The AMD Driver Thread

[Sugarhell]

I don't have any of those Nvidia games, those games have a DrawCall problem, if they are using 15.15 on a 290X they will see an improvements in performance as that driver has better overheads.

But none of that would have anything to do with the GPU its self.

What i did was take a game and an application to address the Tessellation specifically.

Yeah you compared your card with 15.5 vs 15.15

Then you show 3dguru benches that compare a ref 290x with old drivers vs 390x with 15.15.-so much gain.

Your conclusion. Grenada is not hawaii. So much logic. I told you bioshock is not a tesselation hog game. Try crysis 2 or 3

 

[Davedree]

I don't have any of those Nvidia games, those games have a DrawCall problem, if they are using 15.15 on a 290X they will see an improvements in performance as that driver has better overheads.

But none of that would have anything to do with the GPU its self.

What i did was take a game and an application to address the Tessellation specifically. those are my results, take them or leave them.

290's do see improvements with the drivers from the later drivers.
Your tonga gcn 1.2 theory is way off though fella.

I need time to look into your results I'm not dismissing them.

 

[humbug]

290's do see improvements with the drivers from the later drivers.
Your tonga gcn 1.2 theory is way off though fella.

I need time to look into your results I'm not dismissing them.

Well, you saying without even looking at the results, you are not going to retract that now are ye?

Your conclusion. Grenada is not hawaii. So much logic. I told you bioshock is not a tesselation hog game. Try crysis 2 or 3

@ SugarHell, i have have both, if i do test them will you concede?

 

[Davedree]

Well, you saying without even looking at the results, you are not going to retract that now are ye?

Humbug I'm very honest, if i'm wrong i will admit it and say it on here, I have looked at them but I need time to look at them properly as i'm at work.
But this grenada, hawaii not a rebrand has been discussed days ago.

Remember I said Fury behaves like tonga when compared to tahiti in some games, (pointing towards a front end issue). This is related too and a lot of the drivers provided for the reviews weren't even 15.15/15.20 etc they were 15.5.

 

[Sugarhell]

Well, you saying without even looking at the results, you are not going to retract that now are ye?



@ SugarHell, i have have both, if i do test them will you concede?

Same clocks,same machine,same spot,same drivers

 

[humbug]

Humbug I'm very honest, if i'm wrong i will admit it and say it on here, I have looked at them but I need time to look at them properly as i'm at work.
But this grenada, hawaii not a rebrand has been discussed days ago.

Remember I said Fury behaves like tonga when compared to tahiti in some games, (pointing towards a front end issue). This is related too and a lot of the drivers provided for the reviews weren't even 15.15/15.20 etc they were 15.5.

I gave you a perfectly logical reason as to why someone would see an improvement using the 15.15 drivers on the 290X. and then went on to explain what was different about that and what i did differently.

You dismissed it and were somewhat derogatory about it, fella. already decided that you were right and i was wrong... without even looking at it.

So do understand that my confidence in your honesty is a bit low right now.

Same clocks,same machine,same spot,same drivers

Is that a yes? what do you mean by that?

 

[Dygaza]

Hardocp apples to apples: 290x and 390x clock to clock test

http://www.hardocp.com/article/2015/06/18/msi_r9_390x_gaming_8g_video_card_review/9#.VY2gr_mqpBd

 

[Sugarhell]

Hardocp apples to apples: 290x and 390x clock to clock test

http://www.hardocp.com/article/2015/06/18/msi_r9_390x_gaming_8g_video_card_review/9#.VY2gr_mqpBd

i post it already. But according to humbug its not a tesselation game

 

[humbug]

Hardocp apples to apples: 290x and 390x both on 15.15 drivers.

http://www.hardocp.com/article/2015/06/18/msi_r9_390x_gaming_8g_video_card_review/9#.VY2gr_mqpBd

same one Sugarhell already posted, the 15.15 Driver improves DrawCall performance, that has nothing to do with the GPU its self.

For Differences in the GPU you have to look for something like Tessellation, proven to be much better on GCN 1.2. with hairworks off none of those games are heavy on tessellation, they are on drawcalls.

So i used a Game that gives a GCN 1.2 GPU a huge advantage from increased tessellation, that game is BioShock Infinite, and i used Tessmark.

The performance on the 390X is 30% higher than the 290X in Bioshock, using 15.15 on the 290X results in 0 improvement in that game.

 

[TheRealDeal]

Hardocp apples to apples: 290x and 390x clock to clock test

http://www.hardocp.com/article/2015/06/18/msi_r9_390x_gaming_8g_video_card_review/9#.VY2gr_mqpBd

The problem i see with this is that they are not using the drivers designed for the 390x (15.15). Until we see a proper comparison on these drivers then my mind is not made up.

 

[humbug]

The thing is one cannot explain away a 30% increase in performance from the same reviewer apparently using the same 15.15 Driver on both GPU's
Ignore everything i did if you like ^^^ but that....

A lot of reviewers are saying they are the same GPU and most of them are trying to hammer that point home using a method that does not prove it, not by a long shot.

DrawCalls have absolutely nothing to do with the GPU architecture, Hardware journalism these days is in the gutter.

 

[Sugarhell]

The thing is one cannot explain away a 30% increase in performance from the same reviewer apparently using the same 15.15 Driver on both GPU's
Ignore everything i did if you like ^^^ but that....

A lot of reviewers are saying they are the same GPU and most of them are trying to hammer that point home using a method that does not prove it, not by a long shot.

DrawCalls have absolutely nothing to do with the GPU architecture, Hardware journalism these days is in the gutter.

Where is the 30% increase lol? Against the ref thermal throttled 290x? ROfl humbug

 

[Dygaza]

same one Sugarhell already posted, the 15.15 Driver improves DrawCall performance, that has nothing to do with the GPU its self.

For Differences in the GPU you have to look for something like Tessellation, proven to be much better on GCN 1.2. with hairworks off none of those games are heavy on tessellation, they are on drawcalls.

So i used a Game that gives a GCN 1.2 GPU a huge advantage from increased tessellation, that game is BioShock Infinite, and i used Tessmark.

The performance on the 390X is 30% higher than the 290X in Bioshock, using 15.15 on the 290X results in 0 improvement in that game.

Yes, but that guru3d situation where they get that 30% upgrade isn't just from tesselation alone, but also from drawcalls. As they are clearly cpu bound. (Or should we call API bound now?).

I don't believe for a second there is GCN 1.2 tesselators included in grenada. Actually shouldn't that already show in bios. According "The Stilt" he didn't found any real big differencies from bios. Only mentionable differences were memory configs and binning. And in CAC leakage.

http://murobbs.muropaketti.com/posts/1715467766/
use google translator if you trust it's translation from finnish (I don't).

I don't know if 390x can have better tesselation just because of more matured process and different binning.

 

[Mei]

matt stop playing with ur furrys and give me more ark fps!
grr

 

[LtMatt]

matt stop playing with ur furrys and give me more ark fps!
grr

What exactly are you looking for Mei?

 

[Davedree]

same one Sugarhell already posted, the 15.15 Driver improves DrawCall performance, that has nothing to do with the GPU its self.

For Differences in the GPU you have to look for something like Tessellation, proven to be much better on GCN 1.2. with hairworks off none of those games are heavy on tessellation, they are on drawcalls.

So i used a Game that gives a GCN 1.2 GPU a huge advantage from increased tessellation, that game is BioShock Infinite, and i used Tessmark.

The performance on the 390X is 30% higher than the 290X in Bioshock, using 15.15 on the 290X results in 0 improvement in that game.

Humbug ok i've looked at your tessmarks and as noted you saw 15%@15.15 and 18% 15.20 tesselation improvements. You then compared bench results with the drivers and didn't see any gains in the bioshock benchmark on your system and also are comparing your results to the guru3d results. Therefore you are thinking that the grenada has to be something more than gcn1.1. Right welL Nutella had the same theory and idea that it was gcn 1.2 a few days ago and we debunked it but over the witcher 3 tesselation difference.
So trust me this is de ja vu.

Ok back to the guru3d results first,
that bioshock which you are using as your reference for the large gap in performance you forgot to mention this at the bottom of the page

So they are comparing a reference throttling 290x, on older drivers than 15.15 which were used for the 390x only, and with slower clockspeeds and timings on the 4gb of ram.

Now we need to test that with an overclocked 290x at 1100/1525 8gb vs a 390x 1100 1525 8gb, like they said in their disclaimer, (you see that big gap is rapdily shrinking now) was 290x- 62 - 390x - 81 ( predicted in the 70's on the same ref throttling 290x 4gb).

Now with your tessmark results they are fantastic but there is one fundamental problem, you aren't providing a 390 (2560-same clocks and drivers with 4gb to compare to, so whilst you are seeing tesselation improvements it doesn't mean there aren't others factors involved in your lack of game improvement by using the drivers.

Let's not forget that your system could well be holding back your bioshock results as your fx9590 is not running at it's best and also the fx9590 isn't the greatest cpu for testing in some games. If so could you try and overclock your gpu and see if you gain any fps with the cpu settings as before in previous tests but also you need a 390 in your system to compare to for a fair test..

EDITED FOR THE 62 VS 81 gap

 

[Mei]

haha extra!
bronto farming = *sad face*

 

[humbug]

Yes, but that guru3d situation where they get that 30% upgrade isn't just from tesselation alone, but also from drawcalls. As they are clearly cpu bound. (Or should we call API bound now?).

I don't believe for a second there is GCN 1.2 tesselators included in grenada. Actually shouldn't that already show in bios. According "The Stilt" he didn't found any real big differencies from bios. Only mentionable differences were memory configs and binning. And in CAC leakage.

http://murobbs.muropaketti.com/posts/1715467766/
use google translator if you trust it's translation from finnish (I don't).

I don't know if 390x can have better tesselation just because of more matured process and different binning.

If that Driver made no difference to me on an AMD CPU then why should it make a difference to them on an Intel CPU, are Intel CPU's suddenly slower than AMD CPU's?

Humbug ok i've looked at your tessmarks and as noted you saw 15%@15.15 and 18% 15.20 tesselation improvements. You then compared bench results with the drivers and didn't see any gains in the bioshock benchmark on your system and also are comparing your results to the guru3d results. Therefore you are thinking that the grenada has to be something more than gcn1.1. Right welL Nutella had the same theory and idea that it was gcn 1.2 a few days ago and we debunked it but over the witcher 3 tesselation difference.
So trust me this is de ja vu.

Ok back to the guru3d results first,
that bioshock which you are using as your reference for the large gap in performance you forgot to mention this at the bottom of the page

So they are comparing a reference throttling 290x, on older drivers than 15.15 which were used for the 390x only, and with slower clockspeeds and timings on the 4gb of ram.

Now we need to test that with an overclocked 290x at 1100/1525 8gb vs a 390x 1100 1525 8gb, like they said in their disclaimer, (you see that big gap is rapdily shrinking now) was 290x- 62 - 390x - 81 ( predicted in the 70's on the same ref throttling 290x 4gb).

Now with your tessmark results they are fantastic but there is one fundamental problem, you aren't providing a 390 (2560-same clocks and drivers with 4gb to compare to, so whilst you are seeing tesselation improvements it doesn't mean there aren't others factors involved in your lack of game improvement by using the drivers.

Let's not forget that your system could well be holding back your bioshock results as your fx9590 is not running at it's best and also the fx9590 isn't the greatest cpu for testing in some games. If so could you try and overclock your gpu and see if you gain any fps with the cpu settings as before in previous tests but also you need a 390 in your system to compare to for a fair test..

EDITED FOR THE 62 VS 81 gap

Read my reply to Dygaza, and:

yes they tested it "on an older driver set than 15.15" thats what i have been doing.

what you are trying to say here is 15% Tessellation and and some Reference cooler throttling amounts to 30%.

How do you even divide that up?

So 15% points, that would mean that this 290 was running at no more than 850Mhz maximum, lets just assume for a minute this is true, wow. but ok lets entertain it for a bit.

You're left with the other 15%, so 15% Tessellation = 15% FPS, you know for a fact its nothing like that, its doesn't work anything like that, if it did Nvidia would consistently be 400% ahead of AMD, AMD's own R9 285 would be consistently be 400% ahead of all their other GPU's. no. that fact is this is what it takes in Tessellation to get another 10 FPS.

I think What you are doing here now is looking for any reason possible to argue that you were right all along, its all you can do now after calling me out without knowing anything about what is actually going on, you had already accepted one version as a truth and are completely closed to anyother possibility, you are pooling every little thing that you can and exaggerating its impact massively to make up the numbers.

What i want is to get to the truth, so here is my challenge to you, i'm satisfied from the data thats out there and my own tests that the 390X is not just a 290X put in a new box, the performance difference between these two cards is huge right there where you would expect it to be, of course you disagree with that and are trying very hard to explain away all of it.

Do you know what, don't do that, its far to easy to say to someone with hard evidence in proving their point "oh i don't believe you, your wrong" now i figured out a way to test the theory and those are my results, this is a practical experiment that yielded practical results, i'm willing to accept that my results might be wrong, but just sit shooting at the messenger, really? why on earth would you do that?

#It can't be down to CPU bottlenecking with the 15.15 Driver improving it, because id that is what they are experiencing on an Intel CPU then i sure as hell should be, and yet clearly i'm not, i also tested it at 3200x1800P as well as 1920x1080P, its cut my FPS in half from 1080P
#+15% Tessellation i'll give you the benefit of the doubt with this, i think you do understand it better than this, but if you don't Tessellation does not scale FPS anything like this, 15% does not = 30% FPS, its not 15% FPS. its not even 5% or even 2%.
#Thermal throttling, its Throttling by about 30%? i bet if i pull up an AIB 290X the performance result will be pretty much the same, you can do that yourself, go have a look...

I spent a lot of time calling the 390X a total rebox, just like everyone did, cursing it, you name it. that was until i calmed down a bit and thought about it, did a little research and this a test.

Don't be a part of the cesspit politics that fly around far to much in here, hit the reset switch in your head, if you still think i'm wrong, thats fine. prove it. actually prove it instead of just blasting no no no no no at me.

I say again, all i want is to get to the truth, i'd love the idea that i can give my 290 a huge performance boost just by switching drivers.

 

[Davedree]

If that Driver made no difference to me on an AMD CPU then why should it make a difference to them on an Intel CPU, are Intel CPU's suddenly slower than AMD CPU's?

The one game which you decided to use as a conclusion to your studies was Bioshock. You haven't provided evidence of any other games (in particular games with far more tesselation than the (Bioshock is a pretty much a gpu bound game anyway)
I didn't state in my previous post that the whole performance was based on your cpu did I?

Read my reply to Dygaza, and:

what you are trying to say here is 15% Tessellation and and some Reference cooler bottlenecking amounts to 30%.

No you are trying to say that. I'm correcting you when you forgot to realise that there is a difference in driver performance. What you are not getting despite being told numerous times is that 30% is a figure you have made a conclusion out by one reviewer, guru3d and you've stuck to it. We have told you time and time again that the guru3d 290x 62fps vs 81 fps 390x isn't a true representation of a 30% increase because of the factors we've all mentioned. You will not listen to us and you keep going in circles. Despite the other reviewers showing -/+ 15% thereabouts with the driver disparity still in effect. We also discussed this a week before you so don't get all defensive when we have been through the same discussions.

So you provided us the 4 benches you chose to base your results on as conclusive. 1440p Results in red below, you chose to use them so lets use them.

Well thats one, here are the first 4 i pulled from Google all of which confirm what i just tested.

Hawaii in original form did not just get a 30% boost in performance and its not the Drivers that do it, sorry, but your link is a nonsense.


http://www.guru3d.com/articles-pages/msi-radeon-r9-390x-gaming-8g-oc-review,18.html
Guru3d saw 30%
ref 290x vs 390x 81fps-62fps/62*100= 30%

http://www.overclock3d.net/reviews/gpu_displays/msi_r9_390x_gaming_8g_review/11
(take your pick theres a variety of 290s which just proves my point, that the 30% isn't reflective)

http://hexus.net/tech/reviews/graphics/84194-sapphire-radeon-r9-390x-tri-x/?page=5
ref 290x vs 390x saph tri-x 69fps-78fps = 13%

https://www.techpowerup.com/reviews/MSI/R9_390X_Gaming/11.html
ref 290x vs msi gaming 390x 92.6 -105fps =12.3%

How do you even divide that up?

So 15% points, that would mean that this 290 was running at no more than 800Mhz maximum, lets just assume for a minute this is true, wow. but ok lets entertain it for a bit.

You're left with the other 15%, so 15% Tessellation = 15% FPS, you know for a fact its nothing like that, its doesn't work anything like that, if it did Nvidia would consistently be 400% ahead of AMD, AMD's own R9 285 would be consistently be 400% ahead of all their other GPU's. no. that fact is this is what it takes in Tessellation to get another 10 FPS.

So you started off quoting tessmark as an indication of throughput for tesselation on your system. Quoting ''The R9 300 Series Drivers may give the R9 200 Series GPU a little Tessellation boost, but i would suggest thats simply because that driver is designed for the R9 300 Series Tessellation engine which to me looks like a different GPU, it looks to me like a GCN 1.2 GPU''.

Ok let's put it here with some interesting plots.


1.Ok so if the 390x is a Gcn1.2 then explain why in this report that Sugarhell sent to you that you rejected calling his link ''nonsense'', why in this report does it show the scaling of tesselation between a 290x a 390x and a r9 285 and a r9 380, using the 2 different drivers 15.5 / 15.15.
2. The r9 290x is quite a bit below the 390x on 15.5 but at 15.15 it still slighly behind from 1x-8x then they're almost remain in parity beyond 8x.
Hardly anything to shout about but a difference which could be attributed to clocks speeds and internal timings, including the theory about overclocking the tesselator engine on Grenada.
3. The r9 390x if it had massively improved tesselation would be showing far greater gains than the r9 290x with 15.15, and also the r9 285/380 tonga throughout the test range.
4. Tonga beyond 16x has an impressive gain over hawaii/grenada, but it's not so rosey in the lower 1-8x in comparison to Hawaii.
5. Anandtech came to the same conclusion of the differences between tonga and hawaii tesselation. Here's their findings

I think What you are doing here now is looking for any reason possible to argue that you were right all along, its all you can do now after calling me out without knowing anything about what is actually going on, you had already accepted one version as a truth and are completely close to anyother possibility, you are pooling every little thing that you can and exaggerating its impact massively to make up the numbers.

What i want is to get to the truth, so here is my challenge to you, i'm satisfied from the data thats out there and my own tests that the 390X is not just a 290X put in a new box, the performance difference between these two cards is huge right there where you would expect it to be, of course you disagree with that and are trying very hard to explain away all of it.



Do you know what, don't do that, its far to easy to say to someone with hard evidence in proving their point "oh i don't believe you, your wrong" now i figured out a way to test the theory and those are my results, this is a practical experiment that yielded practical results, i'm willing to accept that my results might be wrong, but just sit shooting at the messenger, really? why on earth would you do that?

1.You needed to be aware of factors you weren't accounting for, the clock difference, the driver difference that you didn't see in the guru3d reporting 290x was on old drivers, the range of games the settings, the different timings between the memory.
2.You were basing your 30% as the definitive performance improvement and all your posts reflect that even this one, 30% isn't the true performance difference because of the factors we kept reminding you off. It is more 13-15% and then it's less once we work out the factors.
3.Why you are getting all weird like it's some personal attack I don't know?
I'm not hear to argue any possible way that I was right you were wrong, you just need to admit that you made some mistakes, but you take it personally and then get all defensive trying to hide the fact you made a mistake. You've been caught out before on this. I'm here to fact find and provide an insight into finding the truth and from the previous posts that i have commented on the grenada/hawaii theory you will see we discussed it way before you did.
4. This is what i said ''Grenada is simply a tweaked refreshed Hawaii. Nothing more. It's not a respin with gcn 1.2 otherwise we would see an updated uvd engine. Also if it was tonga then how could people put a bios from a 390 onto the 290. It's simply taking the r9 290x 8gb, and using that for a 2560,2816 390/390x.
They've then played around with internal timings/memory timings and put a bios string on so the recent wonder drivers don't support a 290, they've played with the clocks speeds, but mostly it comes from the drivers.

You see you ignored a lot of my points in that earlier post.

#It can't be down to CPU bottlenecking with the 15.15 Driver improving it, because id that is what they are experiencing on an Intel CPU then i sure as hell should be, and yet clearly i'm not
#+15% Tessellation i'll give you the benefit of the doubt with this, i think you do understand it better than this, but if you don't Tessellation does not scale FPS anything like this, 15% does not = 30% FPS, its not 15% FPS. its not even 5% or even 2%.
#Thermal throttling, its Throttling by about 30%? i bet if i pull up an AIB 290X the performance result will be pretty much the same, you can do that yourself, go have a look...

Like i've said above you go on about the 30% but thats a ref throttling 290x on guru3d, the other reviwers show 13-15%, and the oc3d has a variation of aib 290s which demonstrate my point as they all deliver different fps.

I spent a lot of time calling the 390X a total rebox, just like everyone did, cursing it, you name it. that was until i calmed down a bit and though about it, did a little research and this a test.

Don't be a part of the cesspit politics that fly around far to much in here, hit the reset switch in your head, if you still think i'm wrong, thats fine. prove it. actually prove it instead of just blasting no no no no no at me.

I say again, all i want is to get to the truth,

We have been a week before you, but don't take it personally, you provided in house benchmarks and I appreciate the time and effort with that.
However there's more to it and more testing needs to be carried out before we can come to a conclusion.

 

[humbug]

Actually i provided 4 instances of large performance differences in BioShock, not 1, so i'm not basing it off just one. i'm basing it off what is a common trend

As for your PCHarware slides, they show the same level of results i did, its hardly surprising to me, its a small difference simply confirming what i found and reported, it doesn't change the large performance discrepancies between the two in BioShock, i had the same results, it did not translate.

Your repeating the same thing again, only this time as a huge wall of text, which i did read.

Prove me wrong, don't repeat the same thing over and over again getting more and more elaborate each time.

We have been a week before you, but don't take it personally, you provided in house benchmarks and I appreciate the time and effort with that.
However there's more to it and more testing needs to be carried out before we can come to a conclusion.

I read that its all talk no prof of concept and you seem to be taking this as a front to it

Actually prove there my results as inaccurate, less talk more doing.