AMD Polaris architecture – GCN 4.0

[Ian Evey]

Having now destroyed my backup 7850 these just need to be better than Intel HD 3000 graphics for me to open my wallet Can you deliver that, AMD?

It will be around hd 3000 performance and up to around fury x performance so don't worry.

I have a feeling the price will be around £200 to £500.

 

[Kaapstad]

Having now destroyed my backup 7850 these just need to be better than Intel HD 3000 graphics for me to open my wallet Can you deliver that, AMD?

These are not the GPUs you are looking for move along, to the Vega system !!!

 

[FoxEye]

These are not the GPUs you are looking for move along, to the Vega system !!!

As tempting as that is, I'm going through a phase of breaking everything I touch Cheaper the better, tbh

 

[jupiter_man]

AMD Confirms Polaris GPU Positioning – Polaris 10 Aims The Mainstream Desktop/Notebook Market, Polaris 11 Targets Notebook Market



According to AMD, the Radeon R9 Fury series cards are enthusiast grade products. Everything below them that includes the Radeon R9 390 and Radeon R9 380 series are mainstream cards.

AMD Polaris 10 Allegedely Performs Close To A GTX 980 Ti, Extremely Efficient and Competitive Design
We also want to share some information we learned from our sources about Polaris 11. AMD recently hosted a event in Taiwan to showcase their Polaris GPUs (Polaris 10 and Polaris 11) along with the Radeon Pro Duo card to journalists. We shared slides of the Radeon Pro Duo from that event yesterday. People were able to get some info out of AMD and it seems like the Polaris 10 can be an extremely competitive product.

The AMD Polaris 10 GPU has a maximum TDP of 175W but cards will actually consume much less than that. The GPU was initially built to support HBM memory but AMD chose to go the GDDR5/X route since it offers a better value currently. We will get to see HBM on AMD GPUs when Vega launches but until then, only Fury series will have HBM support. The Polaris 10 GPU is said to have 3DMark Firestrike Ultra performance around 4000 points which is about what a Radeon R9 Fury X and GeForce GTX 980 Ti score. By 4000 points, we don’t mean exactly 4000 but it’s actually quite a bit less but that’s the number we were told.

tbh I am quite disappointed...a new tech with half the size should easily do better than the old tech. I was expecting even the mid range card should be easily 10-20% faster than 980ti / Fury X and new top range should be 30-50% faster making 4K playable.

 

[TheRealDeal]

tbh I am quite disappointed...a new tech with half the size should easily do better than the old tech. I was expecting even the mid range card should be easily 10-20% faster than 980ti / Fury X and new top range should be 30-50% faster making 4K playable.

It's actually more like a 1/3 of the size so pretty impressive if they get near FuryX/ti. Who knows what they showed off whether it was clocked low or limited by a driver to not show off its full potential. We still don't know to much.

 

[AlamoX]

tbh I am quite disappointed...a new tech with half the size should easily do better than the old tech. I was expecting even the mid range card should be easily 10-20% faster than 980ti / Fury X and new top range should be 30-50% faster making 4K playable.

polaris 10 is like 3rd the size of the fury/ti not half, and for playable 4k you wont have that untill they can make chips around 450-500+mm² size.
so 4k not going to be playable untill 2017 with vega/titan if we are lucky

 

[LoadsaMoney]

So according to the article above the Polaris 10 will be close in performance to a gtx980ti/fury X and will be priced in the mainsteam. It also has a max tdp of 175W but will use way less.

This card sounds a lot like the Gtx970/gtx780ti and we all knew how that turned out. Sounds to me if priced right this thing could sell like hot cakes.

Keep in mind Nvidia's chip is a lot bigger so should be faster but will have to be priced higher. It sounds to me like AMD are ready to be very aggressive on pricing.

If they are below Fury X performance, then they aren't going to be worth an upgrade for anyone with a 2/3 series card, they'd be best off going Nvidias Pascals.

 

[AlamoX]

If they are below Fury X performance, then they aren't going to be worth an upgrade for anyone with a 2/3 series card, they'd be best off going Nvidias Pascals.

but they might be worth the upgrade from Nvidia users ranging from 950 to 970, and judging by the market share AMD holds right now it might be worth it.
again it all depands on the pricing, but AMD seem to be going for cost effective, and Nvidia going to top end perf, beside if games keep coming out well optimised for GCN cards, this might be the break AMD need to get some market share back.

 

[Freddie1980]

Good ol Senator Vernak. That recording Sisko and Garek cooked might have been fake but that bomb on his shuttle certainly wasn't.

 

[Freddie1980]

tbh I am quite disappointed...a new tech with half the size should easily do better than the old tech. I was expecting even the mid range card should be easily 10-20% faster than 980ti / Fury X and new top range should be 30-50% faster making 4K playable.

To be fair the 980ti should have been built on 20nm and used less power, the high end enthusiast cards we have at the moment are at unprecedented levels of tdp as result of the failed process node from tscm.

 

[TheRealDeal]

If they are below Fury X performance, then they aren't going to be worth an upgrade for anyone with a 2/3 series card, they'd be best off going Nvidias Pascals.

Fury X can be up to 20-30% faster so if the pricing is right then it could be worth it. Plenty went from a gtx980 to a ti which is around the same gap. Who knows maybe it will be an Overclockers dream as well

The pricing needs to be right if it's not fast enough.

 

[drunkenmaster]

To be fair the 980ti should have been built on 20nm and used less power, the high end enthusiast cards we have at the moment are at unprecedented levels of tdp as result of the failed process node from tscm.

Unprecedented.... yet the same as the previous generations?

Titan X/980ti 250W tdp.

780/780ti/Titan 250W, Titan Z 375W.

690 300W

580 244W, 590 365W.

480 250W

280 236W, 295 289W.


Then AMD side, not dissimilar. Sorry but way below actual single card max and the same as anything up to and including the 480gtx makes the 980ti nothing close to unprecedented TDP wise.

Also for the record there is nothing wrong with 20nm, there were those of us on the forum saying 20nm was dead in the water for GPUs 2-3 years ago for a reason, the entire semi conductor industry knew planar was effectively DOA below lets call it 25nm, Intel's 22nm is extremely generously named, it's WAY closer in size to mature 28nm glofo than 20nm.

20nm isn't a failed process, it's failed hyping from crap computer news sites trying to tell everyone 20nm gpus were coming. Nodes need both double transistor density as well as half the power per transistor for 'normal' scaling. This was never remotely advertised for 20nm, it was never going to come close. 14/16nm finfets provided the power drop while 20nm provided the transistor density, together they break past what is basically a planar barrier around 22nm. 20nm itself brought around 1.8 transistor density but only around 15% or so power drop. Take a 40nm 600mm^2 chip using 250W then on 28nm you can make a 300mm^2 chip with the same amount of transistors but using 125W or make something 600mm^2 with twice the transistors and same power. 20nm would mean that same chip with double the transistors would be 500W(twice the original chip then scaled for power) * 0.85 = 425W. This is why 20nm was a no go, a big chip would have increased power by 70% for maybe 70% more performance.... meaning no performance/watt gain for crappy yields and significantly increased wafer cost.

 

[maonayze]

Well, I'm no graphics technical expert, however it does seem from all we are hearing (If there is any truth to any of it) that both Nvidia and AMD are reverting back to type this time around.

We have AMD aggressively targeting Value/Price/Performance and Nvidia going for the Big Chip/Brute force approach of old or so it seems.

And will that now mean all Nvidia users will completely ignore and dismiss the benefits of lower power consumption as AMD cards will either be on an equal footing or better?

 

[Seanspeed]

i think ppl are expecting way too much from polaris, with a 232mm² i see it more about the 390 perf, and far from fury/ti, maybe nvidia 1080 will be at fury/ti perf or even 20% faster, one thing is sure nvidia mainstream pascal will be no where near as competitive as AMD's price wise, and fury will stay enthusiast card for them untill vega shows up

Well at 2x performance/watt, I'd say that Polaris 10 should probably be slightly ahead of a 390X. Basically bridges the gap between the x90 cards and the Fury line-up, not that there was that big of a gulf in the first place.

Which makes me think it'll be something of an awkward situation.

Will be interesting to see how they handle pricing in that case. The only way they're going to get people excited about it over a 390/390X is through price, really. So I think they'd basically be looking at a similar 'range' of products that they already had, simply with better prices. Which is good. But I think that's going to fail to get people excited.

Nvidia are going to be able to cut prices too. Maybe not quite as much, but they'll potentially have the performance to justify it if the rumors about die sizes are correct between the two.

I'm unsure that's really going to drastically affect the current status quo unless AMD's pricing is so aggressive it completely sweeps the lower/mid range market completely out from Nvidia's feet by a landslide. The problem I see is that Nvidia's current market share, combined with a potential public perception of being the more 'performance-orientated' brand with the new cards, will still affect AMD's ability to do such a thing even if the pricing is drastically in their favor.

 

[Stanners]

Well, I'm no graphics technical expert, however it does seem from all we are hearing (If there is any truth to any of it) that both Nvidia and AMD are reverting back to type this time around.

We have AMD aggressively targeting Value/Price/Performance and Nvidia going for the Big Chip/Brute force approach of old or so it seems.

And will that now mean all Nvidia users will completely ignore and dismiss the benefits of lower power consumption as AMD cards will either be on an equal footing or better?

And vice versa of course. The whole cycle would start again that we have been through many times over the years.

If it were to happen.

 

[Seanspeed]

Well, I'm no graphics technical expert, however it does seem from all we are hearing (If there is any truth to any of it) that both Nvidia and AMD are reverting back to type this time around.

We have AMD aggressively targeting Value/Price/Performance and Nvidia going for the Big Chip/Brute force approach of old or so it seems.

And will that now mean all Nvidia users will completely ignore and dismiss the benefits of lower power consumption as AMD cards will either be on an equal footing or better?

People want lower power consumption AND power. Or efficiency, in other words. The argument about that was never in total isolation like you're making it sound. With Maxwell, Nvidia were able to provide better-than-Titan performance while drastically cutting power consumption at the same time. That's what the 'excitement' about that was. Maxwell was simply a more efficient architecture than what AMD had(has, technically).

If AMD can get on equal footing here with Polaris, then awesome. I dont think anybody will deny that is a good thing. However, if AMD cards use less power but are proportionally less powerful, then no, nobody is going to consider that a 'win' for AMD(though notebook and small form factor users will be happy to have even more options).

 

[flopper]

Well, I'm no graphics technical expert, however it does seem from all we are hearing (If there is any truth to any of it) that both Nvidia and AMD are reverting back to type this time around.

We have AMD aggressively targeting Value/Price/Performance and Nvidia going for the Big Chip/Brute force approach of old or so it seems.

And will that now mean all Nvidia users will completely ignore and dismiss the benefits of lower power consumption as AMD cards will either be on an equal footing or better?

a 250$ 390+ performance card is a golden deal for anyone with a 960/380 or lower card. Unbeatable performance/price delta.

at least speculation is going on while the stars continue to shine brightly

 

[CAT-THE-FIFTH]

Well at 2x performance/watt, I'd say that Polaris 10 should probably be slightly ahead of a 390X. Basically bridges the gap between the x90 cards and the Fury line-up, not that there was that big of a gulf in the first place.

Which makes me think it'll be something of an awkward situation.

Will be interesting to see how they handle pricing in that case. The only way they're going to get people excited about it over a 390/390X is through price, really. So I think they'd basically be looking at a similar 'range' of products that they already had, simply with better prices. Which is good. But I think that's going to fail to get people excited.

Nvidia are going to be able to cut prices too. Maybe not quite as much, but they'll potentially have the performance to justify it if the rumors about die sizes are correct between the two.

I'm unsure that's really going to drastically affect the current status quo unless AMD's pricing is so aggressive it completely sweeps the lower/mid range market completely out from Nvidia's feet by a landslide. The problem I see is that Nvidia's current market share, combined with a potential public perception of being the more 'performance-orientated' brand with the new cards, will still affect AMD's ability to do such a thing even if the pricing is drastically in their favor.

The reason Nvidia did very well is because they got loads of OEM sales in laptops and desktops which destroyed the gains AMD had made with the HD5000 and HD6000 series.

If they get a lot of OEM wins in laptops and desktops it will mean they will gain decent marketshare.

Even,in DIY desktops Nvidia did very well since the GTX970 was priced decently and had a good balance of power consumption and price.

Remember,the HD3000 and HD4000 series were not performance kings but ATI still managed to sell enough cards to keep a reasonable marketshare.

The HD6970 could not beat the GTX580 and so on.

Also,AMD has gained desktop dGPU marketshare in the last two quarters it seems.

IMHO,as long as AMD PR don't screw up with the details,it should help AMD gain marketshare.

 

[Rannoch]

Once proper benchmarks are released I'll believe it.

 

[Seanspeed]

a 250$ 390+ performance card is a golden deal for anyone with a 960/380 or lower card. Unbeatable performance/price delta.

It's a good deal, but it's not that magical unless it's at least clearly 'better than 390' level.

I think if AMD could do a 490 that is very close to a 390X for $200 and then a 490X that beats a 390X for $250, it would be hard for Nvidia to match.

I expect the GTX 1060 to be around GTX 970 performance at least, and that will likely be a $200 card, so that's kind of what AMD should be thinking about.