AMD Polaris architecture – GCN 4.0

[drunkenmaster]

The only way that a chip that small could be faster then the FuryX is if, as Some of us said many, many months ago the performance from this new 14/16nm node is two maybe two and a half times the currant performance. When we said this we were shouted down saying how stupid and ridiculous that idea was.

I really do pray to god that this isn't the bigger of the two Polaris's that are coming sometime before September.

So with all that, a full size 450mm^ part would be rocking at the very least 2* the performance of the 290x that's better than 295x2 performance and miles ahead of a FuryX, which, when it was talked about before, people were told they were crazy for expecting such performance. so if this is right now, then those people were correct before.

I wasn't suggesting that the large cores would be arriving anytime soon, as has been said probably not even this year, as much as we would like it to be different. I'm just trying to say that if as has been suggested that this small chip is going to be up there with the FuryX, then the big chip to follow will be as quick as was suggested previously.

Really, the first thing you claim is that people called you(and un-named others) crazy for believing a new node allows roughly double the performance. Can you point out a post, or someone's name calling you or anyone else crazy?

Or can you only find people telling you that expecting way higher than FuryX performance THIS YEAR wasn't going to happen?

For give or take 20 years a new graphics node enables 80-100% more performance depending on the outgoing and incoming architecture (great to meh reduces performance gain, meh to great increases it). You're posting setting yourself up as the sole voice of reason and that people called you crazy for believing a chip much faster than Fury X was possible at all.

You also combine this with talk of being disappointed if the bigger chip this year isn't that big and ignoring the context of people saying double Fury X performance is likely, just not this year.

I've not seen anyone in any thread on 14/16nm gpus believing that double the performance for similar sized chips isn't possible.

 

[Mauller]

Plus,I think people were being unreasonable in thinking that the GTX980TI and Fury X which launched only in the summer would be replaced within a year?? That's a pretty short lifespan.

I expect we will see the high end GPUs in consumer cards from both towards the end of the year,or early 2017 which would give them around 18 months of lifespan.

The Fury X is an expensive card to manufacture. So it could easily be replaced sooner than later. And AMD don't care when NVIDIA parts were released.

Amd's Polaris range could be small but very performant and priced well. They will more than likely stop manufacturing the majority of their 28nm parts if they are targeting 290x -980ti performance at under £300 etc.

 

[flopper]

There is another way to look at its, 14nm is 2x the transistor density.

The Fury-X is 550mm and has 8.9bn Transistors.

2x 232mm = 464mm = 82% of Fury-X / 7.5bn Transistors.

The 290X is 432mm with 6.2bn Transistors.

So that speculation would have it between a 290X and a Fury-X, or 980 and 980TI. (this without any architectural performance improvement)

Furyx, 598mm.
a 232mm x 2.5 improvement would put in the ballpark of furyx performance.
so a 350us card will be the replacement for todays furyx/980ti.
(if they price it that way that is)

The entusiast replacement for furyX will be a 350mm 14nm chip or so looks to be by far awesome bright Polaris

 

[Overlag]

I doubt going from 28nm to 14nm is going to cause a 2.5x improvement. At most its 2x?

 

[CAT-THE-FIFTH]

I doubt going from 28nm to 14nm is going to cause a 2.5x improvement. At most its 2x?

Maybe its more a 2.5X improvement in power consumption?? Remember AMD is comparing this probably to their cards.

IIRC,it was like a 60W difference between a GTX950 and small Polaris running the same game and the same settings.

If you were to compare that to the R9 270X,which is around the same performance as a GTX950,a 2.5X improvement,would mean an equivalent small Polaris GPU would consume around 50W:

https://tpucdn.com/reviews/EVGA/GTX_950_SSC/images/power_average.gif

Give or take a few watts here and there,that could be the 2.5X improvement.

That would mean a GTX960 4GB power consumption class GPU having R9 290/390 series level performance,ie,having a single six pin PCI-E power connector.

That was also make it between 30% to 50% faster than the R9 380X.

If that is launched around the £150 to £200 mark,it would be a bigger jump than we had during the GTX660 and HD7870 days as these were competing in the £150 to £200 area and the HD7870 originally was around the £200 to £250 mark when it launched.

 

[pmc25]

Well that'd be 4mm2 bigger than a GTX 960, and would put it at approximately 140W TDP class (give or take depending on clockspeed).

With a 2x performance/watt increase, this could be the 'larger' chip and should be easily faster than a GTX 980


Mildly disappointing if this turns out to be the biggest chip we get for the first round. Was hoping to see something in the 290-300mm2 class, since that would reasonably guarantee being 980 TI+ performance.

That's Polaris 11, the second gddr5 chip, which was shown in private at ces and at a private event in December. Doubt it's anywhere near 140W unless clocks are massive.

The HBM2 'big'(ish) chip will be a little later.

 

[CAT-THE-FIFTH]

Looking at the latest TPU review,around R9 290 to R9 390X level performance would make Polaris 11 65% to 85% faster than a GTX960:

https://tpucdn.com/reviews/Sapphire/R9_390_Nitro/images/perfrel_1920_1080.png

If AMD is agressive on pricing and gets this out at between £150 to £200 and a few months before Nvidia get's their 16NM midrange chips out,AMD could get some decent sales out of ity IMHO.

 

[Toss3]

Looking at the latest TPU review,around R9 290 to R9 390X level performance would make Polaris 11 65% to 85% faster than a GTX960:

https://tpucdn.com/reviews/Sapphire/R9_390_Nitro/images/perfrel_1920_1080.png

If AMD is agressive on pricing and gets this out at between £150 to £200 and a few months before Nvidia get's their 16NM midrange chips out,AMD could get some decent sales out of ity IMHO.

Where did you get that the Polaris 11 would perform on the same level as the 290? Remember that is has got 2.5 times the performance per watt, so it's probably going to be closer to the GTX 980 ti.

 

[AlamoX]

AOC's U2879VF 28" 4K 1ms FreeSync monitor for £279 MSRP... Source

 

[pmc25]

AOC's U2879VF 28" 4K 1ms FreeSync monitor for £279 MSRP... Source

Good price. Wake me up when it's 144hz though.

 

[Overlag]

AOC's U2879VF 28" 4K 1ms FreeSync monitor for £279 MSRP... Source

60hz urrg.

 

[AlamoX]

Good price. Wake me up when it's 144hz though.

so you are going to sleep for about a year...we call that a coma

 

[Ayahuasca]

Nah he'll be going to sleep forever as that monitor is never going to be 144hz

 

[pmc25]

There'll be stuff available as soon as Polaris is.

 

[Raikiri]

Good luck running anything demanding at 4k 144FPS.

 

[pmc25]

Good luck running anything demanding at 4k 144FPS.

Not interested in lower refresh rates, so not bothered.

 

[Kaapstad]

There'll be stuff available as soon as Polaris is.

Maybe not.

Small Pascal and Polaris won't be much faster than big Maxwell.

12gb TitanXs will still be more capable cards than Small Polaris and Pascal 8gb cards @2160p.

Things will only start moving when big Pascal and Polaris have been on the shelves a while.

 

[pmc25]

Maybe not.

Small Pascal and Polaris won't be much faster than big Maxwell.

12gb TitanXs will still be more capable cards than Small Polaris and Pascal 8gb cards @2160p.

Things will only start moving when big Pascal and Polaris have been on the shelves a while.

I assume all desktop Polaris SKUs will feature DP1.3. Performance is moot.

 

[Raikiri]

Not interested in lower refresh rates, so not bothered.

So you'll run games at low settings just to get 144FPS?

 

[AllBodies]

Maybe its more a 2.5X improvement in power consumption?? Remember AMD is comparing this probably to their cards.

IIRC,it was like a 60W difference between a GTX950 and small Polaris running the same game and the same settings.

If you were to compare that to the R9 270X,which is around the same performance as a GTX950,a 2.5X improvement,would mean an equivalent small Polaris GPU would consume around 50W:

https://tpucdn.com/reviews/EVGA/GTX_950_SSC/images/power_average.gif

Give or take a few watts here and there,that could be the 2.5X improvement.

That would mean a GTX960 4GB power consumption class GPU having R9 290/390 series level performance,ie,having a single six pin PCI-E power connector.

That was also make it between 30% to 50% faster than the R9 380X.

If that is launched around the £150 to £200 mark,it would be a bigger jump than we had during the GTX660 and HD7870 days as these were competing in the £150 to £200 area and the HD7870 originally was around the £200 to £250 mark when it launched.

It may well be over the whole 14/16 generation, but maybe not in the first batch of cards.

TSMC claim 16FF+ consumes 70% less power than 28nm. So that's a 3.33x performance per Watt increase. Clearly that must be in a best-case scenario, but then 2.5x doesn't seem completely out of the question.

That's Polaris 11, the second gddr5 chip, which was shown in private at ces and at a private event in December. Doubt it's anywhere near 140W unless clocks are massive.

The HBM2 'big'(ish) chip will be a little later.

I mean't 'larger' as in the larger of the two we get in the first batch of cards. Not the large chip of the whole generation.

Also the GTX 960 is 228mm2 (so smaller) and is 120W.

So 140W on 232mm2 seems totally plausible.