• 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

CAT-THE-FIFTH said:
Emm,its simple:
1.)GM206 filled the £100 to £200 range
2.)GM204 fill the £250 to £450 range.

So ultimately,I don't see what the problem is!

You would have smaller Polaris filling the £100 to £200 range and larger Polaris filling the £250 to £450 range.

Remember,Anandtech only estimated the size of around 120MM2 by looking at the finger of the AMD guy. It could easily be closer to 150MM2.

Plus remember,Nvidia themselves class the GTX960 as kind of console level/MOBA market and its easily faster than any of the console GPUs.

Edit!!

MOBA means games like LoL and DOTA2.

FFS,those games will play on an Intel IGP.
Click to expand...

"Power sipping GPU for thin and light notebooks, and mainstream desktops" - AMD.

£200 for that? Nope, nope, nope, nope, NOPE! Remember they demo'd it doing 1080p at 60FPS at medium settings. If it was a beast of a chip they'd have demo'd it against a 970 doing 60FPS at ultra settings.

Remember they said the power savings were huge across the range, so they had no need to demo vs a 950 if the card was capable of much more performance.

Either AMD are /really/ under-selling their little Polaris GPU, or we have an immense gulf in power between the bigger one and its "power sipping" little brother.

Mtom said:
Just pointing out that we don't know if it was a full small chip or already the smallest cut down piece.
Click to expand...

That I'm happy to concede.
 
Last edited:
FoxEye said:
"Power sipping GPU for thin and light notebooks, and mainstream desktops" - AMD.

£200 for that? Nope, nope, nope, nope, NOPE! Remember they demo'd it doing 1080p at 60FPS at medium settings. If it was a beast of a chip they'd have demo'd it against a 970 doing 60FPS at ultra settings.

Remember they said the power savings were huge across the range, so they had no need to demo vs a 950 if the card was capable of much more performance.

Either AMD are /really/ under-selling their little Polaris GPU, or we have an immense gulf in power between the bigger one and its "power sipping" little brother.
Click to expand...

Dude you are over egging it. Nvidia calls the GTX960 a MOBA class graphics card for LoL,DOTA2 and things like Heroes of the Storm. You can run these on an AMD APU at 1080P!!

So imagine what a console level graphics card is by Nvidia standards??

A GTX980TI??

The GM206 is a power sippin GPU which has found itself into many laptops too.

I have a GTX960 - it can run far more intensive games than those.

The HD5000 were power sippin too. Most GCN1.0 parts outside the HD7900 series were massive improvements in performance/watt over the previous Terascale based GPUs.

You have no clue what the actual performance of the small Polaris chip is or even how salvaged it was.

The chip is only running at a low 850MHZ and we don't know how much of it is disabled.

Don't believe me?? The HD7750 and HD7850 were the most efficient AMD GCN1.0 cards of their generation and they were all die salvaged parts. The HD7770 and HD7870 were the parts with the full fat GPUs.

I think you are over-analysing things to the level of pedantry.
 
Last edited:
We'll see. Tbh, after the last couple of releases/ generations by AMD, I am not prepared to give them the benefit of the doubt. I'm fully prepared for disappointment, and I think assuming everything is going to be awesome is the more foolhardy position to take :p

Don't have to remind anyone what's happened to AMD's R&D spend over the last few years...
 
CAT-THE-FIFTH said:
Dude you are over egging it. Nvidia calls the GTX960 a MOBA class graphics card for LoL,DOTA2 and things like Heroes of the Storm. You can run these on an AMD APU at 1080P!!

So imagine what a console level graphics card is by Nvidia standards??

A GTX980TI??

The GM206 is a power sippin GPU which has found itself into many laptops too.

I have a GTX960 - it can run far more intensive games than those.

The HD5000 were power sippin too. Most GCN1.0 parts outside the HD7900 series were massive improvements in performance/watt over the previous Terascale based GPUs.

You have no clue what the actual performance of the small Polaris chip is or even how salvaged it was.

The chip is only running at a low 850MHZ and we don't know how much of it is disabled.

Don't believe me?? The HD7750 and HD7850 were the most efficient AMD GCN1.0 cards of their generation and they were all die salvaged parts. The HD7770 and HD7870 were the parts with the full fat GPUs.

I think you are over-analysing things to the level of pedantry.
Click to expand...

Also the mid range 7870 (after some driver tweaks) was on par with the top 6970.
 
Mtom said:
Also the mid range 7870 (after some driver tweaks) was on par with the top 6970.
Click to expand...

It was actually even better than that. Matched or marginally beat a GTX 580, for just under half the price if memory serves. The 7870 was an amazing card relative to the market at launch.
 
OK, but what would have happened if the 7870 was the best chip in that generation? There was no 7970? Or it came out a year later?
 
FoxEye said:
We'll see. Tbh, after the last couple of releases/ generations by AMD, I am not prepared to give them the benefit of the doubt. I'm fully prepared for disappointment, and I think assuming everything is going to be awesome is the more foolhardy position to take :p

Don't have to remind anyone what's happened to AMD's R&D spend over the last few years...
Click to expand...

Amd have only had one under par gpu compared to Nvidia in Figi but even then it's not a really a slow chip it's just the gm110 cards are better. Sure the launches were not great but both Hawaii(Grenada)/Tahiti have proved they are better chips than gf104(gtx680)/gf110(gtx780ti) and are still keeping pace with gm104 (gtx970/980).
 
The HD7970 only used around a 360MM2 die. Especially as drivers started maturing it comfortably beat a GTX580 which was a 550MM2 40NM monster of a GPU. The HD6970,HD5870 and HD4890 all used sub 400MM2 dies.

The thing is that the first AMD top end GPUs on a new node have for the most part been 300MM2 to 400MM2 and remember GF/Samsung 14NM LPE(and probably 14NM LPP) is supposedly around 10% more denser than TSMC 16NM.

That might not sound much,but it would make a 360MM2 AMD high end GPU closer to 400MM2 if on the 16NM TSMC,and if you add the fact that the HBM means less of the die is taken up by the memory controller,it means they have plenty of transistors to work with.

I expect the large Polaris to be 30% to 40% faster than a Fury X,which is what we saw with the HD7970 when compared with the HD6970 and the HD5870 when compared to the HD4890:

https://www.techpowerup.com/reviews/AMD/HD_7970/28.html
https://www.techpowerup.com/reviews/ATI/Radeon_HD_5870/30.html

Even if that 120MM2 Polaris die had GTX950 level performance,a nearly 400MM2 die version,would probably have no issue gaining 30% extra performance:

https://tpucdn.com/reviews/EVGA/GTX_950_SSC/images/perfrel_1920.gif

Remember,Fiji is massively bottlenecked as a design,so it is leaving quite a bit of potential performance on the table.

The GTX285 to the GTX480 was 30% to 40% and from the GTX580 to the GTX680 it was around 20% or thereabouts at launch:

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_480_Fermi/32.html
https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_680/27.html

That is for first run high end products on a new node.

It could be quite possible Nvidia have a huge die product out for consumers before its smaller chips,but I expect any larger chip productions to be reserved for HPC customers first and for them to release another 300MM2 to 400MM2 GPU for gamers first.
 
Last edited:
either way... I'm going to be selling my 2 x 290x and getting a single new AMD or Nvidia card to play 4k on! Well... Not until after august anyway

The Future is bright!
 
TheRealDeal said:
Amd have only had one under par gpu compared to Nvidia in Figi but even then it's not a really a slow chip it's just the gm110 cards are better. Sure the launches were not great but both Hawaii(Grenada)/Tahiti have proved they are better chips than gf104(gtx680)/gf110(gtx780ti) and are still keeping pace with gm104 (gtx970/980).
Click to expand...

They didn't deliver with cayman either, but it's low performance didn't stop me upgrading from a 4870 to an unlocked 6950. Unfortunately this time round I was stuck on a 7950 and the nano was nowhere to be seen, I held out for the nano but like all the big tonga's they just didn't feel me with good feelings. So I went 2nd hand 970 which at 1080p and clocked to 1600mhz isn't far off the nano anyway. Gm204 was a great execution so was gm200, i'm hoping the bigger polaris will deliver and people need to compare polaris to pascal not maxwell, which amd always seems to be doing.
 
Davedree said:
They didn't deliver with cayman either, but it's low performance didn't stop me upgrading from a 4870 to an unlocked 6950. Unfortunately this time round I was stuck on a 7950 and the nano was nowhere to be seen, I held out for the nano but like all the big tonga's they just didn't feel me with good feelings. So I went 2nd hand 970 which at 1080p and clocked to 1600mhz isn't far off the nano anyway. Gm204 was a great execution so was gm200, i'm hoping the bigger polaris will deliver and people need to compare polaris to pascal not maxwell, which amd always seems to be doing.
Click to expand...

It needs to be compared to whatever Nvidia has out at the time it launches.
 
CAT-THE-FIFTH said:
I expect the large Polaris to be 30% to 40% faster than a Fury X,which is what we saw with the HD7970 when compared with the HD6970 and the HD5870 when compared to the HD4890:
Click to expand...

They also redesigned the engine in a way to elevate for latency and VR.
It could be they added features that simply matured the GCN tech with no significant bottlenecks which is what engineers normally design around.
Nano for example beats the 980ti with a mile with its small form factor and power envelope and adding a Polaris design to HBm2 will rock.

Polaris is up and running and demoed already and it excels what AMD did design it for.

A much brighter future with AMD Polaris
 
"Samsung Electronics announced that it has begun mass producing the industry’s first 4-gigabyte (GB) DRAM package based on the second-generation High Bandwidth Memory (HBM2) interface, for use in high performance computing (HPC), advanced graphics and network systems, as well as enterprise servers. Samsung’s new HBM solution will offer unprecedented DRAM performance – more than seven times faster than the current DRAM performance limit, allowing faster responsiveness for high-end computing tasks including parallel computing, graphics rendering and machine learning"

full samsung press release here

about time, now we got HBM2 ready, so high end GPUs might actualy be possible in 2016.
 
AlamoX said:
"Samsung Electronics announced that it has begun mass producing the industry’s first 4-gigabyte (GB) DRAM package based on the second-generation High Bandwidth Memory (HBM2) interface, for use in high performance computing (HPC), advanced graphics and network systems, as well as enterprise servers. Samsung’s new HBM solution will offer unprecedented DRAM performance – more than seven times faster than the current DRAM performance limit, allowing faster responsiveness for high-end computing tasks including parallel computing, graphics rendering and machine learning"

full samsung press release here

about time, now we got HBM2 ready, so high end GPUs might actualy be possible in 2016.
Click to expand...

Excellent, some actual news in this thread rather than Foxeye's ridiculous conjecture and trolling attempts.
 
You must log in or register to reply here.
Back
Top Bottom