Polaris 10 vs R9 390/390X @ GFXBench

[USER8000]

AMD/NVIDIA performance is not directly comparable here.



No it isn't, not the one in your link neither average R9 380 scores.



Indeed.

Nope.

So,if my link is wrong,so why is yours gospel truth?

This is the whole point of trying to make a thread of something which can be put into one of the other threads. We have one Pascal thread and 50 billion Polaris ones.

You have no clue about:
1.)What clockspeed the card has
2.)How many shaders are active
3.)What even the driver revision is

Those scores are 2/3 of an R9 390X,you listed.

Let's look at the latest TechPowerUp review:

https://tpucdn.com/reviews/ASUS/GTX_950/images/perfrel_1920_1080.png

That would place the Polaris 10 chip at around GTX770,or 10% faster than a GTX960,or basically an R9 380.

Looks,like I need to OC my GTX960 a bit.

:sarcasm:

Edit to post!!

Come to think of it how do we even known these are not R9 380 scores in the first place?

 

[R0H1T]

If (full?)Polaris 10 is slower than the 390, I'll sell my Fury for $100 then eat my own socks.

Even for the regular GDDR5 version, if so when can I buy your Fury :sneaky:

Just for the record I don't expect full Polaris 10, minus GDDR5x, to be slower than the 390

 

[Sweepr]

Glad I haven't sold my 390x yet. I might not want to if your prediction is correct.

Many unknown variables, but I have little doubt the results are legit. Could be the rumoured 2304 SPs @ 800 MHz part leaked a month ago @ SiSoftware (same ID - 67DF:C4).

 

[Gundark]

If Polaris 10 is slower than the 390, I'll sell my Fury for $100 then eat my own socks.

May I please be the first in line to buy your Fury? Man, I now wish that it's really slower, you turned me around.
Edit: R0H1T beat me.

 

[Olecki]

Because AMD keeps talking about increasing the number of VR capable systems. That means more computers with a 390 or better performance.

I will quote :
"If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it's around 7.5 million units,"

So Polaris 10 need to have similar or slightly better performance and much smaller price to increase numer of VR capable computers. It doesn't mean at all that Polaris will be more powerfull than Fury X - this will be Vega territory imho.

 

[ShintaiDK]

Because AMD keeps talking about increasing the number of VR capable systems. That means more computers with a 390 or better performance.

I will quote :
"If you look at the total install base of a Radeon 290, or a GTX 970 or above [the minimum specs required for VR], it's around 7.5 million units,"

So Polaris 10 need to have similar or slightly better performance and much smaller price to increase numer of VR capable computers. It doesn't mean at all that Polaris will be more powerfull than Fury X - this will be Vega territory imho.

He didn't even get the install base right. Plus PR is PR. And there is a long way from Tonga to Hawaii performance wise.

People expecting Fury X performance are dreamers.

 

[Bryf50]

I like how you can automatically tell if a leak is positive or negative based on who posted it. It's convenient.

 

[Kenmitch]

What's this card? 1200p

https://gfxbench.com/device.jsp?benchmark=gfx40&os=Linux&api=gl&cpu-arch=x86&hwtype=dGPU&hwname=AMD%20Device&did=33006520&D=AMD%20Device

What about this one? 1080p

https://gfxbench.com/device.jsp?ben...me=NVIDIA Device&did=32655424&D=NVIDIA Device

 

[linkgoron]

I like how you can automatically tell if a leak is positive or negative based on who posted it. It's convenient.

Yes, I cannot agree more.

EDIT:
Love the $299 tag.

 

[Olecki]

He didn't even get the install base right. Plus PR is PR. And there is a long way from Tonga to Hawaii performance wise.

People expecting Fury X performance are dreamers.

Problem is that giving that kind of performance doesn't increase much numer of VR capable computers. If this will be 400$ card with high power who will buy it? Probably people, which have slower card, but still 'strong enough for VR. To spread VR AMD need to deliver product that will be selling for 200-250$ outside of USA. Please remember that because of taxes rest of the world usually pays 20-30% more.

 

[ShintaiDK]

Problem is that giving that kind of performance doesn't increase much numer of VR capable computers. If this will be 400$ card with high power who will buy it? Probably people, which have slower card, but still 'strong enough for VR. To spread VR AMD need to deliver product that will be selling for 200-250$ outside of USA. Please remember that because of taxes rest of the world usually pays 20-30% more.

GPU cost is pretty much irrelevant with the current VR headset prices.

 

[Olecki]

GPU cost is pretty much irrelevant with the current VR headset prices.

With that hype for VR prices go down pretty fast, especially that more capable PC's means even more VR headsets manufacturers.

 

[DarkKnightDude]

Don't believe anything till the reviews honestly.

 

[tential]

I like how you can automatically tell if a leak is positive or negative based on who posted it. It's convenient.

This is so true it hurts. Honestly some of these threads/posts should be in the amd/Nvidia subsections (the cheerleaders sections?).

 

[Jaydip]

Man, the AMD faithful sure hate leaks that don't favor them, but love those that do.

Rinse and repeat for the NV faithful.

QFT, they were claiming it would trump a 980 TI for $299, now they are in denial mode. OTOH I don't care about these leaks, it will only matter when they are in the hands of professional reviewers.

 

[airfathaaaaa]

QFT, they were claiming it would trump a 980 TI for $299, now they are in denial mode. OTOH I don't care about these leaks, it will only matter when they are in the hands of professional reviewers.

so you are bashing the people that spreaded a rumor by practicly confirming those numbers as true without knowning anything prior to it
while later you jump and say "i wait for real benches"
kinda imploding on your own statement here D:

 

[Flapdrol1337]

How come the 390x is 30% faster than the 390 in this test?

 

[ultron]

These results are unimportant even if they are true. First of all, Polaris cards are not released yet and these benchmarks shows pre-production samples's performance. Their drivers are very immature now, it matters because GCN 4.0 and gcn 3.0 difference far more bigger than GCN 3.0-1.0. They changed a lot of thinks and driver team doesn't have any experience on them. The second thing is these are synthetic tests. Who matters them? The third and most important one is people have wrong expectations from them as far as i see. Die size is 2X smaller in comparison to Hawaii, power consumption decrease is significant. They'll come with new features. The real concern should the value, not the performance. And the 16 nm is very immature and expensive process. Besides the company has R&D expenses that should pay for itself. GPU cost never be cheap if it's about 230 mm^2. Cooling solutions and power units will cost relatively low thanks to low power consumption but it's very hard to be has more value than 390, at least in this year.

Do we have any info about wafer prices and yield for 16 nm? Without knowing this we can't talk about this products before their launch. Can i trust this source:

 

[rainy]

Die size is 2X smaller in comparison to Hawaii, power consumption decrease is significant.

You're comparing two different processes - those 232 mm2 translate to a bit bigger chip than Grenada/Hawaii in 28nm.

And the 16 nm is very immature and expensive process.

Polaris would be made in 14nm (Samsung/GF), Pascal in 16nm of TSMC.

 

[JDG1980]

Either these results are from Polaris 11 and not Polaris 10, or (more likely) the current testing drivers just haven't been optimized in any way for legacy OpenGL on the new architecture. The numbers shown here are actually worse than Tonga in some benchmarks, which just doesn't make any sense; that would imply not only no gains, but a regression in shader IPC without any corresponding improvements in clock speeds. That would be a Bulldozer-level fail, which just doesn't jive with the public demos we've seen so far. If a 2304-shader Polaris part with a 256-bit bus is worse than Tonga, then how can a much smaller corresponding part (estimated 1024-1280 shaders) with a 128-bit bus run Star Wars Battlefront at constant 60 fps? GTX 950 can just barely do it. And Polaris 10 was shown running a scene in Hitman without frame drops that even Fury X has trouble with. This doesn't mean it will be as fast as Fury X in all situations, of course, but it means we should be very skeptical of a result that claims it will be worse than R9 390. Furthermore, AMD is specifically positioning Polaris 10 as a cheap VR card, which means it must be at least as good as 390/970 performance at a minimum.

 

[Jaydip]

so you are bashing the people that spreaded a rumor by practicly confirming those numbers as true without knowning anything prior to it
while later you jump and say "i wait for real benches"
kinda imploding on your own statement here D:

Not really read it again, I love when the AMD brigade always hype the launches like its gonna end NV's dominance but gets disappointed each and every time.Same thing is happening this time too.Well if all the negative rumors should be true for NV what is the problem with this one for AMD?

 

[Sweepr]

BTW the same guy that said nearly 4.000 points @ Firestrike Ultra - the data which copycat websites used to base the 'Geforce GTX 980 Ti' at $299 rumours - also said this:

- Ellesmere Pro: 86% R9 390
- Ellesmere XT: 97% R9 390X

We already know 67DF:C4 is a 2304 SP part according to SiSoftware, and the GFXBench results match almost exactly his expectations for Ellesmere Pro.

 

[maddie]

BTW the same guy that said nearly 4.000 points @ Firestrike Ultra - the data which copycat websites used to base the 'Geforce GTX 980 Ti' at $299 rumours - also said this:

- Ellesmere Pro: 86% R9 390
- Ellesmere XT: 97% R9 390X

We already know 67DF:C4 is a 2304 SP part according to SiSoftware, and the GFXBench results match almost exactly his expectations for Ellesmere Pro.

WOW

Am I seeing a model with HBM. Ellesm. DS = 110% 390X

We also have shader performance regression. The GPU side has learnt a thing from the CPU side. It's about time the divisions shared their expertise. Bulldozer adapted for GPUs.

Ellesm. XT = 97% 390X (2816,176,64,384)

And this is after a node shrink.


YAY. Nvidia forever. They're never gonna catch us now.

 

[airfathaaaaa]

Not really read it again, I love when the AMD brigade always hype the launches like its gonna end NV's dominance but gets disappointed each and every time.Same thing is happening this time too.Well if all the negative rumors should be true for NV what is the problem with this one for AMD?

i would ask for real facts on that post but i know you wont find any simply because you are getting a rumor and spreading it like a virus to suit your views kinda like the ones that kept spamming "pascal on april" then "pascal on may"

 

[Det0x]

"Edit: double post."

Sorry, just wanted more people to see it. Hate it when it gets stuck in the last posts of a page.

Trademarked by S®