Thursday, April 14th 2016
AMD to Launch Radeon R7 470 and R9 480 at Computex
Computex 2016 could see some major consumer graphics action, with AMD reportedly planning to launch two mid-thru-performance segment products on the sidelines of the event - the Radeon R7 470, based on the 14 nm "Baffin" (Polaris 11) silicon, and the Radeon R9 480, based on the 14 nm "Ellesmere" (Polaris 10) silicon. The R7 470 could succeed the R7 370 series in not just performance, but also offer a leap in energy efficiency, with a TDP of less than 50W. The R9 480, on the other hand, could feature a TDP of just 110-135W (R9 380 is rated at 190W).
The R9 480, based on the "Ellesmere" (Polaris 10) is shaping up to be a particularly interesting silicon. It's rumored to feature 2,304 stream processors based on the 4th generation Graphics CoreNext architecture, with 2,560 stream processors being physically present on the chip; and a 256-bit wide GDDR5 (GDDR5X-ready) memory controller. 8 GB could be the standard memory amount. AMD could keep the clock speeds relatively low, with 800-1050 MHz GPU clocks. Imagine R9 390-like performance at half its power-draw.
Sources:
VideoCardz, VR World
The R9 480, based on the "Ellesmere" (Polaris 10) is shaping up to be a particularly interesting silicon. It's rumored to feature 2,304 stream processors based on the 4th generation Graphics CoreNext architecture, with 2,560 stream processors being physically present on the chip; and a 256-bit wide GDDR5 (GDDR5X-ready) memory controller. 8 GB could be the standard memory amount. AMD could keep the clock speeds relatively low, with 800-1050 MHz GPU clocks. Imagine R9 390-like performance at half its power-draw.
97 Comments on AMD to Launch Radeon R7 470 and R9 480 at Computex
The lower power draw should also help AMD get market share in the high margin gaming laptops (assuming it's not far behind Pascal in power efficiency). Currently nearly all gaming laptops are Intel / Nvidia, so AMD really needs better power efficiency to compete there.Not everyone is looking for a power-hungry and very expensive high end GPU, or already owns a high-end GPU. The GTX 970 didn't give much better performance than the GTX 780 Ti, yet it sold like hotcakes.
www.techpowerup.com/reviews/AMD/R9_295_X2/17.html
www.techpowerup.com/reviews/AMD/R9_295_X2/13.html
It's "playable", if you're comfortable with the price of that and a return of 26FPS in Crysis3, but it definitely doesn't "decimate".Based on what?
We need more power for 4k, that much is clear. If the manufacturers decide to throw us 10-20% better performance and put everything else in cutting power consumption, 4k gaming will be a no-go for one more generation. Which isn't a terrible loss, because there aren't many affordable, good 4k monitors on the market anyway.
With the naming here its quite obvious that there will be a 490 in the future.With "decimate" i mean beat every other card out there by a large margin, not every other game out there, 26 FPS in Crysis3 is actuality decimate if the next best thing gets 14 FPS.
IF the 480 has 2304 stream processors and a higher clock ( say 1250 MHz) the compute power will be on pair with the 290X (2816 stream processors at 1 GHz). Add in that the 290X is GCN 1.1 and the rumored 480 is GCN 1.3 there is a good possibility for it to match the 290X.
We can at least agree on the need for more power, but, remember that the name 480 indicates that it is a replacement for the 380, and that card is noticeably weaker than a 290X (the 380 is approximately 60% of the performance of a 290X according to This review)
6870/6850???
this looks very promising, but with no reviews speculating is a bit pointless.
If the R7 470 really performs better than the R7 370 while consuming <50W, they can sell loads of them to budget gamers and laptop manufacturers.
I'd likewise be surprised if the upcoming Pascal card is faster than the 980ti.
www.guru3d.com/articles_pages/hitman_2016_pc_graphics_performance_benchmark_review,7.html
I think DX12 is not really polished enough to demonstrate tangible benefits (10% for 390X and only 5% for Fury X in the above link). Obviously the Async agnostic Maxwell gets zero lift.
Like I said and I always say - we'll need to wait. The games will make a difference to each camp so cherry picking titles never helps (and let's not argue that in showpiece demo's the vendor will choose what suits their marketing best). We need a good spread of neutral and sponsored to see who's fiddled what.
I just hope if AMD come up trumps, they let AIB's release custom's because I don't like stock cards (coil whine, coolers etc).
EDIT:
Found this tidbit. AMD are shoving Async hard because they know Nvidia will not keep up. Funny, Nvidia used to do that with tesselation.
www.gamersnexus.net/news-pc/2309-amd-hitman-dx12-ace-workload-managementIt's going to be like I said ages ago - the games will be chosen by AMD and Nvidia and hobbled accordingly. Welcome to the same future. At least now AMD can hobble back.
But we all lose.
Most likely we will never see a game which does two complete optimized pipelines, so comparisons between Direct3D 11 and 12 within a game wouldn't reflect the API itself.
The other worrying part of the equation is that the sheer amount of work a game developer now has to take onto themselves just to get a DX12 feature rich game out the door and working correctly. As the Hitman dev says:Given the reluctance of a large section of people to move to Win10, it doesn't auger well that devs can continue to use the need for a DX11 codepath as an excuse for sloppy DX12 programming.
Tangental (at best) to this, I found this quote fairly amusingDevelopers wet dream - but no we won't be using it
Vulkan an API for all platforms...that looks like it won't be supported on all platforms !?!
The number of FP32/FP64 units in this chip are likely better utilized than the previous generations of their GCN chips, which didn't change much since its release in 2012.
Spending 4 years or so to find and eliminate bottlenecks in their GCN microarchitecture and make architectural improvements over the original GCN designs should pay off well with these chips.
Also, back when they announced Polaris, they mentioned that Polaris chips will get some form of memory compression to improve memory performance. This in a way implies that they're already hitting the bandwidth limits of GDDR5 with their Polaris 10 cards, which the GTX980 didn't hit even when over clocked.
I predict that the 470 will replace the 390X and that the 480 will replace FuryX performance wise.
You could also make the case that developers are inherently lazy (or too eager to put profit above the gaming experience) - although the only proof for it would be the slew of games that are released with pervasive bug issues, insipid gameplay, lack of originality, the barest amount of effort when porting a game to PC from console, and shorter games with larger DLC add-ons.Not sure how that tracks. Since GCN debuted the FP64 rate has been tied to the market the GPU was primarily aimed for. In 2012, Pitcairn and Cape Verde had 1/16 rate while the more compute minded Tahiti had 1/4. The second GCN iteration was similar - Oland and Bonaire both 1/16 with Hawaii 1/2 (still the GPU basis for AMD's compute cards). The third iteration made 1/16 standard across both Tonga and Fiji. It would be almost a certainty given AMD's "2.5X performance per watt" claim that Polaris will also feature a 1/16 FP64 rate.Seems highly unlikely but I like your optimism.