AMD Polaris architecture – GCN 4.0

drunkenmaster · 14 May 2016 at 01:19

AllBodies said:
Yeah true, especially with the rumours making the rounds that Vega will come sooner than thought. And then on top of that, if that turns out to be true Nvidia will be expected to follow with the 1080 Ti and/or Titan ASAP too.

So in a few weeks it'll be something like "Oh Polaris and middle-Pascal are cool, but HEY EVERYONE what about Vega vs big-Pascal eh? That'll be totally different and super cool right?"

It never ends

WHile not discounting the possibility of a different consumer to enterprise card for high end gaming from Nvidia, there isn't an awful lot so far to suggest that GP100 won't be used for gaming. If a Titan like card launches with GP100 or the much fabled GP102, it will still be a larger core on interposer with HBM2 and GP100 is going to be WAY more profitable for Nvidia. So ultimately GP100 will be Nvidia's priority and Nvidia have already made it pretty clear that won't be capable of launching with any real volume before Q1/17. That is likely the point they see yields being viable for good profit of cards selling at $3k and above. Titan came months after release for the professional segment and 980ti months after that.

Basically I'm not at all expecting anything bigger than the 1080 from Nvidia before at least late Q1/17, if not well into Q2.

layte · 14 May 2016 at 01:21

Mtom said:
Titan X supposed to have been a halo cards, it shouldn't have been trading blows with 390x, it should have beaten it across the board

Hmm this is easier then i thought.

It's interesting what happens when the next chapter is tested and they dont use the canned benchmark.

humbug · 14 May 2016 at 01:25

layte said:
It's interesting what happens when the next chapter is tested and they dont use the canned benchmark.

That ones also bottlenecked...

Seanspeed · 14 May 2016 at 01:41

humbug said:
That ones also bottlenecked...

It is, but there's a bit of difference, isn't there?

humbug · 14 May 2016 at 01:45

Seanspeed said:
It is, but there's a bit of difference, isn't there?

There is a bit yes. Certainly more than the Guru3D one.

Seanspeed · 14 May 2016 at 01:52

humbug said:
There is a bit yes. Certainly more than the Guru3D one.

My point was that the Nvidia vs AMD factor changed considerably just by testing a different chapter.

humbug · 14 May 2016 at 01:57

Seanspeed said:
My point was that the Nvidia vs AMD factor changed considerably just by testing a different chapter.

yeah for sure.

but what worries me is that 980 in the position its in, 8% faster than a Fury-X @ 1080P and yet the 390 Non X is 10% faster @ 4K than that 980 while said Fury-X is only 10% faster than the 390 at 4K...
As for the Nano, what is going on with that?

Its a bit mixed up margins of error all over the place.

drunkenmaster · 14 May 2016 at 01:59

Seanspeed said:
My point was that the Nvidia vs AMD factor changed considerably just by testing a different chapter.

No, the situation changed considerably by.... testing stock AMD cards vs overclocked Nvidia ones. Nvidia likes to pay review sites to test stock cards vs overclocked edition cards to cause reviews to look like that.

Somehow Fury X is still nearly 30% ahead of the 980 at 4k and the has a 26fps minimum vs 21fps for the 980.

CAT-THE-FIFTH · 14 May 2016 at 01:59

humbug said:
That ones also bottlenecked...

And also using factory over clocked cards.

But I can see the thread thrashing happening here.

This is the Polaris thread and some are desperate to thrash this one yet don't seem to want to thrash the Pascal one.

Instead they are desperate to massage their E-PEEN,since their Fury and GTX980TI cards are going to have a death by a 1000 cuts and are going to plummet in value. Comes across as massive purchase justification.

Nobody cares about Fury X and GTX980TI benchmarks in this thread.

People should go and bicker about one year old cards in the old threads.

layte · 14 May 2016 at 02:00

drunkenmaster said:
No, the situation changed considerably by.... testing stock AMD cards vs overclocked Nvidia ones. Nvidia likes to pay review sites to test stock cards vs overclocked edition cards to cause reviews to look like that.

Somehow Fury X is still nearly 30% ahead of the 980 at 4k and the has a 26fps minimum vs 21fps for the 980.

lol

humbug · 14 May 2016 at 02:01

drunkenmaster said:
No, the situation changed considerably by.... testing stock AMD cards vs overclocked Nvidia ones. Nvidia likes to pay review sites to test stock cards vs overclocked edition cards to cause reviews to look like that.

Somehow Fury X is still nearly 30% ahead of the 980 at 4k and the has a 26fps minimum vs 21fps for the 980.

Its a strange mixed bag across the res for all cards, none of these cards are stretching their legs. not beyond the 390 anyway.

edit + yeah overclocked, its not clear...

CAT-THE-FIFTH · 14 May 2016 at 02:06

Who gives a damn?

What had a blasted GTX980TI vs Fury X benchmark gitcto do with Polaris or Vega.

Are some of you just bored??

Or are you just trying to bury an information posted in this thread.

Why don't you polute the Pascal thread with it then?

Don't you see how gullible some if you are - certain people want to get this derailed so any information gets buried while they leave the Pascal threads clear.

Seanspeed · 14 May 2016 at 02:07

humbug said:
Its a bit mixed up margins of error all over the place.

Yet still used as a representative benchmark when convenient.

humbug · 14 May 2016 at 02:07

CAT-THE-FIFTH said:
Who gives a damn?

What had a blasted GTX980TI vs Fury X benchmark gitcto do with Polaris or Vega.

Are some of you just bored??

Or are you just trying to bury an information posted in this thread.

Why don't you polute the Pascal thread with it then?

Ok back on topic then....

D.P. · 14 May 2016 at 02:08

CAT-THE-FIFTH said:
I thought it would be a 75w TDP card. But 60w TDP? That means it's 40% to 70% faster than the bus powered GTX950 while having a 20% lower TDP.

That makes Polaris 11 more a GTX750TI level card in terms of TDP and the R7 360E and GTX750 had similar TDPs and consumed roughly the same amount of power too.

Polaris 10 might be quicker than a R9 390X if this is the case.

And that is all expected given the new process and.prior tumours. In fact I heard 50w before.

AMD will definitely be ahead of nvidia in this segment though

layte · 14 May 2016 at 02:09

If you read their testing methodology (which is not simple as the site is German), they say they use cards that give a representative result for that model of card. I think it's fair to say that the boost clocks shown are well within reach of most if not all versions of said card.

Anyway, apologies for this off topic tangent, I just thought it was an interesting point to make that one sample set is not always representative of the whole.

Kaapstad · 14 May 2016 at 02:13

drunkenmaster said:
No, the situation changed considerably by.... testing stock AMD cards vs overclocked Nvidia ones. Nvidia likes to pay review sites to test stock cards vs overclocked edition cards to cause reviews to look like that.

Somehow Fury X is still nearly 30% ahead of the 980 at 4k and the has a 26fps minimum vs 21fps for the 980.

Prove it

Gerard · 14 May 2016 at 02:16

Do you 2 have to have drama across numerous threads? **** me, when did these stop being cards and start being a religion (posv reference).

D.P. · 14 May 2016 at 02:19

layte said:
If you read their testing methodology (which is not simple as the site is German), they say they use cards that give a representative result for that model of card. I think it's fair to say that the boost clocks shown are well within reach of most if not all versions of said card.

Anyway, apologies for this off topic tangent, I just thought it was an interesting point to make that one sample set is not always representative of the whole.

This is something I said in the pascal thread. Nvidia Lau NH conservative reference cloxks, partners sell factory OC cards for.same price, in fact hard to fund stock clocked 980tis. But dome people wa NH t yo compare stock 980ti vs stock furyX when alms every o89ti has a higher clock.

This will be an even bigger fa got with pascal. Stock clocks of 1.6GHz but factory OC to 2.4GHz. , the reviews will be all.over the place and amd affiliated crew will jump on the clock speed.

drunkenmaster · 14 May 2016 at 02:36

D.P. said:
This is something I said in the pascal thread. Nvidia Lau NH conservative reference cloxks, partners sell factory OC cards for.same price, in fact hard to fund stock clocked 980tis. But dome people wa NH t yo compare stock 980ti vs stock furyX when alms every o89ti has a higher clock.

This will be an even bigger fa got with pascal. Stock clocks of 1.6GHz but factory OC to 2.4GHz. , the reviews will be all.over the place and amd affiliated crew will jump on the clock speed.

One, those clock speeds aren't the standard clock speeds on the non reference card, the card is sold with an overclock and those clocks are beyond that overclock.

Compare overclock vs overclock or stock vs stock, stock vs overclock while also using things like power advantage as a pro point is ridiculous, also for the most part overclocked models cost more than reference cards.