• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Polaris architecture – GCN 4.0

I'll be surprised if it isn't around 1400+MHz though they seem to be focusing on power efficiency - the process can do atleast 1400-1600 with bigger GPUs - and possibly not far off the 2GHz mark when going for clock speeds over balanced power/performance.

EDIT: Given they are on the smaller size core wise they might even manage better than that but it would be pure speculation.
 
Last edited:
Based on the node change and the Finfet tech (a lot less current leakage) it should be higher. Realistically we don't know how higher....but i hope for at least 1200-1300mhz.

I do hope for something clocking at least as well as the 79x0 did and the 78x0. Those could be pushed a fair deal and gave tangible results. Only better clocker ive had would be my old 780 lightning but that was insanely overbuild but man what a clocker that was even for a newbie like me.
 
I like to see how a reviewer going to handle this situation, they cant compare Polaris and Pascal as they adress different markets.

However anyone desire to upgrade that runs 1080p then Polaris is the new thing for them. It might be that anyone with a card up to 390/970 then Polaris is the given choice to upgrade to. Seems like a whole new tactical move from AMD to adress everything better from low end to mid range with price/performance. If thats the best move we will find out.

However it might be, buy Polaris if your a low/mid range gamer - 90% of users are just that.
 
I like to see how a reviewer going to handle this situation, they cant compare Polaris and Pascal as they adress different markets.

However anyone desire to upgrade that runs 1080p then Polaris is the new thing for them. It might be that anyone with a card up to 390/970 then Polaris is the given choice to upgrade to. Seems like a whole new tactical move from AMD to adress everything better from low end to mid range with price/performance. If thats the best move we will find out.

However it might be, buy Polaris if your a low/mid range gamer - 90% of users are just that.

They'd probably concentrate on "60fps gaming" - if 2 cards can hold 60fps with minimal or no drops then they can question spending more on the more expensive one heh.

I think they won't be far of 1070 performance mind - they seem to have put a fair bit of work into low level optimisations that will help to offset some of the die size difference and they may be able to make up a fair bit in clock speed which would be easier with the smaller die.
 
It seems there's going to be quite a gap between P10 and Fury, with nothing in the 490 slot until Vega - which is exactly where the new Nvidia cards will fit I think. It seems weird that AMD are not launching a full top to bottom product stack - it's been a long time since they did that. Maybe it's a rush to get 14nm product out, and the mid/low end is just easier and cheaper.
 
It seems there's going to be quite a gap between P10 and Fury, with nothing in the 490 slot until Vega - which is exactly where the new Nvidia cards will fit I think. It seems weird that AMD are not launching a full top to bottom product stack - it's been a long time since they did that. Maybe it's a rush to get 14nm product out, and the mid/low end is just easier and cheaper.

Another alternative is that P10 will be close enough to Fury/X speeds that AMD can discontinue a large, expensive and frankly poor selling GPU. With Nvidia having such a great success with 980Ti they could not afford to discontinue a £500 GPU that was selling well unless they had something to slot right in at that price. Hence 1080 will continue a very successful price slot for Nvidia.

Fiji is a similar size to Maxwell 980Ti with more expensive (HBM) and is not selling as well, so AMD would be keen to drop it entirely if possible IMHO.

Remember P10 allegedly has the same number of shaders as R9 390 and with clock increases and architectural improvements ~30% increased performance is not a fantasy. That would put P10 at roughly 980Ti/Fury X speeds IMHO but with a much smaller and cheaper chip/VRAM package.

We will know soon enough.
 
Last edited:
Another alternative is that P10 will be close enough to Fury/X speeds that AMD can discontinue a large, expensive and frankly poor selling GPU. With Nvidia having such a great success with 980Ti they could not afford to discontinue a £500 GPU that was selling well unless they had something to slot right in at that price. Hence 1080 will continue a very successful price slot for Nvidia.

Fiji is a similar size to Maxwell 980Ti with more expensive (HBM) and is not selling as well, so AMD would be keen to drop it entirely if possible.

Remember P10 allegedly has the same number of shaders as R9 390 and with clock increases and architectural improvements ~30% increased performance is not a fantasy. That would put P10 at roughly 980Ti/Fury X speeds IMHO.

We will know soon enough.

Fudge the numbers a bit and you get pretty close to FX performance without taking into account any architectural improvements - I reckon it isn't infeasible they can give the 1070 a hard run for its money while being more cost effective.
 
AMD will hold an event at end of may before computex is no news.
Polaris will be mainstream, cheap, and 2/3 size of GP104 so they will be in different tiers and won't really compete are no news either.

I guarantee you the full fat P10 will still beat GP104 in some games / benchmarks, despite being significantly smaller and cheaper.

They won't be the same pricing tier, and will be much less costly to produce, but they are likely to have broadly similar performance.
 
Another alternative is that P10 will be close enough to Fury/X speeds that AMD can discontinue a large, expensive and frankly poor selling GPU. With Nvidia having such a great success with 980Ti they could not afford to discontinue a £500 GPU that was selling well unless they had something to slot right in at that price. Hence 1080 will continue a very successful price slot for Nvidia.

Fiji is a similar size to Maxwell 980Ti with more expensive (HBM) and is not selling as well, so AMD would be keen to drop it entirely if possible.

Remember P10 allegedly has the same number of shaders as R9 390 and with clock increases and architectural improvements ~30% increased performance is not a fantasy. That would put P10 at roughly 980Ti/Fury X speeds IMHO.

We will know soon enough.

Yeah, that makes a lot of sense to me.

My prediction is the best Polaris card will keep up with the 1070 and be priced at around £300 or less, offering the same performance but for ten, maybe twenty quid cheaper than the 1070.

If the above is true, I will likely get the Polaris over a 1070. End of May will be an interesting time :D
 
I'm expecting Polaris to be around £250-300 for 980ti level
I probably wouldn't buy at that though. I'd buy at £200. Will be interesting to see how AMD play the field.
 
Fudge the numbers a bit and you get pretty close to FX performance without taking into account any architectural improvements - I reckon it isn't infeasible they can give the 1070 a hard run for its money while being more cost effective.

From the little information we have seen architecturally AMD haven’t really changed much on Polaris, were getting more of the same from what I can tell. That’s great if a game uses async compute as we know what GNC is capable of in DX12 but it means DX11 games performance will be reliant on programmers and whatever AMD’s driver team can crank out (which can be hit and miss) and for the next year at least DX11 will be the majority API for new releases IMO.

There’s certainly enough anecdotal evidence to suggest AMD’s software team have finally manged to learn how to get the best out of GNC if you look at the performance in some new release games which might explain why they didn’t want to overhaul design. If this works then that’s pretty smart as it saves money but Pascal looks like a beast on paper so it makes you wonder if AMD have been ambitions enough with the new process. Time will tell.
 
IDCP said:
Another alternative is that P10 will be close enough to Fury/X speeds that AMD can discontinue a large, expensive and frankly poor selling GPU. With Nvidia having such a great success with 980Ti they could not afford to discontinue a £500 GPU that was selling well unless they had something to slot right in at that price. Hence 1080 will continue a very successful price slot for Nvidia.

Fiji is a similar size to Maxwell 980Ti with more expensive (HBM) and is not selling as well, so AMD would be keen to drop it entirely if possible IMHO.

Remember P10 allegedly has the same number of shaders as R9 390 and with clock increases and architectural improvements ~30% increased performance is not a fantasy. That would put P10 at roughly 980Ti/Fury X speeds IMHO but with a much smaller and cheaper chip/VRAM package.

We will know soon enough.

I always thought the Fury would have a short life, but I don't want to see Vega go in at that kind of price point. It would still leave a big gap between the 480/490 price wise. Ideally, I'd like to see Vega take a step down from where Fury is now, leaving Nvidia products looking very overpriced and Vega offering powerful performance at the £300-350 mark.
 
From the little information we have seen architecturally AMD haven’t really changed much on Polaris, were getting more of the same from what I can tell. That’s great if a game uses async compute as we know what GNC is capable of in DX12 but it means DX11 games performance will be reliant on programmers and whatever AMD’s driver team can crank out (which can be hit and miss) and for the next year at least DX11 will be the majority API for new releases IMO.

There’s certainly enough anecdotal evidence to suggest AMD’s software team have finally manged to learn how to get the best out of GNC if you look at the performance in some new release games which might explain why they didn’t want to overhaul design. If this works then that’s pretty smart as it saves money but Pascal looks like a beast on paper so it makes you wonder if AMD have been ambitions enough with the new process. Time will tell.

Raja himself has confirmed that there is significant changes to a few areas including how geometry is handled (both frontend and backend) - I doubt the wild claims some are coming out with - partly because a say 80% improvement in that part of the system doesn't equate to an 80% improvement in overall game performance (probably more like 15%) - but there should be a few percent increases from that over like for like with the previous generation.
 
I hope so - then it suddenly means that the Nvidia launch was actually as spoiler against incoming AMD products.

That much was obvious from the beginning. They did a paper launch, and availability will be initially on a very limited basis. Also, as some predicted, it's FINFET Maxwell, and at least to me looks extremely underwhelming.
 
I have a feeling that AMD are going to continue with the Fury line, albeit possibly not in its current form? HBM2 update for it to sit at the top of the AMD tree until Vega hits
 
That much was obvious from the beginning. They did a paper launch, and availability will be initially on a very limited basis. Also, as some predicted, it's FINFET Maxwell, and at least to me looks extremely underwhelming.

:rolleyes:

Must have been a nightmare weekend for you having your world turned inside down!
 
Back
Top Bottom