• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Possible Upcoming AMD Radeon GPUs Spotted: Polaris 10 XT2 and Polaris 12

Raevenlord

News Editor
Joined
Aug 12, 2016
Messages
3,755 (1.17/day)
Location
Portugal
System Name The Ryzening
Processor AMD Ryzen 9 5900X
Motherboard MSI X570 MAG TOMAHAWK
Cooling Lian Li Galahad 360mm AIO
Memory 32 GB G.Skill Trident Z F4-3733 (4x 8 GB)
Video Card(s) Gigabyte RTX 3070 Ti
Storage Boot: Transcend MTE220S 2TB, Kintson A2000 1TB, Seagate Firewolf Pro 14 TB
Display(s) Acer Nitro VG270UP (1440p 144 Hz IPS)
Case Lian Li O11DX Dynamic White
Audio Device(s) iFi Audio Zen DAC
Power Supply Seasonic Focus+ 750 W
Mouse Cooler Master Masterkeys Lite L
Keyboard Cooler Master Masterkeys Lite L
Software Windows 10 x64
when looking up at an hex file taken from macOS Sierra drivers, Anandtech forum user Glo found what could very well possibly amount to upcoming graphics chips from AMD: Polaris 10 XT2 and Polaris 12. We could very well be looking here into an as-of-yet unknown revision of AMD's Polaris 10 architecture, as well as a totally different chip from the already released Polaris 11 and Polaris 10. Maybe Polaris 12 is the mysterious 687F:C1 chip previously benchmarked in AOtS?

Also of note is the referral to Vega 10, lending further credence to reports of an early 2017 release. Given the fact that all three different architectures are referenced in the same hex dump, this could mean that AMD is working on a new 500 line of GPUs for 2017 - possibly to complete a given ZEN platform and giving customers the chance to go all-in on an AMD system, while simultaneously capitalizing on AMD's apparent confidence in ZEN's market reception. In this scenario, AMD's Vega10 would serve as the successor to the Fury series, with Polaris 12 and Polaris 10 XT2 replacing Polaris 11 and Polaris 10 from the product stack. Another scenario is that Polaris 10 XT2 corresponds to a dual-gpu solution, whose rumors have been making the rounds for some time now.



View at TechPowerUp Main Site
 
Last edited:
Yup! I really think this is what's coming:

RX 485 (2304SP @ 1500 MHz with GDDR5X/HBM2) = December for $350

RX 495x2 dual polaris (Hopefully on an interposer so crossfire always works) = December/January for $700

RX Fury X (4096 - 5120 SP with HBM2) = Late Winter or Spring $600
 
Some competition would be good.

Get on with it AMD.
 
Waiting patiently,.......................(taps foot)
 
Yup! I really think this is what's coming:

RX 485 (2304SP @ 1500 MHz with GDDR5X/HBM2) = December for $350

RX 495x2 dual polaris (Hopefully on an interposer so crossfire always works) = December/January for $700

RX Fury X (4096 - 5120 SP with HBM2) = Late Winter or Spring $600

Multi-GPU support from new games has been extremely limited. I highly doubt if AMD will attempt a dual-GPU card, especially with a mid-range chip like Ellesmere. Imagine you paid $500 for your graphics card, and in a game not optimized for multi-GPU (which is nearly every DX12/Vulkan game launched so far), it gives you the performance of a $250 card (barely playable at resolutions above 1080p).

I would want AMD to do something like this:
  • 4,096 "Vega" SP
  • 8 GB, 256-bit, 10 Gbps GDDR5X
  • 256TMU/64ROP
  • 2x ACEs of Polaris 10
  • $399 price
Something like that could disrupt NVIDIA's GP104-based lineup, and also give AMD the ability to wage a price-war against NVIDIA. You can't do something like that with an exotic HBM2/interposer/MCM approach.
 
Don't care about Polaris. I want Vega!
The industry desperately needs the competition.
I understand that AMD flipped their release schedule and got the mid to low end Polaris out first. Now we've been waiting for the higher end and AMD needs to deliver!
 
Last edited:
Yup! I really think this is what's coming:

RX 485 (2304SP @ 1500 MHz with GDDR5X/HBM2) = December for $350
There is no performance/economic reason to put HBM on a small gpu like that. The card would gain almost nothing but lose the main factor 480 has and that's price. GD5X is more likely.
The XT2 could very well be the same 480 gpu, but now that the process has matured, with higher clocks and lower power draw. And newer memory ofc.
 
Tired of the news form AMD... Polaris, Vega, Zen....only leaks, rumors, suppositions, etc, nothing concrete for the past 2-3 years....
BOOORING!
 
Don't care about Polaris. I want Vega!
The industry desperately needs the competition.
I understand that AMD flipped their release schedule and got the mid to low end Polaris out first. Now we've been waiting for the higher end and AMD needs to deliver!

I think any Polaris refreshes are important because those price points are where the majority of buyers live.
Usually with Nvidia, the high tiers get released first and the majority must wait for enough failed silicon to be produced to get the low tier products.
This time with AMD due to two separate architectures the under $300 was released first, forcing high-end enthusiasts to wait. Now you know how the majority feels.
 
Multi-GPU support from new games has been extremely limited. I highly doubt if AMD will attempt a dual-GPU card, especially with a mid-range chip like Ellesmere. Imagine you paid $500 for your graphics card, and in a game not optimized for multi-GPU (which is nearly every DX12/Vulkan game launched so far), it gives you the performance of a $250 card (barely playable at resolutions above 1080p).

I would want AMD to do something like this:
  • 4,096 "Vega" SP
  • 8 GB, 256-bit, 10 Gbps GDDR5X
  • 256TMU/64ROP
  • 2x ACEs of Polaris 10
  • $399 price
Something like that could disrupt NVIDIA's GP104-based lineup, and also give AMD the ability to wage a price-war against NVIDIA. You can't do something like that with an exotic HBM2/interposer/MCM approach.

How many ACE did Polaris have? Fiji had 8?

IMO, all Vega needs to have is a refined smaller process than Fiji (as it obviously does) which should allow far higher clocks. Fiji was a good but 'slow' chip. If it can simply maintain and refine the hardware but boost the clocks, it will be a great design. A 1500Mhz (is that too much a dream?) Vega would be 50% faster clocks than Fiji. Add in 10% from architecture refinement and you get get 60% better than Fiji by doing a Pascal. The benefit of AMD doing a Pascal means it still keeps the compute side intact. My arithmetic is simple but a far faster and refined Fiji is all Vega needs to be.
 
If there really will be improved 480 I will be pissed. Bought mine about month ago :D
 
Well next time don't just buy card 5 months after release date and don't keep it for that long either.
 
Well next time don't just buy card 5 months after release date and don't keep it for that long either.
Exactly. I bought 780 TI just for the time being to wait for the new cards to show up(not mentioning 670 which burned after 2 weeks). I though that improved Polaris will see the light and I only hope it will be worth purchasing. If the single mysterious gpu is POLARIS based chip but tweaked that would mean it is worth to wait for vega. But that depends only of the performance of that RX 490. I'd really want to see all the specs and stuff :)
 
Being this is from Mac I might think at least one of those number could be something like a M295X Mac Edition (Amethyst) chip that was for much of the time exclusive/proprietary.
 
Multi-GPU support from new games has been extremely limited. I highly doubt if AMD will attempt a dual-GPU card, especially with a mid-range chip like Ellesmere. Imagine you paid $500 for your graphics card, and in a game not optimized for multi-GPU (which is nearly every DX12/Vulkan game launched so far), it gives you the performance of a $250 card (barely playable at resolutions above 1080p).

I would want AMD to do something like this:
  • 4,096 "Vega" SP
  • 8 GB, 256-bit, 10 Gbps GDDR5X
  • 256TMU/64ROP
  • 2x ACEs of Polaris 10
  • $399 price
Something like that could disrupt NVIDIA's GP104-based lineup, and also give AMD the ability to wage a price-war against NVIDIA. You can't do something like that with an exotic HBM2/interposer/MCM approach.

ME..
I would want AMD to do something like this:
  • 4,096+ (and or not limited to 6144 ")Vega" SP
  • 8 - 16GB HBM2 2048-4096 Bit 512GBPS/ up to a prefered bandwidth of 1.0 TB/s
  • 256TMU/64ROP or better yet.... 256 TMU's / 128 ROP's... or VEGA XT 320 TMU's / 128+ ROP's
  • 4-6x ACEs of Polaris 10
  • $399-$699 price depending on TMU's/ROP's Amount's
Somthing that really would destroy/DOMINATE an Nvidia GTX 1080Ti or a Current Titan x/z .(a Titan X/Z that is two times of the current Titan X/Z..
 
Let's not buy into Nvidia blaming developers for not optimizing games for SLI excuse.
The real reason for the decline in multi-GPU support is the sharp increase in the GTX 900 and Pascal pricing.
You could buy a Titan X for less than two GTX 1080's cost!
Then at the lower end where SLI makes the most sense for consumers who want to upgrade in installments, they disabled SLI in the GTX 1060.
Thankfully, Radeon allows and even promotes Crossfire of it's RX 480s, which is a great setup for consumers on a budget.
For myself, I could not afford much more than a single RX 480, but I will be able to purchase a second card later after other expenses in my new build are met.
This gives me the ability to game on, while I'm waiting to be able to buy the second card to complete my build.
 
Back
Top