Or the $300 Polaris if they want to do one. We don't know. They said that Polaris would range from $100 - $300.
490 could be Polaris and Vega could use a Fury like monkier, or 490/x could be Vega.
- Status
Or the $300 Polaris if they want to do one. We don't know. They said that Polaris would range from $100 - $300.
490 could be Polaris and Vega could use a Fury like monkier, or 490/x could be Vega.
I think both of you are correct.
The problem with the misunderstanding is that Vega is going to demolish Nvidia this year and next. Some here are stuck on thinking Polaris is the only design being released by AMD and forgetting Vega, the king killer is coming.
I think AMD should be more forthcoming about Vega. Is it a different design than Polaris or just slightly update or what?
Pretty sure they didn't announce Vega until a few months ago, not Jan.
Vega is going to probably be very similar to Polaris, just with HBM2 and bigger die.
http://wccftech.com/amd-vega-gpu-navi-gpu-hbm2-2017-2018/
swilli89
Golden Member
- Mar 23, 2010
- 1,558
- 1,181
- 136
Man, an 80CU, 5120sp vega with HBM 2 + DX12 boost would be unstoppable. I'm hoping AMD doubles up on the full Polaris 10 die (40CU 2560sp).
When I think about it, this actually looks like AMD's strategy..
Polaris 11 20CU 1280SP GDDR5 RX470
Polaris 10 40CU 2560SP GDDR5 RX480
Vega 1x* 60CU 3840SP GDDR5X or HBM2 RX490
Vega 1x* 80CU 5120SP HBM2 Rage PRO??
*not sure which Vega will come first, 10 or 11.
My understanding of all the remarks made by Koduri over the last few months point to a fundamental change in AMD's strategy to GPU development. A much faster change in families and technologies reminiscent of the golden days of GPUs. To those who say AMD can't afford it, my reply is you also said that they couldn't afford Polaris.
The small die strategy as proposed allows the bypassing of Moore's law on GPU design. Since last year, several have speculated about interposers and Multi-GPU. These and other new technologies bypass traditional GPU performance barriers.
Moore's law in action allowed the easiest path to rapid performance increases to be node shrinks. Now, we need other methods and I believe AMD is in the fore here. You have passive and active interposers. 2.5D and 3D stacking. GPU area able to move past traditional 600mm^2 maximum dies. The increase in area with the much lower cost of small dies allow more area expensive power saving techniques to be effective.
Remember the Koduri stated goal of RTG. Him being an engineer and even a bit nerdy, I believe he is very serious and not talking PR.
Koduri quote:
Koduri: We’re just entering the VR era. You see all these VR headsets. If you see what it takes to drive a VR headset, the pixel rate requirements are almost doubled up compared to the previous iteration. That’s driving up demand for discrete graphics quite a bit. Now, if you push that forward, when you get to 16K by 16K resolution, 120Hz, you get to a pixel rate of 6 billion pixels per second. We’re not going to get there if we just rely on Moore’s Law. We have to do disruptive things to get there. That’s the goal of our division, to get to the immersive era. We need to double up our products and technology, step by step. You’ll see the key initiatives and technology this year. You’ll see more next year.
Last edited:
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
It could also be that AMD'S vega was meant to counter 1070/1080 since Polaris seems it was meant to take on 1060/1060Ti.
If Vega is meant to take on big Pascal, what will take on 1070/1080?
Just as likely as Polaris has more shaders than advertised. No?
swilli89
Golden Member
- Mar 23, 2010
- 1,558
- 1,181
- 136
Think about it, the only way there isn't a 40CU Polaris 10 is if AMD is going to charge $100 more for 4gb more ram. We know for a fact that:
- There is a 4gb Polaris 10 @ $199
- That the CEO of AMD specifically said Polaris would be $199-$300
There is no way in hell AMD is going to charge 50% for just 4GB of dirt cheap GDDR5 ram. Based on this very strong assumption, its probable that we are looking at $200 for the 4gb cutdown P10, $249 for an 8gb cutdown P10, and $299 for the fully unlocked 2560sp P10.
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
Your speculation is sound. Another 256 shaders isnt much though on todays shader counts
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
I'm fairly certain Vega will eat Nvidia employees' babies.
And their puppies. No doubt.
Nah. Polaris is for the babies.I'm fairly certain Vega will eat Nvidia employees' babies.
crisium
Platinum Member
- Aug 19, 2001
- 2,643
- 615
- 136
Or $199 2048SP 4GB. Room for a 2304SP 4GB and 2304SP 8GB above, and it doesn't even have to be all the way at $300 since this was more of a general range pricing. There was never $299 MSRP stated.
Which makes more sense:
1. Release a small die with only fused off cores because the time isn't right to sell a higher priced product and make more money.
OR
2. Release a small die with all of it's cores enabled because the competition has already beat them to the punch and are currently walking all over them in performance and it doesn't make any sense that yields would be bad on a ~120w 230mm2 chip.
OR
3. Finally release that full 384-bit Tonga which would have probably been faster than Fiji Nano in CFX and would have ripped a hole in the space time continuum.
Last edited:
3 sounds good
I really wouldn't assume that the yields are terribly great though. Not with this being the first chip of any real size under this process, and it having been transplanted to GF while they were at it into the bargain.
They might well also want the best dies to get a mobile SKU out of this chip.
I really wouldn't assume that the yields are terribly great though. Not with this being the first chip of any real size under this process, and it having been transplanted to GF while they were at it into the bargain.
They might well also want the best dies to get a mobile SKU out of this chip.
I doubt we will see Vega priced above $700. We will probably see small Vega (10) @ 350-450 and large 450-600.
If Nvidia wants to price their cards >$700 let them, the market is tiny and the people buying them already have their minds set on Nvidia being better so why spend the money and resources to try to woo them?
That would be very different indeed to what they did with Fury.
It will of course depend which of the chips they can get up to the 1080/even bigger pascal.
The Vegas that end up lining up vs the 1070/80 will need to undercut their current pricing structure due to coming in rather after their launch, but if they get the performance of one up near the 1080ti/Titan etc it won't be cheap.
crisium
Platinum Member
- Aug 19, 2001
- 2,643
- 615
- 136
We already discussed the important of the halo effect. As bad as it is for consumers, I think if AMD ever gets a card out faster than Nvidia they need to price accordingly. Even Fury X matched 980 Ti price because it technically matched the stock reference performance at 4K and often 1440p. Obviously if they cannot match they will price lower, as is tradition.
Polaris being cheap seems like a return to classic AMD pricing - but not really. It is the successor to the 380 and 7800 series that launched around the same price. It is the competitor to the future 960 successor that launched around the same price. Although since Nvidia raised their prices by $50 ($70-$100 at launch) this time around I guess it is refreshing to see the same prices.
We'll see.
Anyone here knows the specs of the full Polaris line members? P10 and P11.
Clocks, power consumption, shaders, etc?
We know a bit about 1 member, the RX480 card. I think that there are definitely surprises in store. What are they????
Also, what is the lead time for a GPU design? A couple years? Unless there is some naturally ordained parting of performance segments that all follow, I think the GPU companies segment the market as they see fit. Some here and elsewhere say that Polaris is meant to counter 1060/1060Ti or Vega is for the 1070/1080 segment. To believe this is to have a rigid view. Each company developed designs in the manner that they saw most profitable within the philosophy each has. Any competition is from the natural overlap of price, performance and other characteristics.
No one could know very long in advance what the performance of the 1060, 1080, P10 and P11 models were. Certainly not long enough to design a specific competitor. Yes, you can vary certain parameters, but that is all. The core design is pretty well fixed way before any inkling of what the competitor is doing. A perfect example is async compute. Available for years and gradually being approached with all GPUs.
All of this is to show that we only know the most basic stuff about Polaris [deliberately by AMD] and should leave room for surprises beyond the $199 cost.
Clocks, power consumption, shaders, etc?
We know a bit about 1 member, the RX480 card. I think that there are definitely surprises in store. What are they????
Also, what is the lead time for a GPU design? A couple years? Unless there is some naturally ordained parting of performance segments that all follow, I think the GPU companies segment the market as they see fit. Some here and elsewhere say that Polaris is meant to counter 1060/1060Ti or Vega is for the 1070/1080 segment. To believe this is to have a rigid view. Each company developed designs in the manner that they saw most profitable within the philosophy each has. Any competition is from the natural overlap of price, performance and other characteristics.
No one could know very long in advance what the performance of the 1060, 1080, P10 and P11 models were. Certainly not long enough to design a specific competitor. Yes, you can vary certain parameters, but that is all. The core design is pretty well fixed way before any inkling of what the competitor is doing. A perfect example is async compute. Available for years and gradually being approached with all GPUs.
All of this is to show that we only know the most basic stuff about Polaris [deliberately by AMD] and should leave room for surprises beyond the $199 cost.
Understood.
VERY likely that the 480 should be the better choice for 1080p. Very few cases where a R9 390 would be considered inadequate for 1080p, and 480 will almost certainly be faster than the 390. So yes, I'd think it would be hard to justify the 1070 at that resolution. I'm not even considering a 1070 at my 1920x1200 resolution. I think it will end up complete overkill, especially considering the rather massive price difference.
Vesku
Diamond Member
- Aug 25, 2005
- 3,743
- 28
- 86
Another possibility, 8GB 480 with AIO and pre OCed to near max stable clocks. XFX rep on Reddit mused about the appeal of such a card albeit at $400. Not that interesting at $400 but might have a niche at $299.
sze5003
Lifer
- Aug 18, 2012
- 14,320
- 683
- 126
I was planning on getting a 1440p monitor but I also want to upgrade my ram, board, and CPU. Going with an aftermarket 480 will leave me money to do so.
I've spent $500+ before on a gpu and have no issue doing that but seeing how I've just stayed at 1080p, I could have saved myself money.
One thing I do worry about is if I'll be able to crank max settings in my games with the 480 vs the 1070. I've been kind of used to doing that with my 7970ghz now and just not being able to do it in the future would kind of suck.
If the 1070 prices change which I doubt they will, I would be able to justify getting one. Or say if i sold my current card for close to $100.
Now all we need is the benchmarks, deciding the amount of ram will be made when I go to purchase.
- Status
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter