AMD Polaris architecture to succeed Graphics Core Next

Hm that's same or less shaders than PS4 Neo.

It's the same number of shaders as a 380/280X actually. Meaning performance should be 15 - 25% better than the current 280X depending on the clocks. It will undoubtedly be far more efficient power consumption wise though.

A quick calc:

280/380X TF* - 3.5
4XX TF** - 5.5

That would indicate 57% increase in floating point performance (which nobody would believe)

*According to AMD
**According to leak
 
I'm hoping they release a 390x level polaris10 at ~200 dollars
that's the only way consumers get out ahead from all this.

gp104 is impressive as hell but is insanely expensive
amd have nothing for the higher midrange till 2017

As another poster said, getting another hd4870 would be absolutely wonderful for gamers.
Not as powerful, but great performance/dollar.
The hd4870 is still the best value gpu I ever bought. Polaris 10 could be similar, if amd price it correctly
 
It's the same number of shaders as a 380/280X actually. Meaning performance should be 15 - 25% better than the current 280X depending on the clocks. It will undoubtedly be far more efficient power consumption wise though.

A quick calc:

280/380X TF* - 3.5
4XX TF** - 5.5

That would indicate 57% increase in floating point performance (which nobody would believe)

*According to AMD
**According to leak

280 is up to 3.3 TF (1792 SPs x .933 GHz boost clock x 2 = 3,344 GFLOPS), and 380X is 3.9 TF (2048 SPs x .970 GHz x 2 = 3,973 GFLOPS). And yes, that formula is how you calculate FLOPS. A 32CU, 5.5 TFLOPS GPU would run at around 1.35GHz, which is believable for 14nmFF.
 
280 is up to 3.3 TF (1792 SPs x .933 GHz boost clock x 2 = 3,344 GFLOPS), and 380X is 3.9 TF (2048 SPs x .970 GHz x 2 = 3,973 GFLOPS). And yes, that formula is how you calculate FLOPS. A 32CU, 5.5 TFLOPS GPU would run at around 1.35GHz, which is believable for 14nmFF.

True, though I was attempting to indicate that a 57% FLOP increase would not necessarily mean a 57% gaming performance increase. Got my own thinking/words muddled up, apologies. In the unlikely event it does and the card is priced sub $250 (or at the current $230 price for the 380X) then it would be a definite win for consumers
 
I just realized that the 7970 (Ghz Ed.) launched at 499USD in 2012, regular 7970 even at 549USD couple months earlier.

How the fuck was I able to buy one for around 320€ new in late 2012???

Yeah prices spiked something awful. AMD really needs to shake things up.
 
It's the same number of shaders as a 380/280X actually. Meaning performance should be 15 - 25% better than the current 280X depending on the clocks. It will undoubtedly be far more efficient power consumption wise though.

A quick calc:

280/380X TF* - 3.5
4XX TF** - 5.5

That would indicate 57% increase in floating point performance (which nobody would believe)

*According to AMD
**According to leak

this post cant be serious right?
 
This is the strangest thing I have read in a while. What do you mean with otaku-ification of the pc harware market?
I read it as being as the market has stopped growing, the hardware manufactures have to milk their enthusiast market even more. It doesn't help that hardware upgrade junkies fall over themselves for kit that gives marginal gains to what they've already got; witness all those 980ti owners chomping at the bit for a Founder's Edition 1080's.
 
I read it as being as the market has stopped growing, the hardware manufactures have to milk their enthusiast market even more. It doesn't help that hardware upgrade junkies fall over themselves for kit that gives marginal gains to what they've already got; witness all those 980ti owners chomping at the bit for a Founder's Edition 1080's.

The gaming hardware market is still growing so that isn't it.
 
I'm hoping they release a 390x level polaris10 at ~200 dollars
that's the only way consumers get out ahead from all this.

gp104 is impressive as hell but is insanely expensive
amd have nothing for the higher midrange till 2017

As another poster said, getting another hd4870 would be absolutely wonderful for gamers.
Not as powerful, but great performance/dollar.
The hd4870 is still the best value gpu I ever bought. Polaris 10 could be similar, if amd price it correctly

That's that I'm hoping for too, the real midrange has been stagnant for ages now. If you buy anything slower than a 970 or 390 you're barely getting any upgrade at all from 3-4 year-old cards.
 
True, though I was attempting to indicate that a 57% FLOP increase would not necessarily mean a 57% gaming performance increase. Got my own thinking/words muddled up, apologies. In the unlikely event it does and the card is priced sub $250 (or at the current $230 price for the 380X) then it would be a definite win for consumers

It's true that the boost wouldn't be quite that big, but 15-25% is low-balling it. That's also assuming that it even tops out at 32 CUs.
 
The gaming hardware market is still growing so that isn't it.
In relative numbers, yes. Of people owning desktop builds, an increased percentage of them went for enthusiast (and even hyper-enthusiast) level hardware. However, desktop installations at large dropped again. So, the enthusiast segment is growing in an overall shrinking market. I call this phenomenon otaku-ification.

AMD Takes More GPU Market Share From Nvidia In Q1 2016 – Builds Momentum Ahead Of Polaris Launch

http://wccftech.com/amd-takes-gpu-share-nvidia-q1-2016/
Laptop numbers are, sadly, more 'massaged' than your average tourist in a Ho Chi Minh parlour. Most of the "graphics units" in sold AMD laptops are either integrated or low-end DDR3 affairs. I should know; I own one.
 
I'm going AMD if they can match gtx1070 for cheaper and hopefully lower power consumption.

They don't even need to match the 1070 IMO. The need to release a card with 80% of the performance and 60-65% the price of a 1070. That would be a good value mainstream card to me. As for power consumption, it's already low (150W) on the 1070, shouldn't be an issue on Polaris either.
 
I read it as being as the market has stopped growing, the hardware manufactures have to milk their enthusiast market even more. It doesn't help that hardware upgrade junkies fall over themselves for kit that gives marginal gains to what they've already got; witness all those 980ti owners chomping at the bit for a Founder's Edition 1080's.

You say falling over themselves, for some it's likely just a case of putting aside a fraction of that month's pay for a new toy.
 
Pricing will be interesting on these cards. I hope for my wallet that AMD takes a leap down to 380x instead of a 280x or 390. Also it would be nice to get some gtx 1060 info before these release.
 
I hope we will get some Polaris 11 announcements as well. I'm considering upgrading from my 7850 to something that will run, like say, Doom at a decent framerate at 1200p.
 
AMD-Radeon-R9-480-3DMark11-Performance.png
http://videocardz.com/60253/amd-radeon-r9-480-3dmark11-benchmarks

according to this P10 is almost equal to a Fury, grain of salt of course.
 
Looking at the 3dmark pages, both lower results would be below 390x, still I don't know of any released AMD card with memory clocks that high, can you fake something like that?
 
That is a joke. The editor made a deal that if a person's comment got so many likes that he would post a P10 benchmark.

Funnily enough, that probably won't be too far from the truth.

Looking at the 3dmark pages, both lower results would be below 390x, still I don't know of any released AMD card with memory clocks that high, can you fake something like that?

Scores are easily faked, even the ones that might seem legit. It's happened so many times in the past that I've lost count.
 
if this was true and the 1070 turns out to be faster than a titan x for $370, amds product wont be very compelling at $299

Well, yeah, it would fit right under 1070 and directly against 1060 or 1060Ti. Which is what any sane person should expect. There is no reason why AMD would be able to sell a considerably faster GPU at the same price in the mid range. Top end is another story but AMD will be absent from it till Vega.
 
Well, yeah, it would fit right under 1070 and directly against 1060 or 1060Ti. Which is what any sane person should expect. There is no reason why AMD would be able to sell a considerably faster GPU at the same price in the mid range. Top end is another story but AMD will be absent from it till Vega.

Except for the fact that nvidia are gouging the shit out of us (just like amd gouged the shit out of everyone when they were first to market at 28nm)


There's PLENTY of room for price wars and then some.
 
Except for the fact that nvidia are gouging the shit out of us (just like amd gouged the shit out of everyone when they were first to market at 28nm)


There's PLENTY of room for price wars and then some.

No there isn't, if we're talking about the mid range. You haven't noticed my note on the top end being a different story? Polaris is already limited by 1070's $380 price, they can't "gouge" us any higher than that. And $350 ceiling for a new chip on a new process isn't very much.
 
Funnily enough, that probably won't be too far from the truth.

Looking at it further, it appears to be leaked 3DMark scores. But the editor says:

One more thing, according to Sisoft database, C7 has 2304 cores, there is a possibility that there's another P10 chip with 2560 cores (full silicon), even faster than those. Just wanted to make it clear.

So my previous comment was incorrect. But the scores should not be looked at with any validity until they are revealed.
 
if this was true and the 1070 turns out to be faster than a titan x for $370, amds product wont be very compelling at $299

1.26 GHz seems a bit low for 14nm so I would not be surprised to see the final product sporting 1.3 GHz - 1.5 Ghz final clock speeds. It is also possible that there is a 2560 shader part too.

I think 480X will be around Fury X performance for $299. Less than that would be disappointing for a node shrink.
 
Well, yeah, it would fit right under 1070 and directly against 1060 or 1060Ti. Which is what any sane person should expect. There is no reason why AMD would be able to sell a considerably faster GPU at the same price in the mid range. Top end is another story but AMD will be absent from it till Vega.

puts them in a significantly worse spot then they are in right now. if that turns out to be the situation than its clear amd had serious problems transitioning gcn to 14nm. pascal is nothing more than maxwell at higher clocks. gcn 1.2 was already faring better than maxwell so with further improvements to the architecture(something completely absent from gp104) they should have pulled ahead.
 
Top Bottom