AMD Launching Polaris 10 400 Series GPU June 1st At Computex. $299 (rumor)

Except you forget AMD's Crossfire setup is usually pretty damn poorly supported in drivers minus a couple games....
I don't play many games but the ones I do usually are well supported. The ones I do which are not are old and don't need that power. So not a problem, - as I said 'AAA gaming' but should've been more specific.

Also seems they are really pushing their driver support again especially on the Furys and 3xx most of the time for the main titles. But compared to single card, both companies drivers suck. Too bad its the only way to really do 4k justice, no matter what card you choose..
 
More from the higher speed architecture vs low speed higher efficiency POV. Not all P4s were hot beasts. Prescots were though, sure.. Vega will very likely take a big dump over 1080, if the Polaris leaks are accurate.

Polaris also should be either right on the heels or in the same ballpark as 1070 at this rate. So yes, they will compete in the mid range.
The 1080 is really a high midrange card and will be cemented there in a few more months, like the 980 was. Sure it's technically a high end card in the market strata now, but not by design.

Besides, flagships barely matter to the average joe. Less than 1% of people own a Ti on the vaunted steam survey, so who really gives a shit other than us?
 
Besides, flagships barely matter to the average joe. Less than 1% of people own a Ti on the vaunted steam survey, so who really gives a shit other than us?

Two words: Halo Effect
 
AMD doesn't have the capital left to spend on a halo product. They've got to win it back in the mid-range.
 
Just cuz hbm2 got delayed doesn't mean they don't have a halo product. Polaris will tell you what to expect from vega. If 2305 shaders perform closer to fury/furyx. You got one beast of a chip in vega.
 
AMD doesn't have the capital left to spend on a halo product. They've got to win it back in the mid-range.

I do not think it has anything to do with financials, but rather fabrication booking capacity. You can choose to start to manufacture a very complicated chip with a higher reject rate right away, or a smaller chip with a lower reject rate, have the possibility to learn optimization techniques, and prep for your big chips. nvidia cost the former as they usually do, AMD chose the latter this time round. Either way there is only book so much fab time available, especially at the smallest nodes.
 
Its poor planning and thinking HBM2 would be ready in time form Hynix. I'm sure there must be a clause in the contract with Hynix's exclusivity for HBM type ram where AMD must procure a certain amount of chips from them before they can get it from another vendor. Hynix has to cover its own ass too just like AMD covered its ass with the exclusivity portion of the contract.
 
Some quotes from the analyst meeting from Tueday. It seams Polaris will be performance and mainstream as I highlighted below.


So if you imagine, Polaris will ramp over a series of quarters, as we nail that performance and mainstream portion of the market, then you follow it on, and buy a new family in terms of the ultra-enthusiast portion of the market.

...

But as we think about our business model and the things we want to achieve in the near-term for this product launch to go to mainstream first, and then proliferate product upwards and downwards, up into enthusiast and down into value, is the strategy we're pursuing, because we believe that's the right thing for our business model and the right thing in terms of the overarching goal of bringing more performance per dollar that's being offered to the GPU space on the customers that we have there is something that's important to us.
 
Complete bullshit. They're doing it because HBM2 isn't ready yet.

That doesn't mean it won't prove a successful strategy in terms of growing revenue. But it's ludicrous to think they would purposefully concede the high-end and enthusiast markets to Nvidia.
 
Some quotes from the analyst meeting from Tueday. It seams Polaris will be performance and mainstream as I highlighted below.


That is not what he is saying,

But as we think about our business model and the things we want to achieve in the near-term for this product launch to go to mainstream first, and then proliferate product upwards and downwards, up into enthusiast and down into value, is the strategy we're pursuing, because we believe that's the right thing for our business model and the right thing in terms of the overarching goal of bringing more performance per dollar that's being offered to the GPU space on the customers that we have there is something that's important to us. And we think we bring a really strong value proposition with this family.

This is why you don't take things out of context there are many other things that were stated, I don't expect much change from architecture outside of HBM from Polaris to Vega *yes more cores and what not but that's it, Vega will still be GCN 1.4.
 
Last edited:
Heh. The Macau event is called "GCN 4.0". Seems GCNs 2 and 3 went the way of Windows 9.

4th generation GCN.

GCN 1.0. 1.1, and 1.2 were names that tech websites (specifically AnandTech I think) made up and were never internal names from AMD.
 
That is not what he is saying,



This is why you don't take things out of context there are many other things that were stated, I don't expect much change from architecture outside of HBM from Polaris to Vega *yes more cores and what not but that's it, Vega will still be GCN 1.4.

It's exactly what he's saying. After Polaris which is mainstream and performance, comes Vega which is ultra-enthusiast.

Polaris and Vega are two different architectures. Though I am not sure what this has to do with what we are talking about.
 
Last edited:
Architecture is probably the same GCN 4.0 for its internal guts. Minor tweaks for it to use HBM.

Polaris will not fill out the performance nor the enthusiast segments.

There are no other Polaris based chips outside of P10 and P11
 
Some quotes from the analyst meeting from Tueday. It seams Polaris will be performance and mainstream as I highlighted below.

I'm going to say that I don't agree with their stance. I think NV is doing it correctly. Here's why:

NV is launching the 1080/1070 first. This allows them to have a card faster than any prior card (the 1080), and a card that roughly matches the prior halo products (1070 matching 980ti/Titan X) at a lower price. They're saying, "You can get better performance than before, or last year's best for mid-range pricing." This is the precursor to launching the 1060/1050 which fills out and bumps up the quality in the lower-mid range segments. AMD's strategy is essentially taking last year's second tier (390/390x, not Fury lineup), and potentially offering it with a lower price and power consumption. There's nothing "new" here.

That said, the strategy can work. If they truly want to hit mainstream, they need to have GTX 970/R9 390 performance at $199. If the R9 480 is > R9 390, then $229-$249 can work, but not $299. An R9 480x at $299 would be acceptable if the 480 offers at least 390 performance at the lower price point.

Another thing here is about the comments from before about the 480 lineup potentially replacing the x90 lineup. That's possible. The R9 290/290x was the top dog for AMD before it was moved down market as the 390/390x to make room for the Fury lineup at the top. They could be moving the 480 back up a notch, with Vega being the 490/490x. But if this were to happen, the 480 lineup SHOULD be targeting Fury performance. Last year's halo should always be this year's 2nd best (IE, the 1070). I don't think AMD is going to pull THAT off.

But as I've said before, it is fun to speculate and I am not competent enough in this segment to truly have a clue with regards to what will actually happen.
 
Well AMD took a gamble with HBM 2 I think that's why this is happening, and they are now trying to make the best they have work for them. Which it might actually work if there is enough lead time before gp 106 hits, if OEM's get clamped down and they have their orders in, they might not be able to procure as many GP106's since they have to be concerned about their inventory numbers too specially when Back to school is so close. They need to stock up now or actually I think next month to get ready for it. Of course this isn't the first time a midrange launch vs a higher level card launch has been there, but this is a much different time. And the advantages are less too (power/watt)
 
Architecture is probably the same GCN 4.0 for its internal guts. Minor tweaks for it to use HBM.

Polaris will not fill out the performance nor the enthusiast segments.

There are no other Polaris based chips outside of P10 and P11

Yeah everyone thinks Polaris is an architecture now :'( GCN is an architecture, and Polaris and Vega represent the third major revision; hence GCN 3.0, or 1.3, or 3rd Gen GCN
 
Polaris, Arctic Islands, Ellesmere.
I don't know which is which anymore, but I do know those last 2 names will confuse a lot of people.

Polaris is actually "4th generation GCN". Direct from AMD.
 
P10 and ellesmere are the same, they change the chip names internally. So every sku that is P10 is also ellesmere its the Chip name. Its not an architecture.

Raja stated this a while back, they didn't have an "architecture" name like nV had so they added that in, but Polaris to Vega, its the same damn thing, all of them use GCN 4.0, just a few different features mainly the memory.
 
P10 and ellesmere are the same, they change the chip names internally. So every sku that is P10 is also ellesmere its the Chip name. Its not an architecture.

Raja stated this a while back, they didn't have an "architecture" name like nV had so they added that in, but Polaris to Vega, its the same damn thing, all of the use GCN 4.0, just a few different features mainly the memory.

I don't know why they don't name their architectures honestly.

Polaris sounds much better than GCN 4.0
 
Back
Top