Friday, May 13th 2016
More Polaris10 and Polaris11 Specifications Revealed
Industry sources revealed to TechPowerUp some pretty interesting specifications of AMD's two upcoming GPUs based on the 4th generation Graphics CoreNext "Polaris" architecture. The company is preparing a performance-segment GPU and a mainstream one. It turns out, that the performance-segment chip, which the press has been referring to as "Ellesmere," could feature 32 compute units (CUs), and not the previously thought 40.
Assuming that each CU continues to consist of 64 stream processors (SP), you're looking at an SP count of 2,048. What's more, this chip is said to offer a single-precision floating point performance of 5.5 TFLOP/s, as claimed by AMD. To put this into perspective, the company had claimed 5.2 TFLOP/s for the "Hawaii"/"Grenada" based FirePro W9100, which launched earlier this February, and that SKU featured all 2,816 SP present on the chip. So this chip is definitely faster than most "Hawaii" based SKUs.While "Hawaii" based SKUs feature TDP of no less than 250W, the new chip has a TDP rated no higher than 150W. AMD could pull off a "single 8-pin power connector" feat like NVIDIA, with quite some headroom to spare. The chip features a 256-bit wide GDDR5/GDDR5X memory interface, and 8 GB could be its standard memory amount. The first SKUs based on this chip could feature 7 Gbps GDDR5 memory.
AMD will upgrade the feature-set to include HVEC/H.265 hardware encode/decode acceleration, DisplayPort 1.3, and HDMI 2.0a outputs.
The smaller "Polaris" chip scheduled for 2016, which the press has been referring to as "Baffin," could feature 14 compute units, working out to a stream processor count of 896. It will be a mainstream chip, succeeding the "Tobago" silicon, which drives the current R7 360 series SKUs, although it wouldn't surprise us if it outperformed bigger chips, such as the "Trinidad" based R7 370 series. This chip has its peak single-precision floating-point performance rated at 2.5 TFLOP/s. Its TDP is rated at just 50W, and it is expected to feature a 128-bit wide GDDR5 memory interface, holding 4 GB of memory.
Assuming that each CU continues to consist of 64 stream processors (SP), you're looking at an SP count of 2,048. What's more, this chip is said to offer a single-precision floating point performance of 5.5 TFLOP/s, as claimed by AMD. To put this into perspective, the company had claimed 5.2 TFLOP/s for the "Hawaii"/"Grenada" based FirePro W9100, which launched earlier this February, and that SKU featured all 2,816 SP present on the chip. So this chip is definitely faster than most "Hawaii" based SKUs.While "Hawaii" based SKUs feature TDP of no less than 250W, the new chip has a TDP rated no higher than 150W. AMD could pull off a "single 8-pin power connector" feat like NVIDIA, with quite some headroom to spare. The chip features a 256-bit wide GDDR5/GDDR5X memory interface, and 8 GB could be its standard memory amount. The first SKUs based on this chip could feature 7 Gbps GDDR5 memory.
AMD will upgrade the feature-set to include HVEC/H.265 hardware encode/decode acceleration, DisplayPort 1.3, and HDMI 2.0a outputs.
The smaller "Polaris" chip scheduled for 2016, which the press has been referring to as "Baffin," could feature 14 compute units, working out to a stream processor count of 896. It will be a mainstream chip, succeeding the "Tobago" silicon, which drives the current R7 360 series SKUs, although it wouldn't surprise us if it outperformed bigger chips, such as the "Trinidad" based R7 370 series. This chip has its peak single-precision floating-point performance rated at 2.5 TFLOP/s. Its TDP is rated at just 50W, and it is expected to feature a 128-bit wide GDDR5 memory interface, holding 4 GB of memory.
133 Comments on More Polaris10 and Polaris11 Specifications Revealed
As for the Polaris 10 cards, we know that these cards will be at least as fast as the 390/390X and will cost a lot less than Nvidia's offerings. Both the 480 and the 480X will most likely outperform the 980 TI/Titan X in existing/upcoming DX12 games since the 390X already does that. All these factors make the 480/480X cards way more appealing to your average PC gamer, especially to those of us who aren't in the US.
What "here and now" are you talking about?
Around Xmas it will likely be cheaper as they have been out for a while and the market settled after both releases.
The 480X (40 of 40 CU activated) may then follow, have 2560 shaders and be nearly as fast as 980 Ti or even faster, for about 300-400$, effectively competing against GTX 1070.
Also I think the "Baffin" chip with 896 shaders is a non-X version with deactivated CU's too, so the full version may have 1024 or 1280 shaders and be a lot faster.
The GDDR5 @ 1750 MHz on a 256 bit bus should btw be enough to drive the 480/X, I expect revamped delta compression that is comparable to that of Maxwell or better, effectively increasing bandwidth per bit of GCN 4 cards a lot.
GDDR5X at 10 GHz and 256 bit = GDDR5 at 5Ghz and 512 bit (290x ram bandwidth).
two 290x plays everything i have trown at it at 2560 x 1600, one will do the same at 1080p, if these rumors are to be trusted the cards based on polaris 10 will outperform a 290x, at 150W they just need to price it right and you have the new price/performance king. As a fun side note, the 295x2 outperforms anything from nvidia when CF works, IF AMD gets the CP profiles on polaris 10 right two of them have the potential to take on a 1080 GTX.
I think that some people are missing is that is a seperation in customers requirements being 1080p, VR and 4K (In some games VR will require more that 4K gaming.
There will be released a product from both companies;
1 for 1080p FPS gaming ala GTA V, the witcher / BF4
1 for VR experiences right now in 90FPS / low req 4K gaming
1 30-40FPS+ for 4K
GTX1080 is aiming for highest end polaris is aiming for the low end in this spectre :0
VEGA is aimed to compete with 1080 and to be honest right now 1080 is hitting in a weird spot of nothing; to weak for 4K gaming at 60FPS+ and WAY too powerful for 1080p; 1440p and high-requirements VR-games will benfit from it but right now these games are not made yet and not that many people have 1440p screen. 1080Ti and VEGA 2017 Q1 refresh will be ready for VR maxed out, and 4K 60FPS-ish.
(ofcourse there will be lower-lower-end and entusiast aswell)
Now it's just meh after meh.... Fury is cool as hell, nobody gives it respect though. Still cooler than the 1080, even if it's not as fast.
We need bigger interposers! HEY! The 1930's called, they want their fiberglass back!
Assuming that Polairs has delta color compression, which Hawaii lacks, as well as better shader efficiency and Geometry throughput than Hawaii, then This should give Polaris a 25% performance advantage over Hawaii at the same flops rating. So a 480X should have: 5.5/5.2 *1.25 = ~1.3 the performance of the 390X. That is ~FuryX ish performance, which isn't too shabby. The 480 pro should be slightly faster than the 390X in this case. my point still holds; the 1070 at 2K or 4K will lose to the 980 TI/Titan X.
This is a smart move by AMD, because most likely their architecture will be inferior to Nvidia's but they will have several months where they have the best and newest midrange cards, since they'll be competing against Nvidia's Maxwell. How long has it been since that was the case?
FPS/$ will be at least as good as Nvidia's 1070. I predict Polaris 10 will match or slightly exceed 390/970 performance for $250 or less. It will be great if you are in that budget range.
:P
I don't think the architecture is inferior, when a 2048 shader card will be faster than a 390X, the opposite is true. My predictions are just more optimistic than yours. "Match or slightly exceed 390/970 performance" is a pretty negative guess for a new chip. It must be clearly faster than 390X, everything else would be a catastrophe. Also read the news again, I think you missed the tflops part of it, your guess doesn't make much sense.
Vega will be the more interesting part of this, I expect up to 2 times the CU amount of Polaris 10, meaning between 4096 and 5120 shaders and meaning the GTX 1080 will be easily crushed. After all, the GTX 1080 is only ~25% faster than Titan X, and maybe 5% faster than 980 Ti custom cards. The 1080 custom will maybe be 20% faster than 980 Ti custom, meaning that Pascal isn't as strong as the hype and Nvidia PR makes you believe. Pascal is pretty weak, I expected a lot more from a 16nm chip for 599$+. My hopes are on the Vega GPU of AMD now, this will be a real game changer if my expectations are true, unlike the GTX 1080 which is a small upgrade.
Anyway, we will see.
PS
www.amd.com/en-gb/products/graphics/notebook/r9-m200
What matters is the theoretical single precision FP rating and how efficient the micro-architecture is. If the
5.5 TF rating is true, then the Polaris 10 cards are definitely faster than the 390/390X. How much faster depends on how good Polaris is in comparison to Hawaii, which is Vanilla GCN.
you can expect the 1080Ti and VEGA refresh to be a better than the first "generation". Im not sure if you think my example deserve a generalization or if you are trolling, but to state the obvious; there is no garanty that the VEGA/1080Ti will ENSURE 60 FPS+. But you can expect them to be around 2,5 times faster than say GTX980. That will for the first time mean that almost ANY game is playable and alot with 60FPS-ish. With one single card.