[TonyMacx86]Possible Polaris 12 GPU and Refreshed Polaris 10 spotted in macOS Sierra drivers.

[Glo.]

You guys are confusion technology with actual hardware ...

No, they can't reuse the GPU part of an APU for doing a desktop GPU.

Yes, they can reuse the technology to spin a new desktop GPU.

But by reusing the technology, they will need to re-optimize everything from scratch ... it's not like you can copy/paste from an APU. Certainly, some blocks could be reused 1:1 but not the entire GPU part.

Nobody says that is the case. More likely the same case as is with Polaris 10, and PS4Pro GPU. One have inspired another. There is no reason why would Project Scorpio GPU not reused in some form in desktop GPU, if there is need for it.

 

[Phynaz]

Zen APUs use Vega architecture.

Has this actually been confirmed by AMD? AMD has had a significant time lag between introducing a GPU architecture and having it show up in a CPU.

 

[Glo.]

Has this actually been confirmed by AMD? AMD has had a significant time lag between introducing a GPU architecture and having it show up in a CPU.

http://videocardz.com/62250/amd-vega10-and-vega11-gpus-spotted-in-opencl-driver

Stoney Ridge - same family as Ellesmere(Polaris).
Raven Ridge - Same family as Greenland/Vega.

 

[Phynaz]

http://videocardz.com/62250/amd-vega10-and-vega11-gpus-spotted-in-opencl-driver

Stoney Ridge - same family as Ellesmere(Polaris).
Raven Ridge - Same family as Greenland/Vega.

So "no" then, you don't have an AMD source.

Considering Polaris hasn't made it to a CPU yet, I don't expect Vega to be in a CPU this year. After all, it took two years from a GPU based upon GCN 3 to make it into a CPU.

I don't see how AMD qualifies the GPU and CPU that contains the GPU at the same time. As I said, it has always taken AMD at least a year to get a new GPU into a CPU. Remember, the graphics group is a different organization within AMD. After the graphics guys make a product, the the CPU guys get it for integration into a CPU. And that's a two year job.

 

[Piroko]

Based on how AMD has handled GCN in the past I'd say there are two logical slots where a third Polaris SKU would slot in:

The first slot is below Polaris 11, at ~90mm², ~1.8b transistors, 512 shaders and 64 Bit GDDR5 bus. This would be a cost effective replacement of Oland and Cape Verde and a good replacement for the 15W to 25W volume SKUs on mobile. It's also a market that hasn't seen Pascal chips yet, increasing the competitiveness of Polaris.

The second slot is where Bonaire slotted in in 2012, at 160 to 180mm², ~4.5b transistors and 1500 to 1800 shaders, coupled with a 128 Bit GDDR5(x) bus. This would slot into the 55 to 75W mobile segment. The rather low bandwith per shader would be fine for a chip that doesn't see clocks beyond 1100 MHz, further pushing the mobile orientation.

 

[Valantar]

Nobody says that is the case. More likely the same case as is with Polaris 10, and PS4Pro GPU. One have inspired another. There is no reason why would Project Scorpio GPU not reused in some form in desktop GPU, if there is need for it.

Actually, there is a reason. It's called Vega, and is supposed to be launching soon, with at least two chip sizes (although not at the same time), both larger than P10, along with improved perf/W. Making a larger Polaris chip makes very little sense in light of this.

 

[Glo.]

If you want to know what is the biggest argument for Polaris 12 to be bigger than Polaris 10: Smallest GPU AMD makes is already in MacBook Pro.

What compute could use even smaller than that? Mac Mini - Polaris 11, Mac Pro? Polaris 10. iMac: Polaris 10 and 11. What could they provide that would need that small and that low power GPU, if P12 would be smaller than P11?

And for one thing, Mac Pro - they do need better GPU than Polaris 10.

And best part. There are two kexts, with framebuffers. AMD4100 does not have any traces and relation to Mac Pro 7.1 trash can design. It is most likely related to iMac hardware. It contains Ellesmere Pro and XT deviceID's. But no Polaris 10XT2. It also contains full personality for Polaris 11 used in MBP.

AMD9500 on the other hand can be related to Mac Pro trash can personality. It has specific design(6 DP) output. It contains the GPUs in question: Vega 10, Polaris 12, and Polaris 10XT2.

 

[Blitzvogel]

Based on how AMD has handled GCN in the past I'd say there are two logical slots where a third Polaris SKU would slot in:
The first slot is below Polaris 11, at ~90mm², ~1.8b transistors, 512 shaders and 64 Bit GDDR5 bus. This would be a cost effective replacement of Oland and Cape Verde and a good replacement for the 15W to 25W volume SKUs on mobile. It's also a market that hasn't seen Pascal chips yet, increasing the competitiveness of Polaris.

Polaris 11 already fills this void on the MacBook Pro as the Radeon Pro 450 as it has 640 SPs. The less desirably performing silicon from the foundry could be sold as 640 and 512 SP + 128 bit GDDR5 or DDR3 SKUs for the low end. "Radeon R450" and "Radeon R440" seem like appropriate designations. However, I wouldn't be surprised if AMD is afraid of such SKUs cannibalizing RX460 sales.

I'm wondering if the full 1024 SP Polaris 11 SKU will eventually make an appearance in desktop SKUs. The best silicon is obviously being reserved for Apple, with the secondary going to notebooks, but the rest would be fine for a "Radeon RX 465" or "Radeon RX 460XT" edition. There is a big void between the RX460 and RX 470 currently. While Polaris 11 couldn't hope to fight the GTX 1050Ti, it can fight the basic 1050.

 

[Headfoot]

I just find it very unlikely there is going to be another die size when we're on the precipice of 2 more dies under the Vega family. It just doesn't make sense. I'd imagine it's a revision that ends up being 465 or 485.

 

[Keysplayr]

So its as I thought. Polaris 12 is bigger than polaris 10 simply because people want it to be. Im not saying it isnt , im saying that the "evidence that as been presented is pretty weak at best and reeks ofwishful thinking. There have been excuses of chronological order, but why only now? And reasoning that the smallest gpu amd makes is already in a macbook?
Things for mac are their own animal and I wouldn't bet on that at all.

 

[dacostafilipe]

So its as I thought. Polaris 12 is bigger than polaris 10 simply because people want it to be. Im not saying it isnt , im saying that the "evidence that as been presented is pretty weak at best and reeks ofwishful thinking.

There's no evidence in this thread, only speculation ...

And reasoning that the smallest gpu amd makes is already in a macbook?
Things for mac are their own animal and I wouldn't bet on that at all.

Think about that.

Polaris 11 was specially crafted for Apple ( die height reduction?).

Is it that hard to think that AMD could create a new die for the future iMac and Ma Pro? Personally, I think it's possible.

Edit: Just to say there are two "new" chips. P12 and P10XT2. Impossible that both are "only" a refresh.

 

[Gigamaxx]

http://videocardz.com/62250/amd-vega10-and-vega11-gpus-spotted-in-opencl-driver

Stoney Ridge - same family as Ellesmere(Polaris).
Raven Ridge - Same family as Greenland/Vega.

I'm seeing lots of references to Raven in the 9500 controller kext.

 

[Valantar]

Polaris 11 was specially crafted for Apple mobile ( die height reduction?).

There, I fixed it for you.

Once production volumes pick up, we'll no doubt see Polaris 11 in a whole host of more or less thin-and-light laptops (such as the Asus UX series that have used GTX 940s and the like).

 

[Gigamaxx]

My wild guess, Polaris 12 is for a new iPad Phablet 12" 4K.

 

[Valantar]

I just find it very unlikely there is going to be another die size when we're on the precipice of 2 more dies under the Vega family. It just doesn't make sense. I'd imagine it's a revision that ends up being 465 or 485.

Exactly. The only possible new Polaris die size I can imagine is an in-between one (~1800SMs), to allow for a cheaper SKU to blow the GTX 1050Ti out of the water/fit comfortably in an iMac's limited thermal dissipation. A low-end one (~600SM) might possibly make sense if it came with a deal to sell a ton of them to various laptop makers. But in general, not. A bigger one? In light of Vega being so, so close to launch, and from all AMD has said being a better/more power efficient architecture? Not a chance. It wouldn't be worth developing an expensive chip that would probably be more power hungry than it's bigger brother - it would be laughed off the stage before it even got started. If AMD needs an in-between card between the 480 and 490, cut-down Vega is the only thing that makes sense.

 

[Glo.]

My wild guess, Polaris 12 is for a new iPad Phablet 12" 4K.

I have to disagree on this completely .

Hello, welcome to the forum. Its actually you who have posted the information about Vega 10, Polaris 12, and Polaris 10XT2 in TonyMacx86 forums .

About the Raven Ridge traces: Bits and Chips claims that Apple will use Zen based APUs in MacBook Pro and MacBook. So that might actually validate their claims.

 

[Gigamaxx]

I have to disagree on this completely .

Hello, welcome to the forum. Its actually you who have posted the information about Vega 10, Polaris 12, and Polaris 10XT2 in TonyMacx86 forums .

About the Raven Ridge traces: Bits and Chips claims that Apple will use Zen based APUs in MacBook Pro and MacBook. So that might actually validate their claims.

It was a wild guess, how about a 15" Phablet? The new iPhab!

I also see Falcon along with Raven, does that ring any bells?

 

[RussianSensation]

Exactly. The only possible new Polaris die size I can imagine is an in-between one (~1800SMs), to allow for a cheaper SKU to blow the GTX 1050Ti out of the water/fit comfortably in an iMac's limited thermal dissipation. A low-end one (~600SM) might possibly make sense if it came with a deal to sell a ton of them to various laptop makers. But in general, not. A bigger one? In light of Vega being so, so close to launch, and from all AMD has said being a better/more power efficient architecture? Not a chance. It wouldn't be worth developing an expensive chip that would probably be more power hungry than it's bigger brother - it would be laughed off the stage before it even got started. If AMD needs an in-between card between the 480 and 490, cut-down Vega is the only thing that makes sense.

"Turns out, as Ars Technica explains, Apple had to take this decision to ensure that all variants of the 15-inch MacBook Pro are able to drive two 5K monitors at the same time. The dedicated intel GPU on the Skylake chips only supports up to three displaystreams, with one of them being used by the internal Retina Display. Powering a 5K display would require two display streams since the DisplayPort v1.2 spec does not offer enough bandwidth to run such a high-resolution display. Apple gets around this shortcoming by merging two DisplayPort 1.2 streams into one. This essentially means that the Intel HD GPU inside the quad-core Skylake chips on the new 15-inch MacBook Pro could not have powered more than one 5K display at once. In comparison, the AMD GPU used by Apple offers six display streams, which allows them to support up to two 5K monitors simultaneously. With one display stream being used for the internal display and four for powering two 5K monitors, the sixth and left over stream could still be possibly used to power an additional 4K monitor."
http://www.iphonehacks.com/2016/11/...olaris-gpu-15-inch-macbook-pro-touch-bar.html

The only way NV is getting inside the iMac is if they improve their multi-monitor support and/or if they give Apple prices AMD cannot match. AMD dGPUs are a perfect match for Apple who continue to squeeze suppliers while raising prices on its computers. Apple also likes to support OpenCL over CUDA. It is true that Apple likes to give contract wins to AMD or NV to get the best pricing but it seems.

 

[happy medium]

I'm guessing these are Polaris 12 gpu's

http://www.kitguru.net/components/g...mobile-gpus-for-macbooks-arent-that-powerful/

 

[kraatus77]

Those are p11. not p12.

 

[HurleyBird]

Honestly, it's hard to make heads or tails of this. It's probably not a refresh of P11 because they also confirm a refresh of P10 that's named differently, and a new Polaris chip so close to the Vega launch doesn't really make sense in the first place. There isn't much room at all between Polaris 10 and 11, and going above or below the current stack doesn't make that much sense either. I can't think of anywhere obvious for this chip to go, but I can think of a few improbable possibilities (some of which may be combined), but keep in mind that my confidence in any of these being correct is certainly not high enough to put any kind of wager on:

1. A really, really small chip, below P11, to address the "lets take advantage of the technologically impaired" market. Incredibly hard to see what Apple would want with that though.
2. A mobile focused GPU with an HBM memory controller to address the ultra book market - which somehow slipped and wasn't ready for the new Macbook Pro, and two of which may be used in the Mac Pro.
3. A smaller GPU based on Vega's new graphics IP version but without the HBM, because AMD decides for whatever reason to separate the Polaris and Vega code-names based on the presence of an HBM memory controller instead of graphics IP.
4. A different, most likely older code name for Vega 10 that slipped through the cracks. This would hint at very little architectural difference between Polaris and Vega if true.
5. The above, but a typo instead of an older code name.
6. Some kind of process pipe cleaner and/or insurance policy. Either from a different foundry, or on a different node.
7. A bigger GPU that slots between P10 and small Vega. Perhaps in part designed as a backup plan should the Vega GPU family run into serious delays. This only makes sense if Vega 10 and 11 are way bigger than we think they are so this in-between-er has room to coexist. This is the most awesome possibility, but not the most likely.
8. A Polaris chip with tangible DP performance to fill the HPC gap until a DP oriented Vega chip comes out. Or maybe some other boutique HPC oriented thing. Not sure why Apple would want it, but I guess it could kind-ish make sense in the Mac Pro. Maybe? I guess.

I suppose time will tell. I think the chances of it being one of the more interesting possibilities is higher than it normally would be just because the more "normal" options make so little sense.

 

[AtenRa]

So "no" then, you don't have an AMD source.

Considering Polaris hasn't made it to a CPU yet, I don't expect Vega to be in a CPU this year. After all, it took two years from a GPU based upon GCN 3 to make it into a CPU.

I don't see how AMD qualifies the GPU and CPU that contains the GPU at the same time. As I said, it has always taken AMD at least a year to get a new GPU into a CPU. Remember, the graphics group is a different organization within AMD. After the graphics guys make a product, the the CPU guys get it for integration into a CPU. And that's a two year job.

GCN 1.2 Hawaii (R9 290X) was launched in October 2013,
Kaveri APU launched in January 2014, only two months after Hawaii.

 

[Glo.]

It was a wild guess, how about a 15" Phablet? The new iPhab!

I also see Falcon along with Raven, does that ring any bells?

Falcon Ridge is previous name for Thunderbolt 2 controller, from Intel.

 

[Shivansps]

If P11<P10, why whould P12>P10? makes no sence.

 

[Det0x]

If P11<P10, why whould P12>P10? makes no sence.

Read previous page