• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Polaris performance numbers perf/$

schizo said:
That pricing would be disastrous. Next-generation, same performance, exact same price. Oh, and it runs cooler and is more power-efficient-- but who really gives a crap about that on the desktop?

A very low-power GPU with 390x-level performance would be interesting for laptops, but that assumes GP106 doesn't compete in that space. Nah. Prices have to come down.
Click to expand...
People in the mid range and low end do, as do OEM's, power consumption is very important them since they can cut down on component costs on desktops thus dropping entire system costs.
 
schizo said:
That pricing would be disastrous. Next-generation, same performance, exact same price. Oh, and it runs cooler and is more power-efficient-- but who really gives a crap about that on the desktop?
A very low-power GPU with 390x-level performance would be interesting for laptops, but that assumes GP106 doesn't compete in that space. Nah. Prices have to come down.
Click to expand...

This whole roller coaster ride of rumours starting with 980TI performance then the price then back and forth between performance and configuration. These are mid range cards and we all know it , Vega (greenland) will be with HBM2 that is the one to watch performance wise.

What you sum up here is exactly what most of us are thinking then again most would not buy a such a card even if it was cheap and there is the difference between the demographic of [H]ardocp and what AMD sales are they will more then likely sell a boatload of these cards even tho "none of us" find them appealing ..
 
Well better than semi accurate, Fuad learned his lesson in predicting lol, Charlie still hasn't.
 
Daniel_Chang said:
Your first link compared cards with the following core overclocks:
GTX 970 - 4.95%
GTX 980ti - 11.78%, 12.16%, and 12.53%
.
Click to expand...

The 970 and 980ti have different base/boost clocks so everything you say is invalid; just because their relative overclocks are different it means nothing, you want to compare clock for clock, not relative overclock to relative overclock.

The 980ti boosts to 1076, the 970 to 1178

Last time I repeat myself to you because I know frustration is going to kick in very fast, you've been stuck on this since yesterday, it's not that hard.
 
I don't know about you guys but 390x performance for $300 is a little less of a performance increase than I expect, but then again the power savings are the star of the show. Don't forget AMD is behind both in terms of power efficiency and die area efficiency, it will be a resounding success if they achieve their targets in one iteration

Disappointing for people wanting higher performance at same $

And, btw, only 390x performance in compute bound situations

In geometry bound situations it should be twice as fast
 
I find it very hard to believe that the power savings will be worth the extra $100 this card will cost over GP106 (assuming GP106 isn't a dog) at the same performance level on desktops. Yes, that means OEMs can skimp on their PSUs. But we're talking about a hundred dollars.

Mobile, it has a chance for sure.
 
schizo said:
I find it very hard to believe that the power savings will be worth the extra $100 this card will cost over GP106 (assuming GP106 isn't a dog) at the same performance level on desktops. Yes, that means OEMs can skimp on their PSUs. But we're talking about a hundred dollars.

Mobile, it has a chance for sure.
Click to expand...

GP106 isn't just one SKU, there gonna be at least two versions (960/950)

Assuming GP104-150 isn't a x60Ti or something, that leaves GP106 vs Polaris 10/11
~

If the new x70 matches the 980ti (one performance bracket up, same price approximately as 970.

Then the new x60 based on GP106 should offer 970/980 tier performance
 
Keep in mind that you can buy a GTX970 or R9 390 today for $300.

My guess (and we're all just speculating here) is that both Polaris10 and GP106 come in at 390/GTX970 level performance, MSRP around $250. I would like to see $200, and I think that's possible too, but a bit optimistic.

Then again AMD keeps talking about bringing VR to the mainstream, and $200 is the mainstream pricepoint.
 
schizo said:
Keep in mind that you can buy a GTX970 or R9 390 today for $300.

My guess (and we're all just speculating here) is that both Polaris10 and GP106 come in at 390/GTX970 level performance, MSRP around $250. I would like to see $200, and I think that's possible too, but a bit optimistic.

Then again AMD keeps talking about bringing VR to the mainstream, and $200 is the mainstream pricepoint.
Click to expand...

bringing VR to the mainstream is typical marketing BS.

If they wanted to bring VR to the mainstream they should focus on producing VR headsets that cost less than 600$
 
Yes, of course. AMD has a great offering at the 390 performance level with Polaris10, which coincidentally is what you need for current-gen VR. Polaris 10 is a tiny chip, cheap to make, and sips power. That's what AMD is bringing to the table.

"Bring VR to the mainstream" is AMD's way of spinning the fact that they have no next-gen high-end or enthusiast GPUs to offer until Q1 2017 because they bet on HBM2.

That said, a 390-class card for $200 would be one tasty meatball.
 
that is true, even if card prices drop there is still a cost to VR that is much higher.
 
schizo said:
Yes, of course. AMD has a great offering at the 390 performance level with Polaris10, which coincidentally is what you need for current-gen VR. Polaris 10 is a tiny chip, cheap to make, and sips power. That's what AMD is bringing to the table.

"Bring VR to the mainstream" is AMD's way of spinning the fact that they have no next-gen high-end or enthusiast GPUs to offer until Q1 2017 because they bet on HBM2.
Click to expand...

What I think when AMD says :

"Bring VR to the mainstream"

masturbation_fb_201361.jpg
 
schizo said:
Yes, of course. AMD has a great offering at the 390 performance level with Polaris10, which coincidentally is what you need for current-gen VR. Polaris 10 is a tiny chip, cheap to make, and sips power. That's what AMD is bringing to the table.

"Bring VR to the mainstream" is AMD's way of spinning the fact that they have no next-gen high-end or enthusiast GPUs to offer until Q1 2017 because they bet on HBM2.

That said, a 390-class card for $200 would be one tasty meatball.
Click to expand...
Well I wouldn't say cheap to make, it probably costs the same amount as a 390x to make. Which by today's standards yeah it should be in the mid range price targets now.
 
razor1 said:
that is true, even if card prices drop there is still a cost to VR that is much higher.
Click to expand...
Also we're still in first-gen VR. You still get the screen-door effect using them. Next year second-gen VR will come out with doubled resolution per eye, and a 390/970 won't exactly keep up.

razor1: True, but prices will continue to drop as the 16nm process matures. Can't say that about 28nm. That tiny die will pay off for AMD.
 
razor1 said:
Well I wouldn't say cheap to make, it probably costs the same amount as a 390x to make. Which by today's standards yeah it should be in the mid range price targets now.
Click to expand...

Polaris 10 is 1/4 the size and requires half the power regulation and cooling. It will be a much cheaper card to manufacturer.
 
schizo said:
Also we're still in first-gen VR. You still get the screen-door effect using them. Next year second-gen VR will come out with doubled resolution per eye, and a 390/970 won't exactly keep up.
Click to expand...

mainstream VR is several years away, and I'm much more interested in innovative gameplay than cool graphics, I already know it'll be years before VR can achieve graphic fidelity games have today
 
Yeah this is why I think marketing for VR is just a bit too early still a couple of quarters away.
 
Quix said:
Polaris 10 is 1/4 the size and requires half the power regulation and cooling. It will be a much cheaper card to manufacturer.
Click to expand...

It's not a 1/4 the size unless Hawaii was 920mm^2 and nobody told me

Polaris is roughly half the size, but transistor cost has increased, so if it has equal numbers of xtors ( 6.2Bn) it'll cost more than 390x to produce.
 
Quix said:
Polaris 10 is 1/4 the size and requires half the power regulation and cooling. It will be a much cheaper card to manufacturer.
Click to expand...
hmm its not 1/4 the size its 1/2 the size and that won't cut down costs much, cost savings will come down to layers of PCB and traces due to the lower power consumption which is very little in comparison to all the components.
 
razor1 said:
Yeah this is why I think marketing for VR is just a bit too early still a couple of quarters away.
Click to expand...
"Bring VR to the mainstream" is pure marketing. The only things you can take from it are:

1) AMD probably doesn't have high-end/enthusiast parts right now
2) The mainstream parts they do have will be priced appropriately for the mainstream (meaning $250 at most)
 
I think so too but I don't see why they didn't aim for high end markets, they should be able to do what nV is doing, yeah there might be extra costs due to lower yields, and power consumption is going to be higher due to the vram they would be using vs. HBM but looking at Polaris (rumored specs) they should be able to something out with respectable power consumption vs the competition, even if its 10% higher in that category it doesn't really matter since the performance is more important.

I think they saw something when they crunched the numbers (profits, cost etc.) which held them back.

Of course once we see what Vega brings to the table that will give us a better idea.
 
razor1 said:
I think so too but I don't see why they didn't aim for high end markets, they should be able to do what nV is doing, yeah there might be extra costs due to lower yields, and power consumption is going to be higher due to the vram they would be using vs. HBM but looking at Polaris (rumored specs) they should be able to something out with respectable power consumption vs the competition, even if its 10% higher in that category it doesn't really matter since the performance is more important.

I think they saw something when they crunched the numbers (profits, cost etc.) which held them back.

Of course once we see what Vega brings to the table that will give us a better idea.
Click to expand...

I just hope they change the name because every time someone says Vega I think of this
1353.gif
 
It's definitely the numbers; Raja Koduri spoke about it in that PCper interview. They aren't focusing on those huge dies anymore, the economics just don't make sense. They want to make tiny dies cheaply with higher yields and put a bunch of them on a card.

For Vega, they bet on HBM2 and it just isn't ready yet.
 
schizo said:
It's definitely the numbers; Raja Koduri spoke about it in that PCper interview. They aren't focusing on those huge dies anymore, the economics just don't make sense. They want to make tiny dies cheaply with higher yields and put a bunch of them on a card.

For Vega, they bet on HBM2 and it just isn't ready yet.
Click to expand...

I don't think it's just about HBM2, yields mainly but anyway, yeah they are betting on smaller dies now, but the thing is they have been a disadvantage in this respect for the last few years. They can't just make their dies smaller for free, something's gotta give, what will it be ? Obviously there will be a die shrink because of the process, that's not what I mean, I mean how will they compete with NV at die size parity ? NV has a totally different architecture with well-known concessions made to reduce die area
 
Ieldra said:
I don't think it's just about HBM2, yields mainly but anyway, yeah they are betting on smaller dies now, but the thing is they have been a disadvantage in this respect for the last few years. They can't just make their dies smaller for free, something's gotta give, what will it be ? Obviously there will be a die shrink because of the process, that's not what I mean, I mean how will they compete with NV at die size parity ? NV has a totally different architecture with well-known concessions made to reduce die area
Click to expand...

Easy. You can buy 2 AMD cards for around the same price as an Nvidia flagship. AMD and Nvidia are promising that mGPU is going to be the way to VR nirvana in the future. If the cards are around 150w in stock form you could run 4 in a system with a 850w power supply.

R.I.P. Corsair 1500w power supplies. :(
 
Personally I do not agree with the seemingly revulsion (and by this I refer to both extremes with Polaris 10 being in the 390x range as opposed to higher. That performance target at lower price in my opinion is much better to a larger amount of parties then a hypothetical higher target and higher price.

Of course the variable in this is how efficient the cost structure actually is and how much of it is transferred to the consumer.

At the end I'd prefer AMD go back to their pre-GCN strategy versus post-GCN.
 
cageymaru said:
Easy. You can buy 2 AMD cards for around the same price as an Nvidia flagship. AMD and Nvidia are promising that mGPU is going to be the way to VR nirvana in the future. If the cards are around 150w in stock form you could run 4 in a system with a 850w power supply.

R.I.P. Corsair 1500w power supplies. :(
Click to expand...


In the premise of something will be usable in the future like mGPU is never a good way to market cards (or any product for that matter) at any price range
 
razor1 said:
In the premise of something will be usable in the future like mGPU is never a good way to market cards (or any product for that matter) at any price range
Click to expand...

Call me cynical, but I won't even consider multi-gpu until I see a steady stream of games that support it from day one... scaling issues aside, you have higher failure rate, higher power draw , motherboard requirements, case size requriements
 
Same here tried it once with the 7800gt and never looked at multi gpu since. Too many compatibility issues for the extra costs.
 
Ieldra said:
Call me cynical, but I won't even consider multi-gpu until I see a steady stream of games that support it from day one... scaling issues aside, you have higher failure rate, higher power draw , motherboard requirements, case size requriements
Click to expand...
That's not cynical, that's completely reasonable.

Cynical would be saying that AMD's push towards multi-GPU comes from the fact that they don't have a large die enthusiast GPU on offer. And I think that's a small part of it, but I believed Raja when he said small dies are the future. If AMD puts multiple GPUs in the console upgrades, and DX12 support works as promised, they might come out on top-- eventually.
 
schizo said:
That's not cynical, that's completely reasonable.

Cynical would be saying that AMD's push towards multi-GPU comes from the fact that they don't have a large die enthusiast GPU on offer. And I think that's a small part of it, but I believed Raja when he said small dies are the future. If AMD puts multiple GPUs in the console upgrades, and DX12 support works as promised, they might come out on top-- eventually.
Click to expand...

Rajas Talk about small dies are the future is pretty shallow and I'll explain why

Small dies are preferable because given fixed defects per wafer, you have average less defects per die with small dies.

However

Two 300mm dies stuck on an interposer will take up more space than one 600mm

Four 150mm dies will need a bigger interposer than two 300mm dies

Interposers are expensive
 
limitedaccess said:
Personally I do not agree with the seemingly revulsion (and by this I refer to both extremes with Polaris 10 being in the 390x range as opposed to higher. That performance target at lower price in my opinion is much better to a larger amount of parties then a hypothetical higher target and higher price.

Of course the variable in this is how efficient the cost structure actually is and how much of it is transferred to the consumer.

At the end I'd prefer AMD go back to their pre-GCN strategy versus post-GCN.
Click to expand...

The largest AMD base members are actually around r9 290/390 cards, they need to give a reason for upgrade to that people, they need their own people to spend money upgrading with truly performance reasons and not by only decreasing power consumption, someone who have already a system with a 290/390 or 290X/390X need a reason to upgrade, people with gtx 970/gtx980 need a reason to upgrade that's how you atract people to buy your products not only saying "hey I offer your performance at lower power requirement, come here and side grade your card to save 1$ monthly" nope just nope.

They can have whatever they performance/watt want but the main focus should be always increase performance over what we have now, just look at intel mainstream CPU. They keep selling as hot cakes not matter if the performance upgrade year to year is ~10% they keep decreasing power consumption and still increasing iGPU performance and CPU performance, even enough to increase the base clock up to 4ghz. So everyone happy.. but sacrifice performance for power is a no go way in my opinion, I would prefer if they offer a high performance/high powered GPUs to everyone and a lower performance/low powered version of the same card to OEM and so on, so let's say a heavy underclocked version card in core and memory can do the job offering good performance/watt..
 
schizo said:
That's not cynical, that's completely reasonable.

Cynical would be saying that AMD's push towards multi-GPU comes from the fact that they don't have a large die enthusiast GPU on offer. And I think that's a small part of it, but I believed Raja when he said small dies are the future. If AMD puts multiple GPUs in the console upgrades, and DX12 support works as promised, they might come out on top-- eventually.
Click to expand...

Basically DX12 is not the problem it has to be done in gaming engines as soon as that happens in Vulkan/DX12 or in Mantle when using LiquidVR then there is no more drawback in using multiple GPU as we know it ...
 
Pieter3dnow said:
Basically DX12 is not the problem it has to be done in gaming engines as soon as that happens in Vulkan/DX12 or in Mantle when using LiquidVR then there is no more drawback in using multiple GPU as we know it ...
Click to expand...

That's just VR though, so people with VR headsets should be happy, all five of them :p
 
You must log in or register to reply here.
Back
Top