AMD Polaris architecture – GCN 4.0

[andybird123]

The 40fps is also a guess - using Robert's figures of 62.5fps at 1080p, 1.83x scaling works out to 34fps

 

[ICDP]

So,if the GTX1070 is 60% faster than a GTX970,that means it is faster than a GTX1080?

That would rate the GTX1070 as 83% to 91% of a GTX1080,or 10% to 20% faster than a Geforce Titan. Ehh,no.

Seriously? If GTX 1080 is 100% and 970 is 52% that means 1080 is almost double the performance, or 100% faster. GTX 1070 is ~20%-25% slower than 1080, ergo 1070 is ~60%-70% faster than 970.

You are seeing 970 at 52% and 1080 at 100% and thinking 48% faster and that's not how percentages work.

I'm also not comparing price vs price but percentage increase on direct replacement GPUs, 380 vs 480 and 970 vs 1070. Admittedly the 380 to 480 increase is not known 100% but if the assumption that 480 = 390/970 then it's only ~40% faster on average over 380.

 

[AllBodies]

The 40fps is also a guess - using Robert's figures of 62.5fps at 1080p, 1.83x scaling works out to 34fps

The 40 fps comes from a test on the AotS database from a single 480.

Also the 34 fps can't be right because that'd be horrendous due to AMDs DX12 advantage. If it scored the same as a stock 970 in AotS, then it'd be much slower than a 970 in DX11.

 

[meldarthx]

The 40fps is also a guess - using Robert's figures of 62.5fps at 1080p, 1.83x scaling works out to 34fps

don't you mean 1.51x?

 

[Ferrari100]

What little reliable actual information from AMD suggests it is atleast 970 performance (capped 1440p game performance) - beyond that hard to nail down so just going based off the minimum at this point. Without at the very least a complimentary pair of cards at mainstream and lower high end points AMD is screwed in terms of product strategy to be frank.

Seriosuly?

'Without at the very least a complimentary pair of cards at mainstream and lower high end points AMD is screwed in terms of product strategy to be frank.'

Polaris has way more than a $1 billion in semi custom design wins already.
This GPU will sell boatloads on PC. Lets not forget, the real money is in consoles for AMD. Even the original PS4 / XBOX was pulling in $100 a sale at the start, newer versions will match, maybe even exceed that rate.

The result of AMD's console wins this time round is the reason they can offer VR entry level GPUs and reasonable performance to the masses at $200. That's a great thing for all gamers, VR or not.


I would say it's quite the opposite to being 'screwed' whether or not they release a higher end part - which will be great if they do but no big concern as the 480 will fly off shelves as is.

 

[chroniclard]

er, those figures are better than a Titan X results @ 1440p..........

am I missing something....

 

[TheRealDeal]

Who's to say that with the new design the 480's strong point may even be dx11 compared to cards like the 390/x. The new design should see it leverage more of it's performance in dx11 compared to these cards. In dx12 the 390/x get to stretch there legs and may be faster here than the 480. As most games atm are Dx11 come reviews it could look pretty good if this is the case.

 

[ICDP]

er, those figures are better than a Titan X results @ 1440p..........

am I missing something....

No but I think this is the problem. There are conflicting/confusing numbers coming from AMD and the guesses range from 970/390 like performance right up to Nano/Fury. If it's the higher end of that assumption then it will be a good result. If on the lower it's it's pretty poor to be honest IMHO.

 

[Rroff]

Seriosuly?

Yes seriously if they want to break the cycle and bring back home significant PC market share they have their strategy all wrong if they don't have a complimentary pair of cards at the right market points.

Might sound far fetched or me being negative but I'm looking at it realistically - if you go back over my posts around here in the past I've been pretty much dead on when AMD's market approach will and won't work in the past.

 

[CAT-THE-FIFTH]

snip

No,becuase you are not again reading what I am saying. Its not 40% over an R9 380 or GTX960,and remember I actually have a GTX960,with mates with GTX970 and R9 390 cards. I think you might find I wish they were only 40% faster than my card especially with games like ARK.

Either way,enough now.

Mods said not to bicker now literally two pages ago. So we can agree to disagree and leave it at that.

 

[Ferrari100]

Yes seriously if they want to break the cycle and bring back home significant PC market share they have their strategy all wrong if they don't have a complimentary pair of cards at the right market points.

Might sound far fetched or me being negative but I'm looking at it realistically - if you go back over my posts around here in the past I've been pretty much dead on when AMD's market approach will and won't work in the past.

Well lets hope for the better good of the gaming industry you are wrong this time.
Negativity NEVER helps, especially without facts.

 

[CAT-THE-FIFTH]

Yes seriously if they want to break the cycle and bring back home significant PC market share they have their strategy all wrong if they don't have a complimentary pair of cards at the right market points.

Might sound far fetched or me being negative but I'm looking at it realistically - if you go back over my posts around here in the past I've been pretty much dead on when AMD's market approach will and won't work in the past.

Dude seriously - JPR figures state 85% of sales under $300. I think of the 15% remaining 3% is over $449.

That 3% makes the most noise on tech forums. Nvidia started gaining marketshare from AMD with the GTX750TI.

This is when the rot started - all those OEMs could have 200W PSUs instead of the 300W ones and save £5 on production costs.

The same goes with laptops. AMD hardly moved forward with performance in that segment as performance/watt stalled with GCN starting with the R9 290 and R7 260 series onwards.

Nvidia didn't. So many laptops and desktops have GM107 or GM206 based parts in them.

The GTX750TI when it came out was not that great value - it couldn't beat many of the GTX660 and R9 270 cards for a few quid more,let alone the R7 260X.

But even I mentioned back then,Nvidia was going to do well in prebuilt desktops and laptops,and they did.

Edit!!

Even going back to the HD2000 series disaster.

ATI released the HD3870.

It could barely beat an 8800GT and Nvidia had the 8800GTS 512MB,8800GTX 768MB and 8800Ultra 768MB.

ATI could barely beat the 4th tier Nvidia card.

I had both the HD3870 GDDR4 and 8800GTS 512MB.

Yet,ATI marketshare never went under 30% and that is because the RV670 was a decent seller in prebuilt PCs,etc too down to the low price.

 

[Rroff]

Well lets hope for the better good of the gaming industry you are wrong this time.
Negativity NEVER helps, especially without facts.

Agreed on the aspect that the gaming industry needs a strong AMD.

 

[Mauller]

The 40fps is also a guess - using Robert's figures of 62.5fps at 1080p, 1.83x scaling works out to 34fps

But those averages were taken from results where it was quite CPU bound in some situations, so the single card performance could be higher than what you have calculated.

The engine may well be capable of 1.83x scaling, but it wont always be that dependent on how CPU bound it is.

 

[Rroff]

Dude seriously - JPR figures state 85% of sales under $300. I think of the 15% remaining 3% is over $449.

That 3% makes the most noise on tech forums. Nvidia started gaining marketshare from AMD with the GTX750TI.

There is more to making sales in a particular segment than just having a strong card in that segment - that 3% you mention for instance is quite important in driving the overall image of the brand, etc. which has a knock on effect.

 

[CAT-THE-FIFTH]

There is more to making sales in a particular segment than just having a strong card in that segment - that 3% you mention for instance is quite important in driving the overall image of the brand, etc. which has a knock on effect.

It doesn't look at my edit to the previous post. If you look at the period starting the 2900XT to the HD6970,ATI only had the fastest card once and they still defended their marketshare.

The ATI HD3870 was noticeably slower than an 8800GT(I had an HD3870 and a 8800GTS 512MB,so it was easy to check),and they never dipped below 30% and that was with Nvidia having 4 faster cards and if anything AMD has concentrated on the highend more times since GCN since then.

Yet,they lost marketshare.

AMD trying at all costs to match Nvidia on the highend is costing them.

They went off from the "small die strategy" to large dies and the costs have ballooned up for them and so is the money they make is even less.

AMD has to destroy Nvidia at the high end for enthusiasts to ever buy them,and if they match them it is not good enough.

Best for them to start small again and fight smaller challenges IMHO.

 

[TheRealDeal]

There is more to making sales in a particular segment than just having a strong card in that segment - that 3% you mention for instance is quite important in driving the overall image of the brand, etc. which has a knock on effect.

There is but people around these forums and most likely others are buzzing about the price/performance. The great sales of the gtx970 were nothing to do with the gtx980. It was more what people were getting for the money. I see nothing different here if the price and performance is spot on. Would it help if vega was out and beating up the gtx1080/70 most likely.

Amd have been going head to head with Nvidia to no avail so are trying something different. I think they can steel some sales here and maybe get some big contracts with this approach.

 

[bru]

Well just looking at threads like this one and others on different forums, AMD's idea seems to have backfired somewhat. They seem to have wanted to cause debate, about the RX480 performance, but everyone seems to be talking about NVidia cards instead. Of course if it wasn't AMD's idea to get people talking about the RX480 performance then they should have made their performance comparison nice and plain and simple, to avoid all this debate in the first place.

 

[ICDP]

Yes seriously if they want to break the cycle and bring back home significant PC market share they have their strategy all wrong if they don't have a complimentary pair of cards at the right market points.

Might sound far fetched or me being negative but I'm looking at it realistically - if you go back over my posts around here in the past I've been pretty much dead on when AMD's market approach will and won't work in the past.

I agree with this assessment. Though AMD don't need to trade blows at the top to win market share. Now if it was pushing the mainstream price/perf to a new tier such as Nano/Fury level then yeah it would be great. So I'm hoping a $300 level 480X type card exists that gives existing 970/390 owners a upgrade path.

If AMD think releasing a GPU that gives GTX970/ R9 390 performance for £50 cheaper is going to win them significant market share they are on a poor strategy. We have had that level of performance for ~£230 - £250 now for well over a year.

 

[CAT-THE-FIFTH]

Well just looking at threads like this one and others on different forums, AMD's idea seems to have backfired somewhat. They seem to have wanted to cause debate, about the RX480 performance, but everyone seems to be talking about NVidia cards instead. Of course if it wasn't AMD's idea to get people talking about the RX480 performance then they should have made their performance comparison nice and plain and simple, to avoid all this debate in the first place.

On most tech and gaming forums people are quite positive about it - places like PCMR on Reddit,etc. Remember 85% of consumer card sales are under $300 and that is from JPR.

This is actually one of the more negative places about it. The most vocal negative people tend to be a minority of enthusiasts who never would buy this kind of card anyway. It was the same negativity some had about the HD3850 and HD3870 and even the HD4850 and HD4870. Certain enthusiasts were negative about those too since they were not top dog but they did fine.

For them £50,here,£75 there,£40 here is nothing for them when for lower end parts it is actually a big difference. This is the same people who probably would never see the point of a Core i3 when a Core i5 is "only" £50 more.

People knew very well these cards would probably be close to R9 390X level for months - what AMD said months ago,the leaks. FFS,we knew Polaris 10 was 2304 shaders ages ago. I pretty much said it would be R9 390 to R9 390X for a very long time. It might even get close to a Fury if it was exceptional,which I started to doubt anyway.

Many were saying if it is R9 390/R9 390X level it needs to be £200ish or under otherwise it is a tad meh. This was stated many a time in this thread. I agreed too.

So,know we know it starts at £165 to £190,now the goalposts have changed.

So,anything AMD does won't be good enough and that is what AMD needs to fight more IMHO.