NVidia driver 337.50 beta released!

[Ed1]

Just fyi, it isn't a "forum rumor," hardocp stated in their review that their 780ti was reading higher boost clocks and 3 degrees higher temps than previous drivers. Since I don't speak German, I can't confirm the actual translation, but I could shoot an email to h to confirm.

These drivers seem to have issues with AB and other OC utilities for some people ,hopefully WHQL will be ok .

HardOCP used reference 780ti , most people in forums use 3rd party coolers/cards so they don't get to high 80's .
I have not seen anyone mention higher clocks in forums .
Probably just bug in driver .

 

[Despoiler]

If you want to make a fair comparison in the best light possible for both vendors, then Windows 8.1 is absolutely necessary..

Absolutely necessary why? The only thing that needs to happen is that AMD benches in Win8.1 need to be compared to Nvidia benches in Win8.1. The same goes with Win7. A complete review would include both OSs. The install base for Win7 is vastly higher than Win8/8.1. Win7 has 62% and Win8 has 22% according the the Steam hardware survey.

I'm still waiting for someone to show min fps and frame latency reductions on the order of Mantle's improvement.

 

[VulgarDisplay]

my 780 is not clocking any higher and it's going to the same exact boost clock as always. and lets not be dramatic here as hard is reporting just a 13 mhz increase from when they originally reviewed that 780 ti. And I really doubt the temps are being reported in a perfectly controlled environment. Plus if the card is being utilized more with the new driver then temps would go up a degree or two.

Let me guess your card is over clocked and using an aftermarket cooler?

 

[Hitman928]

my 780 is not clocking any higher and it's going to the same exact boost clock as always. and lets not be dramatic here as hard is reporting just a 13 mhz increase from when they originally reviewed that 780 ti. And I really doubt the temps are being reported in a perfectly controlled environment. Plus if the card is being utilized more with the new driver then temps would go up a degree or two.

More specifically, [H] said that their card was going above the 84 C thermal limit implemented by previous drivers, so without an increase in the thermal limit, pushing the card harder wouldn't result in higher temps because it would throttle to maintain 84 C. This would also allow [theoretically] for more than ~13 MHz of increased performance if it was able to maintain a higher boost for longer due to the increased temp limit.

These drivers seem to have issues with AB and other OC utilities for some people ,hopefully WHQL will be ok .

HardOCP used reference 780ti , most people in forums use 3rd party coolers/cards so they don't get to high 80's .
I have not seen anyone mention higher clocks in forums .
Probably just bug in driver .

I wouldn't be surprised if it was just some bug in the driver or something due to their particular setup or a combination of factors, would be nice for them or someone else to really test it though and show some temp/clocks over time. Either way, I see it as a minor thing, just wanted to mention that it didn't come out of thin air and would like someone to follow up on it with some actual testing is all.

 

[Carfax83]

Absolutely necessary why? The only thing that needs to happen is that AMD benches in Win8.1 need to be compared to Nvidia benches in Win8.1. The same goes with Win7. A complete review would include both OSs. The install base for Win7 is vastly higher than Win8/8.1. Win7 has 62% and Win8 has 22% according the the Steam hardware survey.

This has nothing to do with install base. It's about showing the highest possible performance on both platforms. If AMD gets to use the Mantle path, then NVidia should get to use the DX11.1 path.

Anything else is just unfair..

 

[Carfax83]

Man it's too bad Watch Dogs isn't out yet. That game would have been perfect to test with these drivers, as it's supposed to be very CPU heavy and uses a multithreaded renderer..

 

[Sinthoras]

This has nothing to do with install base. It's about showing the highest possible performance on both platforms. If AMD gets to use the Mantle path, then NVidia should get to use the DX11.1 path.

Anything else is just unfair..

Actually, AMD gave Nvidia the permission to use the Mantle API.
And it's not unfair, AMD played this well and spend much time creating Mantle.
It would cost Nvidia the same amount of time to create something of their own. They decided to optimize DX11 as reaction. You can't forbid a company trying to get on the top.

 

[Deders]

Not that I know, I've only seen one site testing HT off vs on. But I forgot the name of that site.

Most bigger sites have been reviewing with a GTX 780 Ti in SLI and an i7 3930K, that's why I like the one on blackholetec testing with a GTX 760.

Yeah, testing with top of the range equipment kind of misses the point with Mantle and possibly these drivers if they are supposed to be decreasing CPU overhead.

 

[blackened23]

Actually, AMD gave Nvidia the permission to use the Mantle API.

This is actually not the case. The API is not public and AMD has not released it to anyone, nor did they give anyone permission to use it except their inner circle of developers under the AMD GE program.

John Mcdonald, who worked at NV and now at Valve - he posts at Neogaf and has stated pretty much in no uncertain terms that AMD is not releasing Mantle to anyone. It is not "open" in any sense of the word. The API is not available to anyone except their inner circle of AMD GE developers, so despite what their marketing would have you believe, it isn't true.

Which is why NV is pushing so hard for DX12 beta to happen sooner rather than later.

 

[Hitman928]

This has nothing to do with install base. It's about showing the highest possible performance on both platforms. If AMD gets to use the Mantle path, then NVidia should get to use the DX11.1 path.

Anything else is just unfair..

I could see it both ways. For instance, if you are a gamer on Win7 and every reviewer only tested the "best possible scenario" for each card, if you then bought an Nvidia card, you would be disappointed because you wouldn't have the performance you were shown (at least in games where it matters). However, if you bought an AMD card, you would have the same performance you expected because Mantle works the same either way.

I think it's important to show both cases. If nothing else, most reviewers are trying to earn a living and therefore need to show tests that are relevant to their readership. If they only tested in Win8.1, they would lose relevance with many potential readers. There's nothing wrong with saying under X conditions this is the outcome but under Y conditions the outcome is changed by a factor of Z. It's akin to testing different resolutions, if you only tested all cards at 4k because one brand does better at higher res, you would lose a lot of relevance to most people.

 

[Carfax83]

Actually, AMD gave Nvidia the permission to use the Mantle API.

Source? Everything I've read implies that AMD has no intention of opening up Mantle to any other IHVs. It was never meant to be open source as evidenced by the fact that there still isn't any publically available SDK..

 

[Despoiler]

This has nothing to do with install base. It's about showing the highest possible performance on both platforms. If AMD gets to use the Mantle path, then NVidia should get to use the DX11.1 path.

Anything else is just unfair..

Unfair? The responsibility of both companies to serve their user base on whatever platforms they support. As both companies support Win7 and Win8 the performance on both platforms matters. You can't honestly be arguing that Win7 users should totally disregard the system performance benchmarks that matter to them and only consider Win8 performance. Nvidia framed this driver as competing with Mantle, but I guess what you are saying is they need a footnote to those claims *Except in Win7 where we accept total defeat.

 

[Hitman928]

Source? Everything I've read implies that AMD has no intention of opening up Mantle to anyone. It was never meant to be open source as evidenced by the fact that there still isn't any publically available SDK..

I don't think anyone knows the extent of AMD's intentions with Mantle at this point in time, especially with the dx12 announcement. The fact that the SDK isn't public yet isn't important, the API is still under development and won't have a public SDK until its ready, that's a normal progression of things. The impression AMD/Dice gave was that once the SDK is public, they would allow anyone to design for it, hardware or software, but again, who knows, we'll have to wait and see what happens.

Either way, Nvidia most likely would have to adjust their architecture to take full advantage of Mantle and I don't see them doing that, even if they had the opportunity, not with dx12 coming. Unless Mantle becomes a de facto standard in the industry (doubtful), I don't think you will see Mantle running on Nvidia hardware.

 

[blackened23]

I don't think anyone knows the extent of AMD's intentions with Mantle at this point in time, especially with the dx12 announcement. The fact that the SDK isn't public yet isn't important, the API is still under development and won't have a public SDK until its ready, that's a normal progression of things. The impression AMD/Dice gave was that once the SDK is public, they would allow anyone to design for it, hardware or software, but again, who knows, we'll have to wait and see what happens.

Either way, Nvidia most likely would have to adjust their architecture to take full advantage of Mantle and I don't see them doing that, even if they had the opportunity, not with dx12 coming. Unless Mantle becomes a de facto standard in the industry (doubtful), I don't think you will see Mantle running on Nvidia hardware.

It's ridiculous to think that any other vendor could use Mantle anyway, even if AMD made it open (they didn't). That would require other vendors to adopt an AMD like GCN architecture.

As much as AMD would love to have everyone by the cajones in doing that, it just isn't going to happen. Intel and NV aren't going to develop AMD based GCN hardware just to use Mantle, even if it were open. But it isn't open in any sense of the word, anyway, so it's a moot point on both the hardware and software fronts.

And that isn't necessarily a bad thing. When DX12 hits, it won't matter since everything will move over to DX12, so for the time being Mantle is simply AMD's proprietary value added feature to help them sell GPUs. And they have a year or so to do their thing with Mantle before DX12 hits. Which is fine really, Mantle will have extremely limited impact compared to any revision of D3D.

 

[Sinthoras]

It's ridiculous to think that any other vendor could use Mantle anyway, even if AMD made it open (they didn't). That would require other vendors to adopt an AMD like GCN architecture.

As much as AMD would love to have everyone by the cajones in doing that, it just isn't going to happen. Intel and NV aren't going to develop AMD based GCN hardware just to use Mantle, even if it were open. But it isn't open in any sense of the word, anyway, so it's a moot point on both the hardware and software fronts.

And that isn't necessarily a bad thing. When DX12 hits, it won't matter since everything will move over to DX12, so for the time being Mantle is simply AMD's proprietary value added feature to help them sell GPUs. And they have a year or so to do their thing with Mantle before DX12 hits. Which is fine really, Mantle will have extremely limited impact compared to any revision of D3D.

Well, the reason why I thought it, was because I read some time ago that Nvidia had been discussing a long time to start supporting Mantle in their next architecture or not. Seems like I imagined it or something...

 

[Hitman928]

It's ridiculous to think that any other vendor could use Mantle anyway, even if AMD made it open (they didn't). That would require other vendors to adopt an AMD like GCN architecture.

As much as AMD would love to have everyone by the cajones in doing that, it just isn't going to happen. Intel and NV aren't going to develop AMD based GCN hardware just to use Mantle, even if it were open. But it isn't open in any sense of the word, anyway, so it's a moot point on both the hardware and software fronts.

Until they actually release the SDK, we don't know if this is true or not. For instance, Dice said that Nvidia's architecture already supports most of what you need for Mantle. So AMD could offer different levels of compatibility, much like dx12. AMD could also try and make a play into the mobile space where they may have more luck convincing others of Mantle's benefits. Of course ,this all could be made moot depending on what happens with dx12, we won't know until we get there.

 

[Despoiler]

Either way, Nvidia most likely would have to adjust their architecture to take full advantage of Mantle and I don't see them doing that, even if they had the opportunity, not with dx12 coming. Unless Mantle becomes a de facto standard in the industry (doubtful), I don't think you will see Mantle running on Nvidia hardware.

They wouldn't have to make any architecture changes. Mantle's just has to be able to A) recognize their hardware Nvidia wants to support and then b) map hardware functions to the API calls, C) have a driver. The reason why it probably won't happen isn't technical it's that Nvidia may have to reveal architectural details to AMD as they control the source code. That and they would be supporting a competitor's idea.

 

[Ed1]

More specifically, [H] said that their card was going above the 84 C thermal limit implemented by previous drivers, so without an increase in the thermal limit, pushing the card harder wouldn't result in higher temps because it would throttle to maintain 84 C. This would also allow [theoretically] for more than ~13 MHz of increased performance if it was able to maintain a higher boost for longer due to the increased temp limit.



I wouldn't be surprised if it was just some bug in the driver or something due to their particular setup or a combination of factors, would be nice for them or someone else to really test it though and show some temp/clocks over time. Either way, I see it as a minor thing, just wanted to mention that it didn't come out of thin air and would like someone to follow up on it with some actual testing is all.

All things equal once you get into high temps (above 70's , going higher on temps will lower clock speed (13mhz) unless you raise temp slider on boost2 . Even then higher temps = lower clocks (if you cool it below 70 it run faster) .

This is just some glitch IMO in driver, I doubt Nvidia did any thing on purpose to try and trick higher clocks . Such small increase and again first I have heard of that .

 

[Hitman928]

They wouldn't have to make any architecture changes. Mantle's just has to be able to A) recognize their hardware Nvidia wants to support and then b) map hardware functions to the API calls, C) have a driver. The reason why it probably won't happen isn't technical it's that Nvidia may have to reveal architectural details to AMD as they control the source code. That and they would be supporting a competitor's idea.

6 of one, half a dozen of another. Either nvidia would have to adjust their architecture for mantle or amd would have to adjust mantle to nvidia architecture, getting nvidia documentation in the process. I don't see either as being likely at this point.

 

[Hitman928]

All things equal once you get into high temps (above 70's , going higher on temps will lower clock speed (13mhz) unless you raise temp slider on boost2 . Even then higher temps = lower clocks (if you cool it below 70 it run faster) .

This is just some glitch IMO in driver, I doubt Nvidia did any thing on purpose to try and trick higher clocks . Such small increase and again first I have heard of that .

So nvidia cards start to thermally regulate when temps are in the 70's? Didn't know that, seems pretty conservative for a modern process.

 

[Ed1]

So nvidia cards start to thermally regulate when temps are in the 70's? Didn't know that, seems pretty conservative for a modern process.

Kind of , boost1 (6xx series) once you get to 70c down one step (13mhz) , then at 80c another down 13mhz , so with 6xx you want cool temps .

With 7xx they went to boost2 with option to control temp limits so if you raise limit your raise where it throttles instead of old temp bias .

So with older boost1 (6xx) when you OC you might throttle back in clock either from TDP power limit or temp limit point . so they linked it with boost2 to simplify .

 

[guskline]

All things equal once you get into high temps (above 70's , going higher on temps will lower clock speed (13mhz) unless you raise temp slider on boost2 . Even then higher temps = lower clocks (if you cool it below 70 it run faster) .

This is just some glitch IMO in driver, I doubt Nvidia did any thing on purpose to try and trick higher clocks . Such small increase and again first I have heard of that .

Water cooling my GTX780 really negates the thermal issue.

 

[Deders]

Mine never goes above 68c on air

 

[Carfax83]

Unfair? The responsibility of both companies to serve their user base on whatever platforms they support. As both companies support Win7 and Win8 the performance on both platforms matters. You can't honestly be arguing that Win7 users should totally disregard the system performance benchmarks that matter to them and only consider Win8 performance. Nvidia framed this driver as competing with Mantle, but I guess what you are saying is they need a footnote to those claims *Except in Win7 where we accept total defeat.

I think you need to reread what triggered this conversation. One member said that BF4 was faster on NVidia than AMD despite Mantle, and another member disagreed.

So we're not discussing platforms here, only raw performance. And in that context, it becomes appropriate to compare cards in their best light. AMD with Mantle, and NVidia with DX11.1..

Anything else is disingenuous..

 

[Ed1]

Mine never goes above 68c on air

Yes, there no problem if you get a dual fan cooler . My 660ti PE/OC was like 68-70c stock.
all I did is make custom fan profile and it doesn't even hit 60c hardly .