Pretty sure all initial SKUs will be Samsung. Later launches (APUs) may be GF. I don't think there will be any HBM based SKUs for Polaris at TSMC.
Last edited:
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Polaris architecture – GCN 4.0
- Thread starter Boomstick777
- Start date
Pretty sure all initial SKUs will be Samsung. Later launches (APUs) may be GF. I don't think there will be any HBM based SKUs for Polaris at TSMC.
Yes it will be Samsung+Glofo, as those are the same process.
There were rumors about Tsmc manufacturing the low end cards, but as the one they showed is a small GPU and its not Tsmc this rumor is busted it seems.
It's not necessarily even a small GPU. Potentially it could be Greenland using FRTC and heavy power gating. Though I'm not sure such a large GPU could scale down quite that efficiently.
You can, it's called a 980ti. If you expect something similar in performance but cheaper, I'd say that's unlikely - maybe towards the end of the year.
Wow so AMD might be first to the punch for a change.
I do hope so
Even if NVIDIA manage to make it out ahead or around the same time (which I doubt), I think they'll have a fraction of the supply.
It wouldn't surprise me if AMD will now use Samsung HBM2 as well as SK Hynix, now that there's the fab agreement. Potentially they could take everything in house with Samsung - moving the interposers from UMC too - which would save a lot of costs.
Given that NVIDIA are in the process of patent trolling Samsung (and losing badly), I doubt Samsung would have many qualms about screwing NVIDIA over re: supply of HBM2, as they're a tiny customer anyway.
This may be why there's been all the rumblings about only the top Pascal SKU having HBM, and the rest using either GDDR5 or GDDR5x.
Wow so AMD might be first to the punch for a change.
I do hope so
Same here, fingers crossed
Hopefully it does. If the core was to output a constant compared to previous, 28nm cards, that's the equivalent of lot of pushing power on 14/16nm.
Not interested in power saving unless it translates into more powerful chips and better headroom tbh
"In any case, the GPU RTG showed off was a small GPU. And while Raja’s hand is hardly a scientifically accurate basis for size comparisons, if I had to guess I would wager it’s a bit smaller than RTG’s 28nm Cape Verde GPU or NVIDIA’s GK107 GPU, which is to say that it’s likely smaller than 120mm2. This is clearly meant to be RTG’s low-end GPU, and given the evolving state of FinFET yields, I wouldn’t be surprised if this was the very first GPU design they got back from Global Foundries as its size makes it comparable to current high-end FinFET-based SoCs. In that case, it could very well also be that it will be the first GPU we see in mid-2016, though that’s just supposition on my part."
By Anandtech.
Either way i look at it the difference is closer to 50%...nowhere near double.
I know 90% of your posts are trolling, but:
1) Anandtech are frequently wrong and if this chip is GF, it's a very big surprise. If it is GF it doesn't mean that production chips will be.
2) I seriously doubt any of the initial chips will be GF as it'd be a nightmare logistically. Their 14nmFF plant is in New York. For any volume product (as opposed to test chips) they'd have to ship the GPU from NY to UMC in Taiwan or Samsung in South Korea to have it mounted on the interposer, then have the interposer shipped to China or Malaysia to be mounted on the PCB and have the cooler mounted. Zen or APUs make more sense for GF, though Zen is also now expected to at least initially be Samsung too.
3) Greenland is definintely Samsung. It's their big chip. It would defy all logic if any of their big chips (or maybe '16 chips at all) were TSMC.
Last edited:
so why do people think AMD's stuff is coming first??
wasnt so long ago people were saying nvidia's is 1st maybe as early as feb/march?
im surprised we seeing tech demos so early but i dont think we should read too much into that
Well as i see it.
Samsung/Glofo is a much safer bet than TSMC.
Two factories with the same tech, twice the output, against tsmc one.
Samsung manufactures HBM as well, so they can use those in house...maybe move the interposer to samsung as well.
Samsung and NV are shooting lawsuits on each other, so hbm manufacturing is a weapon in samsungs hands against nvidia.
I think they will try to use the Samsung/Glofo same process to the quantity advantage vs tsmc/nvidia.
That would be the logical step...
Gpu - samsung (or glofo for some chips)
Hbm - samsung (+hynix for quantity)
Interposer/assembly - samsung
While nv can do the tsmc-hynix-umc rounds.
Last edited:
Soldato
- Joined
- 29 Aug 2010
- Posts
- 8,284
- Location
- Cornwall
Remember when the Nvidia 900 series came out and used less power and "everyone" was saying that nobody cares about power efficiency we just want a powerful card.
Now it seems AMD are releasing a power efficient card it seems like actually lots of people do suddenly care about this. Now a suggestion of not caring about power efficiency is met with silly suggestion of a 5KW card...
Makes me wonder if all the people saying that "we need AMD to stay in business to keep Nvidia honest" would suddenly have the opinion of "let Nvidia die we only need 1 vendor!" if it ever looked likely.
Within reason I'm not too bothered about power usage. I'd rather have a ~300W card that can do 120fps at 4K in BF5 (or whatever) than a 125W card that can do 55fps at 4K.
However if it's a case of a 375W card doing 165fps @ 4K or a 250W card doing 145fps @ 4K the 250W does make more sense. So basically if the extra power is justified by a relative boost in performance, then I'd rather have more performance at the cost of more power draw.
Hopefully the die-shrink and move to HBM2 will give decent power savings, but there's surely more saving to be had with the right architecture design.
Before we can really judge how well AMD have done in this regard we need to see how well Nvidia do. Might give us a sense of what gains you get from die-shrink and HBM2.
I just hope the new cards will smash 4K so we can look forward to 120/144Hz 4K screens, if that means they draw around 300W then so be it.
Now it seems AMD are releasing a power efficient card it seems like actually lots of people do suddenly care about this. Now a suggestion of not caring about power efficiency is met with silly suggestion of a 5KW card...
Makes me wonder if all the people saying that "we need AMD to stay in business to keep Nvidia honest" would suddenly have the opinion of "let Nvidia die we only need 1 vendor!" if it ever looked likely.
Within reason I'm not too bothered about power usage. I'd rather have a ~300W card that can do 120fps at 4K in BF5 (or whatever) than a 125W card that can do 55fps at 4K.
However if it's a case of a 375W card doing 165fps @ 4K or a 250W card doing 145fps @ 4K the 250W does make more sense. So basically if the extra power is justified by a relative boost in performance, then I'd rather have more performance at the cost of more power draw.
Hopefully the die-shrink and move to HBM2 will give decent power savings, but there's surely more saving to be had with the right architecture design.
Before we can really judge how well AMD have done in this regard we need to see how well Nvidia do. Might give us a sense of what gains you get from die-shrink and HBM2.
I just hope the new cards will smash 4K so we can look forward to 120/144Hz 4K screens, if that means they draw around 300W then so be it.