performance over time, which was the goal of the test, as explained in the review.
I never said it wasn't important, but complaining that I could simply not read the article is a bit silly when:
Nothing in the title or summary on the site indicates the DX12 performance was mostly left out.
The introduction indicates that DX12 games were included in the testing(and you've got some DX12 titles running in DX11).

It should be obvious, but I wouldn't be shocked if readers generally cut straight to the "meat" of an article after skimming through the summary. You're right that you did mention on the first page that it was primarily focused on DX11 testing... perhaps it might have simply been prudent to indicate as such in the summary or introduction then? I never said DX11 shouldn't have been tested, I never said DX11 games shouldn't have been tested, but it is simply an odd read when the big deal about this was DX12 performance and with some of the DX12 games there are only results listed for DX11. Sure, you could put the blame on the reader for skimming the article and not examining everything in its entirety, but you could also make it crystal clear right off the bat in the summary and intro what the focus of the article is actually about rather than solely being about nvidia's very specific claim.
 
there would also be no reason to post in this thread if you find no interest or point in the article.

Some businesses find feedback from their customers to be useful, especially when it comes to what type of content is help and what and is not. You may disagree with the feedback which is absolutely within your right, but discouraging your audience from providing feedback seems a little short sighted.

There is no ill will meant, simply an opinion on what type of content is helpful.

IMO, DX11 performance is still important, and worthy of being tested, especially if a driver is touted as being a performance driver, as it was. It was only later that it was found out it was more specific to DX12, than anything.

Maybe the early press materials you received were different, but everything shown to the public about this driver was very specific about it's DX12 performance focus. Even the marketing slides you included only mention the focus on DX12.
 
Some businesses find feedback from their customers to be useful, especially when it comes to what type of content is help and what and is not. You may disagree with the feedback which is absolutely within your right, but discouraging your audience from providing feedback seems a little short sighted.

There is no ill will meant, simply an opinion on what type of content is helpful.



Maybe the early press materials you received were different, but everything shown to the public about this driver was very specific about it's DX12 performance focus. Even the marketing slides you included only mention the focus on DX12.

We did indicate this in the evaluation. The goal was stated, the explanation for the setup was stated, and the fact we will be doing a specific DX12 roundup re-evaluation was stated. So it's all explained in the review.

This follow-up did not take long to put together, it was meant as a quick follow-up to our driver evaluation in February using this latest driver, the DX12 evaluation will take longer, it's a whole different kind of evaluation. We are going to compare DX12 vs. DX11 performance in games, and there are several of them. But all this has been said already.

I'm sorry you did not get anything out of this evaluation today, luckily reading it is optional.
 
We did indicate this in the evaluation. The goal was stated, the explanation for the setup was stated, and the fact we will be doing a specific DX12 roundup re-evaluation was stated. So it's all explained in the review.

This follow-up did not take long to put together, it was meant as a quick follow-up to our driver evaluation in February using this latest driver, the DX12 evaluation will take longer, it's a whole different kind of evaluation. We are going to compare DX12 vs. DX11 performance in games, and there are several of them. But all this has been said already.

I'm sorry you did not get anything out of this evaluation today, luckily reading it is optional.
Ya know, you could have simply noted the constructional criticism that I explained, instead of continuing to take it personally for some reason. Right?
 
Ya know, you could have simply noted the constructional criticism that I explained, instead of continuing to take it personally for some reason. Right?

Your feedback is noted. Let's get this thread back on topic now.

For those wondering about the setup, and why things are they way they are, explanation is in the review.

DX12 testing is already underway, as of today.
 
I can personally say in Rise of the Tomb Raider it is a noticeable increase, 4x ssaa is alot smoother than what it was before the update and it is at 60 FPS most of the time now, whereas before the update it averaged around 45-55 most of the time. Granted, I am on a 1080 p 60hz TV so I dont know how high my FPS could go and how it would act on higher resolutions. I do know though that in the next review you all will see a good increase. Honestly it was like I did a slight hardware upgrade in this game.
 
I can personally say in Rise of the Tomb Raider it is a noticeable increase, 4x ssaa is alot smoother than what it was before the update and it is at 60 FPS most of the time now, whereas before the update it averaged around 45-55 most of the time. Granted, I am on a 1080 p 60hz TV so I dont know how high my FPS could go and how it would act on higher resolutions. I do know though that in the next review you all will see a good increase. Honestly it was like I did a slight hardware upgrade in this game.

I'm having a performance increase that is around 10% in DX12 playing RotTR. I have a GTX980Ti. Even in the benchmark, it shows lows in the 40fps where before there were sometimes single digits. Playing at 1440P.
 
Reading is fundamental. Now please get onto the topic of what the review is about, instead of what it is not about. Thanks.
 
Good follow up on the previous review and some promising results for DX 12 and Vulkan (few titles that is). I fail to understand some of the criticism dealing with testing beyond the initial or first part to this? Keeping it tight and consistent with the previous review makes a more comprehensive and accurate account of driver performance over time. The title of the review is very clear and I would be more critical if the testing was different then the first review being that it is a follow up to the first one. Anyways, yes I would like to know if the performance gains in DX 12 also meant some DX 11 gains as well (now I know they do not).

Look forward to the next review dealing particularly with DX 12. If the performance improvements are that good with DX 12 games and I hope to see games not listed with DX 12 also tested to see if there are improvements or not. Reason being is that if Nvidia can optimize per game in DX 12 like DX 11 then what will prevent them to optimize all games like they did before? Which leads to AMD - no DX 12 advantage.

I am more looking at Nvidia DX 11 performance optimizations as a positive in that Nvidia will get to the user the most out of their video card sooner rather than later - That to me is better than seeing performance out of your video card a year or more later when it had that ability to begin with. Don't get me wrong, if more performance can be gain then both companies need to do that as much as possible - just looks like Nvidia is more proficient and quicker to the punch.
 
Did you checked if the infamous telemetry is still installed and if is still possible to remove it?
 
So I am not too impressed with the numbers. it seems odd to advertise these drivers as a big DX12 improvement if we are literally talking about 3to9 frames a second better.
 
I'd love to see 980 TI figures, Nvidia is notorious for lack of improvements on last gen cards. And because I have a 980ti!
There is a performance test of GTX 1080 and 980ti with this new driver on Anandtech.
 
So did they say how they made this miraculous discovery of hidden performance that was untapped for so long?

Maybe they can fix their driver interface next, still from early 2000 design and locks up screen for a about a minute

I totally agree and here I thought I was the only one with that issue. It's always been embarrassing when I brag about my rigs and then want to make an adjustment and then wait, wait, almost there, its thinking about, there's the CP!
The only thing I've noticed is that the cards out of the box come up quick but as soon as you start adding updates(even with clean installs) CP starts to slow down.
 
Thank's Kyle for these driver reviews. I sometimes cringe when I see that notification on the task bar from NV anymore. These save me a lot of grief with which updates I really should focus on.

If I have time I think I'll update the 1080rig this weekend. Those improvements in ROTTR look very tempting. Until now DX12 for it was something I wanted to work but was still crash happy and problematic.
 
Back
Top