• 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 don't understand why people expect so little of Polaris 10. I mean even a straight die shrink of a 390X would be ~220mm2 and would obviously be the speed of a 390X.

P10 is a little bigger than that and it will no doubt have efficiency improvements which I would hope at least minorly improve this, then with a pretty good clock bump on top of that.

If it came out any slower than a 390X minimum I would be very disappointed.

Yeah, pointed this out several times myself.

Also if they did just do the straight-shrink it would clock higher too due to the 14LPP process.

So you'd expect a ~220mm2 card which was 15-20% faster than the 390X.

P10 is larger and an upgraded architecture, so people should be expecting it to be around FuryX/980Ti level as the minimum.
 
Its from leaks of a 2048 shader Laptop part which benched at around a 390/X.

The fact that its a Laptop GPU somehow got discarded and the whole thing spiralled into hyperbole.
 
Yeah, pointed this out several times myself.

Also if they did just do the straight-shrink it would clock higher too due to the 14LPP process.

So you'd expect a ~220mm2 card which was 15-20% faster than the 390X.

P10 is larger and an upgraded architecture, so people should be expecting it to be around FuryX/980Ti level as the minimum.

I completely agree. I can't see it being less than around there which would put it close to the 1070 probably at a lower price point.

Bring on the competition, it's good for all of us.
 
Yeah, pointed this out several times myself.

Also if they did just do the straight-shrink it would clock higher too due to the 14LPP process.

So you'd expect a ~220mm2 card which was 15-20% faster than the 390X.

P10 is larger and an upgraded architecture, so people should be expecting it to be around FuryX/980Ti level as the minimum.

So have I, I cannot see the point if it is worst than a shrink of a 390x as that will be a big fail.
humbug said, Its from leaks of a 2048 shader Laptop part which benched at around a 390/X.

The fact that its a Laptop GPU somehow got discarded and the whole thing spiralled into hyperbole.

I also think this as well it is just that they improved so much with next GCN generation that it should show and people need to remember that AMD do have a better showing in dx12 even though it is still early and with 2 consoles with GCN it could stay that way for a while.
 
One things you guys have to consider is the density they are actually going to use in this card is unknown.

Also if rumours are to be believed, then p10 will have 2560 stream processors, which is less than full fat hawaii 2816. So we can't just neceserilly claim that a die shrunk 390x is the minimum level (before efficiency and clockspeed).

I'm definitely hoping it's a good card though!
 
One things you guys have to consider is the density they are actually going to use in this card is unknown.

Also if rumours are to be believed, then p10 will have 2560 stream processors, which is less than full fat hawaii 2816. So we can't just neceserilly claim that a die shrunk 390x is the minimum level (before efficiency and clockspeed).

I'm definitely hoping it's a good card though!

But why can't we? I mean if they did all of this for Polaris, then it turned out to be worse than just a direct die shrink why wouldn't we be disappointed by it?
 
But why can't we? I mean if they did all of this for Polaris, then it turned out to be worse than just a direct die shrink why wouldn't we be disappointed by it?


I you could definitely be disappointed by it.
I was more referring to how since it was more than half the size of a 390x, then the slowest it could possibly be is above a 390x then adding the clockspeed and efficiency improvements to get the final performance.

It's not impossible that it could not be bigger than a 390x in raw physical game rendering terms (ie slower if using the same GCN and clock) but with technological advancements bringing the extra performances.

Kind of how like the 1080 isn't at a baseline a shrunken titan X in terms of core count even if the card is estimated to be slightly more than half the size.

Hopefully IPC improvements and clockspeed make for a great card, but there's no guarantee it'll be 980ti level at a minimum.
 
Last edited:
The Mobile GPU benchmark in question was 2048 shaders @ 1375Mhz = 390/X

Mobile parts are cut down parts from full fat and underclocked.

So we can speculate 2560 shaders @ 1600Mhz

Or put another way that is an R9 390 @ 1600Mhz

So, R9 390 is 1000Mhz + 600Mhz = + 60% @ 0.7 Scaling = R9 390 + 42% performance, that would put it right around a Titan-X, perhaps slightly more, which is exactly the 3DMark 11 score that was hinted at in that tweet.
 
I you could definitely be disappointed by it.
I was more referring to how since it was more than half the size of a 390x, then the slowest it could possibly be is above a 390x then adding the clockspeed and efficiency improvements to get the final performance.

It's not impossible that it could not be bigger than a 390x in raw physical game rendering terms (ie slower if using the same GCN and clock) but with technological advancements bringing the extra performances.

Kind of how like the 1080 isn't at a baseline a shrunken titan X in terms of core count even if the card is estimated to be slightly more than half the size.

Hopefully IPC improvements and clockspeed make for a great card, but there's no guarantee it'll be 980ti level at a minimum.

They have already said that GCN 4.0 is significantly more efficient and I would be amazed if there weren't other minor tweaks done and if they only use the same clock as the 390X then it would be very strange to me.

I understand your point though, and I would possibly agree if it wasn't pretty much a certainty that GCN 4 was much more efficient. That alone plus clock bumps pretty much assure that this will be very positive for gamers looking for bang for buck.
 
The Mobile GPU benchmark in question was 2048 shaders @ 1375Mhz = 390/X

Mobile parts are cut down parts from full fat and underclocked.

So we can speculate 2560 shaders @ 1600Mhz

Or put another way that is an R9 390 @ 1600Mhz

So, R9 390 is 1000Mhz + 600Mhz = + 60% @ 0.7 Scaling = R9 390 + 42% performance, that would put it right around a Titan-X, perhaps slightly more, which is exactly the 3DMark 11 score that was hinted at in that tweet.

Plus arch improvements of course.

Should be much more interesting than some are thinking.

Also wonder how much overclocking headroom it'll have. I've been reading even the Asus Strix (8+6 pin) 1080 can't make it to 2.1 GHz sustained clock in games, so the 1080 seems to be pretty close to max at stock.
 
The Mobile GPU benchmark in question was 2048 shaders @ 1375Mhz = 390/X

Mobile parts are cut down parts from full fat and underclocked.

So we can speculate 2560 shaders @ 1600Mhz

Or put another way that is an R9 390 @ 1600Mhz

So, R9 390 is 1000Mhz + 600Mhz = + 60% @ 0.7 Scaling = R9 390 + 42% performance, that would put it right around a Titan-X, perhaps slightly more, which is exactly the 3DMark 11 score that was hinted at in that tweet.

What was the score?
 
Plus arch improvements of course.

Should be much more interesting than some are thinking.

Also wonder how much overclocking headroom it'll have. I've been reading even the Asus Strix (8+6 pin) 1080 can't make it to 2.1 GHz sustained clock in games, so the 1080 seems to be pretty close to max at stock.

I said this a while ago that the 1080 cards were near their limit for OC'ing, this may be true of the 1070 also, i remember someone writing an article on it a week or so back, basically saying the cards would never be high overclockers as Nvidia had been very aggressive on the base and boost clocks.

Hopefully the Polaris stuff takes the crown(s) this time round for AMD, and the Vega stuff builds on it and is truly stunning performance wise, i dont mind paying a bit extra for a top tier AMD card this round if its the top performance card between both brands, i refused to pay for that abysmal Fiji rubbish they churned out, but i always said if they came back strong they will always get my money before Nvidia.
 
The Mobile GPU benchmark in question was 2048 shaders @ 1375Mhz = 390/X

Mobile parts are cut down parts from full fat and underclocked.

So we can speculate 2560 shaders @ 1600Mhz

Or put another way that is an R9 390 @ 1600Mhz

So, R9 390 is 1000Mhz + 600Mhz = + 60% @ 0.7 Scaling = R9 390 + 42% performance, that would put it right around a Titan-X, perhaps slightly more, which is exactly the 3DMark 11 score that was hinted at in that tweet.
Mobile parts are *usually* cutdown and underclocked.

We've seen lately that they've been putting in some complete desktop-class GPU's in them, though. And if AMD are trying to really push efficiency, it wouldn't surprise me if we saw some full P10-equipped laptops, similar to how we saw them with 980's.
 
I think we're all getting a bit too optimistic about it now. Realistically I expect top Polaris to be no more than 20% faster than 390x.
 
I think we're all getting a bit too optimistic about it now. Realistically I expect top Polaris to be no more than 20% faster than 390x.

Well even if it was, that still puts it at stock 980Ti and FuryX level.

So then if it was priced £300 or less, you'd be looking at a big winner.
 
I think we're all getting a bit too optimistic about it now. Realistically I expect top Polaris to be no more than 20% faster than 390x.

For £300 i'd be happy with that given the 390X is = 980 which is 970 + 15%.

An 8GB card that plays 1440P near to what my 970 plays 1080P gets my £300.
 
Last edited:
Back
Top Bottom