Thursday, August 3rd 2017

AMD RX Vega 56 Benchmarks Leaked - An (Unverified) GTX 1070 Killer

TweakTown has put forth an article wherein they claim to have received info from industry insiders regarding the upcoming Vega 56's performance. Remember that Vega 56 is the slightly cut-down version of the flagship Vega 64, counting with 56 next-generation compute units (NGCUs) instead of Vega 64's, well, 64. This means that while the Vega 64 has the full complement of 4,096 Stream processors, 256 TMUs, 64 ROPs, and a 2048-bit wide 8 GB HBM2 memory pool offering 484 GB/s of bandwidth, Vega 56 makes do with 3,548 Stream processors,192 TMUs, 64 ROPs, the same 8 GB of HBM2 memory and a slightly lower memory bandwidth at 410 GB/s.

The Vega 56 has been announced to retail for about $399, or $499 with one of AMD's new (famous or infamous, depends on your mileage) Radeon Packs. The RX Vega 56 card was running on a system configured with an Intel Core i7-7700K @ 4.2GHz, 16 GB of DDR4-3000 MHz RAM, and Windows 10 at 2560 x 1440 resolution.
The results in a number of popular games were as follows:

Battlefield 1 (Ultra settings): 95.4 FPS (GTX 1070: 72.2 FPS; 32% in favor of Vega 56)
Civilization 6 (Ultra settings, 4x MSAA): 85.1 FPS (GTX 1070: 72.2 FPS; 17% in favor of Vega 56)
DOOM (Ultra settings, 8x TSAA): 101.2 FPS (GTX 1070: 84.6 FPS; 20% in favor of Vega 56)
Call of Duty: Infinite Warfare (High preset): 99.9 FPS (GTX 1070: 92.1 FPS; 8% in favor of Vega 56)

If these numbers ring true, this means NVIDIA's GTX 1070, whose average pricing stands at around $460, will have a much reduced value proposition compared to the RX Vega 56. The AMD contender (which did arrive a year after NVIDIA's Pascal-based cards) delivers around 20% better performance (at least in the admittedly sparse games line-up), while costing around 15% less in greenbacks. Coupled with a lower cost of entry for a FreeSync monitor, and the possibility for users to get even more value out of a particular Radeon Pack they're eyeing, this could potentially be a killer deal. However, I'd recommend you wait for independent, confirmed benchmarks and reviews in controlled environments. I dare to suggest you won't need to look much further than your favorite tech site on the internet for that, when the time comes. Source: TweakTown
Add your own comment

169 Comments on AMD RX Vega 56 Benchmarks Leaked - An (Unverified) GTX 1070 Killer

#76
bug
"RejZoR said:
Of course there's more discussion about Vega in recent months. Why would I talk about a year old product (Pascal) that I don't care since it has all been discussed like trillion times already? VEGA is the shit everyone talks about now. Including me, you and everyone else.
Ok, try this then: in how many posts were you critical of Vega?
Posted on Reply
#77
the54thvoid
"bug said:
Ok, try this then: in how many posts were you critical of Vega?
He can't be critical of Vega - it's not been released yet...
Posted on Reply
#78
oxidized
"Vya Domus said:
I have no idea how they payed anyone , but one thing is clear : they did their best to shove things like GameWorks and other things ( which are in my opinion shady tactics ) such as making developers put ludicrous amounts of tessellation because they knew AMD wasn't as efficient at it.

You may think these are perfectly legit methods but that doesn't change the fact that yes , Nvidia did do their best to make AMD look bad.
But we're talking about reviews here not nvidia, it's pretty reasonable that a company like nvidia or even amd, will try to make the other look bad, it's how it works, and it's a completely different thing from "nvidia is bribing reviewers, and review websites", completely different, so in the end you'll have a game that runs better on nvidia rather than amd, and the reason can be whatever you want it to be, but the fact remains, that game ACTUALLY runs better on nvidia rather than amd, same story if AMD did that, for example doom, only has vulkan and opengl, but why exactly? I'm pretty sure that if it ran on dx11 nvidia would have the upper hand there too. Needless to remind that vulkan is pretty much an iteration of mantle, which was used on bf4? or 3 i can't remember. and boosted amd performance, so what are we talking about here?


You people should really stop evaluating companies based on their behaviour, they're lucrative companies, they'll always do shady things no matter nvidia amd intel or whatever you want, there's nobody better than the other in doing such, there's only who does a better job in terms of product, and since one can't rely on their tests, because ofc they'll be biased, one has to rely on 3d party tests or even personal tests.
Posted on Reply
#79
bug
"the54thvoid said:
He can't be critical of Vega - it's not been released yet...
It's been announced it will draw a lot of power, it's been announced that far from offering breakthrough performance it will offer something Nvidia gave us last year and it's been announced it will not be particularly cheap either. I think there were enough reasons to at least express concern for someone not explicitly rooting for AMD.
Posted on Reply
#80
Vya Domus
"oxidized said:
But we're talking about reviews here not nvidia, it's pretty reasonable that a company like nvidia or even amd, will try to make the other look bad, it's how it works, and it's a completely different thing from "nvidia is bribing reviewers
But if say even as little as 1/10 games featured in a review have gone through "Nvidia's handy optimizations" , doesn't that mean this bias gets carried through ?

"oxidized said:
so in the end you'll have a game that runs better on nvidia rather than amd, and the reason can be whatever you want it to be, but the fact remains, that game ACTUALLY runs better on nvidia rather than amd
OK so you're fine with software and hardware that is gimped on purpose as long as it gives you the illusion of better ? Whatever floats your boat I guess.

"oxidized said:
same story if AMD did that, for example doom, only has vulkan and opengl, but why exactly? I'm pretty sure that if it ran on dx11 nvidia would have the upper hand there too? Needless to remind that vulkan is pretty much an iteration on mantle, which was used on bf4? or 3 i can't remember. and boosted amd performance, so what are we talking about here?
Every single time someone goes out to mention how Vulkan/Doom is biased towards AMD tells me how much do they really understand from all of this.

I have already gone through this in a recent discussion about how Vulkan DOSE NOT favor anyone. Nvidia is at the top of charts aren't they ? Did they need DX 11 ? No , because Vulkan is great at taking advantage of modern hardware. DX11 doesn't do that anymore.
Posted on Reply
#81
Sasqui
Ok, I'm off to sell my brand new GTX 1070.

Seriously...
Posted on Reply
#82
B-Real
"oxidized said:

Now i'd been reading all kind of reviews in the months before buying the new card, and in pretty much ALL of them the 1060 was faster on most of the games tried in the benchmark it was something like 60/40 in favour of nvidia, but the price of that gtr black edition was just too good, and you know what happened next. So what?
Well, it's very easy to answer: You looked at the initial reviews.
Actually since the december re-tests of RX480 vs 1060, they are equal in DX11. Check the reviews, usually on YT but there are updated reviews on sites like Hardware Canuks, for example. The tipical ~10% advantage of the 1060 in DX11 titles has evaporated to 0-2%, not to mention the DX12 advantage of RX480. So yeah, you were unlucky with the coil whine RX480, but overall, you got an overall bit worse card (for more money).
Posted on Reply
#83
neatfeatguy
Just like all pre-released info on a new GPU lineup - I'll just bypass the info here and wait for actual benchmarks from reliable sites once NDA is done.

I do, however, greatly enjoy reading the bias response people have on both sides of the fence when going off of "leaked" or "rumored" data. You get those that will defend the data to the end and those that will dispute it with all their might. Most of the comments are laughable and make it for good reading material.
Posted on Reply
#84
bug
"Sasqui said:
Ok, I'm off to sell my brand new GTX 1070.

Seriously...
Because of 4 leaked benchmarks? You know what they say about a fool and their money...
Posted on Reply
#85
Fouquin
"bug said:
Because of 4 leaked benchmarks? You know what they say about a fool and their money...
I'm fairly certain the "Seriously..." was in an exasperated tone in response to the cherry picked benchmarks, and that they were being sarcastic about selling their brand new GTX 1070.

I could be wrong, but it fits the tone of the rest of the comments on this article.
Posted on Reply
#86
Th3pwn3r
"RejZoR said:
Ok, if being quiet and not bitching over AMD all the time automatically means I'm favoring AMD, so one would assume I'm constantly saying how garbage NVIDIA is, right? Well, good luck finding that, coz you won't find any (and don't pull stuff out of context). Only time I ever mentioned downsides about NVIDIA was about particular features or FX and Kepler series. And how Pascal is ridiculously fast, but doesn't really bring any tech that would excite me. It's just a very fast card. And that's about it. Wouldn't that, I don't know, kinda make me you know, neutral? Just because I can find positives in otherwise underwhelming launch of RX Vega, that doesn't mean I'm a fanboy.

It almost makes me want to buy RX Vega just to piss people off here at TPU. Seeing all of you implode here would be the best s***t ever.
Lol, you have changed your mind at least five times about getting Vega. You even made an upset post that you were getting a 1080ti and that was it. Then you said you were getting the liquid cooled Vega...blah,blah.

I don't think you're too pro AMD but you sure are extremely indecisive.

I on the other hand will buy Vega if it can beat the 1080ti. If not I'll buy another Nividia card(1080ti).
Posted on Reply
#87
Dimi
There is no Vega 56 at 399$ available. There probably NEVER will be due to mining.

Hell average price for 580's is 500$!

Besides, who wants to buy crap cooled blower style standard editions anyway.

Half a dozen aftermarket 1070's were available for 349.99 or lower at one point. Don't forget this.
Posted on Reply
#88
ppn
Maybe if it can get closer to 1080 with memory OC to 512GB/s compared to default of 410 that is 10% performance for free, the 399$ mark is perfect for a new product. GTX 1070 will be the new 250$ card soon enough so no point comparing it to that.
Posted on Reply
#89
Vya Domus
"Dimi said:
Half a dozen aftermarket 1070's were available for 349.99 or lower at one point. Don't forget this.
Yeah , were. If Vega 56 pricing will get screwed up , it'll be just as screwed up as 1070 pricing.
Posted on Reply
#90
cdawall
where the hell are my stars
"bug said:
Because of 4 leaked benchmarks? You know what they say about a fool and their money...
To be fair I am selling off my 980Ti collection on ebay just in case the price tanks
Posted on Reply
#91
oxidized
"Vya Domus said:
But if say even as little as 1/10 games featured in a review have gone through "Nvidia's handy optimizations" , doesn't that mean this bias gets carried through ?



OK so you're fine with software and hardware that is gimped on purpose as long as it gives you the illusion of better ? Whatever floats your boat I guess.



Every single time someone goes out to mention how Vulkan/Doom is biased towards AMD tells me how much do they really understand from all of this.

I have already gone through this in a recent discussion about how Vulkan DOSE NOT favor anyone. Nvidia is at the top of charts aren't they ? Did they need DX 11 ? No , because Vulkan is great at taking advantage of modern hardware. DX11 doesn't do that anymore.
It's not an illusion of better, IT IS better, the way they achieved this isn't relevant. as long as they didn't to anything illegal, and they didn't last i checked.
Vulkan, and Doom specifically run much better on AMD hardware, i understood everything i assure you.
Vulkan favors AMD whether they like it or not, so much they sold doom with amd cards for a while, what are we talking about cmon, are you only sharp when it suits you? Vulkan favors AMD even because (and i'll say it again) it's mantle's offspring.
I'd say pascal is more modern than polaris seen consuption and everything else related.

This last thing actually reminds me of those people that when it's pro AMD it's sure and can't be wrong, otherwise it's wrong, so if a game runs better on an AMD cards then it means the game is good, it's well optimized and well developed to gain from newer hardware, if the game runs better on nvidia but be badly optimized and doesn't scale well enough with new hardware.

So that's practically AMD fanboy, what else would you need?


"B-Real said:
Well, it's very easy to answer: You looked at the initial reviews.
Actually since the december re-tests of RX480 vs 1060, they are equal in DX11. Check the reviews, usually on YT but there are updated reviews on sites like Hardware Canuks, for example. The tipical ~10% advantage of the 1060 in DX11 titles has evaporated to 0-2%, not to mention the DX12 advantage of RX480. So yeah, you were unlucky with the coil whine RX480, but overall, you got an overall bit worse card (for more money).
No i looked even at the most recent reviews i could find, and the 1060 was still above, and on the same percentage, the only thing going really head to head with the 1060 is the 580, i looked at hardware canuck's re-review, and it's probably the only one that showed those shift, which i could find anywhere else besides, dx12 worked better on 480, that is true, and also you're probably true about saying i got the worst bang for the buck between those 2, but 1060 had more valid AIB choices and the overall consumption is lower on 1060, so in the end i'd say it's on par.
Posted on Reply
#92
Vya Domus
"oxidized said:
It's not an illusion of better, IT IS better, the way they achieved this isn't relevant. as long as they didn't to anything illegal, and they didn't last i checked.
It's not illegal but it hampers progress and it annoys me because it is not benefiting the consumer in way.

"oxidized said:

Vulkan, and Doom specifically run much better on AMD hardware, i understood everything i assure you.
Vulkan favors AMD whether they like it or not, so much they sold doom with amd cards for a while, what are we talking about cmon, are you only sharp when it suits you? Vulkan favors AMD even because (and i'll say it again) it's mantle's offspring, i'd say pascal is more modern than polaris seen consuption and everything else related.
You are simply wrong , like I said I already had a pretty lengthy discussion on this matter , not gonna have it again.

Vulkan does not favor AMD , hell think about this : The president of The Khronos Group who developed Vulkan works at Nvidia. Whoop de doo , you still think it favors AMD ? Nvidia along side other members and partners from all around the computer graphics technology world would let such a ridiculously biased thing happen ? Just because a piece of hardware performs better in a certain application doesn't mean that application favors it. Come on man. And do I have to remind you again that Titan Xp is at the top of the charts in Doom and Vega with the same raw power doesn't. Is that bias ?

"oxidized said:

This last thing actually reminds me of those people that when it's pro AMD it's sure and can't be wrong, otherwise it's wrong, so if a game runs better on an AMD cards then it means the game is good, it's well optimized and well developed to gain from newer hardware, if the game runs better on nvidia but be badly optimized and doesn't scale well enough with new hardware.

So that's practically AMD fanboy, what else would you need?
Doom is well optimized to run on any newer hardware , Nvidia and AMD.
Posted on Reply
#93
Nkd
"Assimilator said:
Do you even proofread? The word you're looking for is "median".

And you're also completely ignoring that GTX 1070's MSRP is $349, i.e. $50 less than Vega 56, which is extremely fair considering the (supposed) relative performance of these cards - in fact I'd say the 1070 still wins on price/performance if these Vega 56 numbers are truthful. So calling Vega 56 a "GTX 1070 killer" is laughable.

The price of GTX 1070 cards has only been pushed up because of the cryptomining BS. Vega 56 is unlikely to offer a better hashrate-per-watt than GTX 1070, which means GTX 1070 prices will stay high and they will continue to be bought in volume by miners, whereas Vega 56 will be bought in much smaller quantities by gamers. So NVIDIA still wins in terms of the pure numbers game, and therefore in revenue.

You can't blame NVIDIA that they made a great card that is in such high demand, regardless of the reason, that it commands a nearly 25% price premium over its MSRP.
1070 MSRP was never reduced it was gtx 1080 MSRP. for 1070 It was always 399.99, they did start going on sale before mining craze hit.
Posted on Reply
#94
B-Real
"Dimi said:
There is no Vega 56 at 399$ available. There probably NEVER will be due to mining.

Hell average price for 580's is 500$!

Besides, who wants to buy crap cooled blower style standard editions anyway.

Half a dozen aftermarket 1070's were available for 349.99 or lower at one point. Don't forget this.
So one on side, you are talking about Vega will not be available for 399$, because 580 IS 500$ (which is simply a lie, because you can get RX580s for 310-320$, but yeah, its way too expensive for the initial price) and on the other side you say that 1070s WERE available for $350 at one point. :D OMG, help me please.
Posted on Reply
#95
oxidized
"Vya Domus said:
It's not illegal but it hampers progress and it annoys me because it is not benefiting the consumer in way.
I'd say it hampers the competitor, not really the consumer directly, but i can give you that.

"Vya Domus said:

You are simply wrong , like I said I already had a pretty lengthy discussion on this matter , not gonna have it again.

Vulkan does not favor AMD , hell think about this : The president of The Khronos Group who developed Vulkan works at Nvidia. Whoop de doo , you still think it favors AMD ? Nvidia along side other members and partners from all around the computer graphics technology world would let such a ridiculously biased thing happen ? Just because a piece of hardware perform better in a certain application doesn't mean it favors it. Come on man.
And do I have to remind you again that Titan Xp is at the top of the charts in Doom and Vega with the same raw power doesn't. Is that bias ?
Means nothing, i'm just saying Vulkan favors AMD, not that it's meant to favor AMD, whether or not it's meant, it still favor greatly AMD cards performance, and not for the things you say, because there's really nothing superior on polaris compared to pascal, hell even Vega isn't superior to pascal if these latest results become true. TitanXp is a much much more powerful chip than any Vega, i take it you're not serious.

"Vya Domus said:

Doom is well optimized to run on any newer hardware , Nvidia and AMD.
Doom is excellently optimized, but still on par GPUs does much much better on AMD, it's AMD friendly environment if you don't want to call it bias.
Posted on Reply
#96
Prince Valiant
If the games AMD performs well on are AMD friendly that must mean that the games it performs bad on are Nvidia friendly.
Posted on Reply
#97
Vya Domus
"oxidized said:
TitanXp is a much much more powerful chip than any Vega, i take it you're not serious.
"oxidized said:
but still on par GPUs does much much better on AMD, it's AMD friendly environment if you don't want to call it bias.
Oh really , is that so ? Vega 64 has about 14 TFLOPS , right about where Titan Xp sits as well, die sizes are similar too. So yeah , they are on par in terms of power. Yet , if Vulkan favors "greatly" AMD how come it doesn't beat it ? You are on witch hunt mate , Doom/Vulkan is impartial to hardware.
Posted on Reply
#98
B-Real
"oxidized said:

No i looked even at the most recent reviews i could find, and the 1060 was still above, and on the same percentage, the only thing going really head to head with the 1060 is the 580, i looked at hardware canuck's re-review, and it's probably the only one that showed those shift, which i could find anywhere else besides, dx12 worked better on 480, that is true, and also you're probably true about saying i got the worst bang for the buck between those 2, but 1060 had more valid AIB choices and the overall consumption is lower on 1060, so in the end i'd say it's on par.
Then I have no idea what you checked. First is Hardware Canuks as a written review, YT videos like HW Unboxed (<div class="youtube-embed" data-id="CiYQqNiqQKU"><img src="https://i.ytimg.com/vi/CiYQqNiqQKU/hqdefault.jpg" /><div class="youtube-play"></div><a href="https://www.youtube.com/watch?v=CiYQqNiqQKU" target="_blank" class="youtube-title"></a></div>), getting 1% in favor of the 1060 and first it was a 12% difference. Yeah, it consumes 30-40W more at default (which is said by HW Unboxed that shouldn't be a deal breaker), but actually, JayzTwoCents' XFX 480 video shows that under stress test, that RX480 eats about 95 to 120W. And with Crimson, AB etc. you can control the more hungry RX480s like the Sapphire or MSI. Also, ones like MSI got later BIOS updates that got their power consumption lower.

So yeah, overall the RX480 (including the DX12 and Vulkan games) is a somewhat faster (or equal) card than the GTX 1060 since about 8 months now, and before that mining fever, it was about 30-40 bucks cheaper. At start, I would say they were egal (2016 summer), considering cost, performance, power consumption, but since decembers Crimson drivers, RX480 is simply the better choice. And for the RX580... yeah, its actually a faster card than the 1060. And it was sold for around 480 prices till the fever hit... Yeah, it consumed more than the RX480, it's for sure.
Posted on Reply
#99
Gasaraki
"Dimi said:
There is no Vega 56 at 399$ available. There probably NEVER will be due to mining.

Hell average price for 580's is 500$!

Besides, who wants to buy crap cooled blower style standard editions anyway.

Half a dozen aftermarket 1070's were available for 349.99 or lower at one point. Don't forget this.
No one knows if Vega will be good for mining so you can't say that. Vega is a totally different architecture.
Posted on Reply
#100
oxidized
"Prince Valiant said:
If the games AMD performs well on are AMD friendly that must mean that the games it performs bad on are Nvidia friendly.
No? Not necessarily, since often 1060 performed better than the 480 by 4/5%, so no, that doesn't show any friendly environment for nvidia.

"Vya Domus said:
Oh really , is that so ? Vega 64 has about 14 TFLOPS , right about where Titan Xp sits as well, die sizes are similar too. So yeah , they are on par in terms of power. Yet , if Vulkan favors "greatly" AMD how come it doesn't beat it ? You are on witch hunt mate , Doom/Vulkan is impartial to hardware.
Oh yeah? TFLOPS mean nothing, 1060 had less than 480, still, a faster card, in the beginning, and now even if the gap reduced a bit. TitanXP is a much faster card than any Vega, at least on games that is, rest i don't care, since we're talking about that. So again, Vega doesn't beat TitanXp under Vulkan because TitanXP is much faster that that on games. And NO, Doom isn't impartial at all, neither is Vulkan for the moment, in future we'll see.

"B-Real said:
Then I have no idea what you checked. First is Hardware Canuks as a written review, YT videos like HW Unboxed <div class="youtube-embed" data-id="CiYQqNiqQKU"><img src="https://i.ytimg.com/vi/CiYQqNiqQKU/hqdefault.jpg" /><div class="youtube-play"></div><a href="https://www.youtube.com/watch?v=CiYQqNiqQKU" target="_blank" class="youtube-title"></a></div>, getting 1% in favor of the 1060 and first it was a 12% difference. Yeah, it consumes 30-40W more at default (which is said by HW Unboxed that shouldn't be a deal breaker), but actually, JayzTwoCents' XFX 480 video shows that under stress test, that RX480 eats about 95 to 120W. And with Crimson, AB etc. you can control the more hungry RX480s like the Sapphire or MSI. Also, ones like MSI got later BIOS updates that got their power consumption lower.

So yeah, overall the RX480 (including the DX12 and Vulkan games) is a somewhat faster (or equal) card than the GTX 1060 since about 8 months now, and before that mining fever, it was about 30-40 bucks cheaper. At start, I would say they were egal (2016 summer), considering cost, performance, power consumption, but since decembers Crimson drivers, RX480 is simply the better choice.And for the RX580... yeah, its actually a faster card than the 1060. And it was sold for around 480 prices till the fever hit... Yeah, it consumed more than the RX480, it's for sure.
No sir, the 1060 was and is a faster even if sligthly card compared to 480, JayzTwoCent's 480 was a pretty lucky one since he could overclock it much while leaving untouched the voltage, i couldn't find anyone else that reached those frequencies, also there's nothing to control, Polaris is just designed with a higher TDP (232mm^2 vs 200mm^2, which is 16% bigger die) than Pascal, there's nothing to adjust. So again, overall the 480 isn't a faster card than 1060, it's the opposite, it's true the 480 increased performance overtime but couldn't still catch the 1060, put it as you want, and again what kept AMD floating was price/performance, nothing else really.
The 580 was maybe a little faster, costing more, and not at 480 price as you say, because i checked for that (i actually was planning to buy it) and it cost like a good AIB 1060, while having additional power draw, and it needed a bigger heatsink due the the higher TDP, which not everyone did, like sapphire on the pulse version.
Posted on Reply
Add your own comment