Thursday, April 13th 2017
AMD's RX 500 Series Specifications, Performance Leaked
A leak of what appears to be AMD's presentation on the Radeon RX 500 series has brought confirmation on specifications and details of the new line-up - which includes the RX 580, RX 570, the (until now) missing RX 560, and the RX 550. It would seem AMD has now opted for a new, dual-fan reference design, instead of their usual single-fan, blower-style coolers.
The RX 580 has a base clock of 1257 MHz, and a boost clock of 1340 MHz (74 MHz greater than the RX 480's 1266 MHz). It's a Polaris chip alright, packing the same 36 Compute Units (2304 Stream Processors, and up to 8 GB of GDDR5 memory across a 256-bit interface. AMD apparently decided to compare the RX 580 to the R9 380, which allows the company to show some relevant performance improvements (which wouldn't be possible with the RX 480, now would it.)
The RX 570 also keeps the same core configuration as the RX 470 it substitutes, but with an increased boost clock up to 1244 MHz (by 38 MHz over the reference values for the RX 470) over 32 Compute Units, and will be offered in 4 GB and 8 GB variants.
The RX 560, on the other hand, is where AMD has decided to mix up the game a little, with the RX 560 seeing an increased Stream Processor count over its previous-gen counterpart: 1024 against the RX 460's 896. It's TMUs also see an increase from 56 to 64, though the number of ROPs remains the same. It also features the greatest boost clock increase of the entire RX 500 series, at 57 MHz (to 1257 MHz on the RX 560 over the RX 460's 1200 MHz.) This card will see configurations with up to 4 GB memory over a 128-bit bus.
Finally, the new kid on the block, the RX 550 comes in at 512 Stream Processors (8 CUs) clocked in at 1183 MHz boost. Its 2 GB of memory seem appropriate to the crunching prowess of the little card that will, which AMD is positioning for competitive MOBA's and lower-requirements gaming.
Sources:
Videocardz, Jisakutech
The RX 580 has a base clock of 1257 MHz, and a boost clock of 1340 MHz (74 MHz greater than the RX 480's 1266 MHz). It's a Polaris chip alright, packing the same 36 Compute Units (2304 Stream Processors, and up to 8 GB of GDDR5 memory across a 256-bit interface. AMD apparently decided to compare the RX 580 to the R9 380, which allows the company to show some relevant performance improvements (which wouldn't be possible with the RX 480, now would it.)
68 Comments on AMD's RX 500 Series Specifications, Performance Leaked
Anyhoo, I'm eagerly waiting on what the partners will do with the RX 580. What is also great that the stock coolers look semi decent in terms of cooling capability for once.
Rx 450 = HD 7770 Ghz (about 1.2 Tflops, more bandwidth)
Rx 560 = HD 7870 Ghz (about 2.5 Tflops, more bandwidth)
Rx 570 = HD 7970 w/ o/c (less Tflops; 4.5, more bandwidth)
Rx 580 = HD 7990 (ok maybe a 12 GB version with c/f support)
HD 2900XT with 512 MB of GDDR3 and 0.475 TFlops of powa!!
So blame Oculus for that shitshow of a marketing spin if you really need to lay blame. :P
Reality ...
I am a huge fan of their reference single fan blower. Although loud, it saved me from having to use a heater whilst gaming during winter :P
RX 500 series naming should have been left for future products(e.g after Polaris is DONE as VEGA is a specific family designation) now it will be RX 600 series all the way to maybe RX 900 series and they will have to rinse and repeat totally new naming/numbering system when they could have stretched it out a bit more to not have to spend billions more on marketing etc *face palm*
RX 450-455..460-465..470-475..480-485..not complicated for billion+ dollar company to have figured out(well their marketing teams aside that is numbskulls they seem to be) likely would have been the best way forward(if they so choose to use a cream of the crop super TWKR/custom clocked version type thing, THEN they could add the X to them to denote even higher performance levels while not confusing end consumers for nothing)
RX 480 can keep up with a 485(though likely consume much more power /heat to reach these higher clocks if at all STABLE, 485x as a simple example out of reach for any 480 to hit via much more "cherry binned" chips in their use type deal) If the GCN version they are using currently(supposedly GCN 4.0)is "so adaptable with tons of performance left to be tapped" then they could have/should have went this route.
I appreciate the "jump" i.e higher out of the box speeds and lower power/temps to do this(claimed lower at launch pricing, but I seen the BS that was spewed for the so called $199 launch price for 480..yeh)
Numbering used is moronic, at a simple level, 400 series specific die, 500 series should be new specific die, not just a "optimized/tweaked" version of the old. like maybe RX 500 series should have been relegated naming for 10nm/7nm generation instead it likely will create undo confusion when someone who has a very nice RX 480 is actually outpacing others with a brand new not the greatest 580(not all 480s will hit these speeds, some will for sure, most top out ~1350 no extra power, ~1400 is the upper limit, and the few will do 1425, so these will be faster at less power to do this hell there is some shaddy custom cards will not even maintain their from factory below 1300Mhz speeds)
In one aspect you know it is a "new card" because it is called RX 580 as example whereas the "old" was called RX 480, but, it damn well better be faster than the old not a year earlier cards using similar number. IMO the numbering would have made far more sense for current/new buyers to choose buying a RX 4x0 or a RX 4x5.
That being said, I base on price, quality etc, I held off buying an RX 480 because living here in Canada, prices pretty much have been sitting at a constant base cost of ~$329+(4gb versions are even more so fubared pricing overall do not get me started on the 470s which are far more wonky pricing) plus tax/shipping (~$10-$40+ more than should be) for a decent model not just reference version.
Prefer blower style BUT not when they have no real ability to deal with heat they are producing(no heatpipes cmon lol) let alone having possible out of spec 6pin when should be an 8 which vast majority of all modern computer owners can use, even including a ~$0.20 molex to 8 pin would have been a great thing all makers to have in the box considering the price gouging they have been doing this generation(certainly ensure WILL run at rate speeds FFS)
One in this thread said something about, why buy this when any 480 can easily do this, my MSI is much faster than this LMAO
Yep some awesome 7700k you "lucked out" will hit 5.2+Ghz low voltage/heat standard cooling, but that is by far not an average..RX 480 tops out shy of ~1400 MAX majority 1375 specific game/benchmarks or ~1350 or so 24/7(extra voltage doesn't always help there is upper limit of how much extra power can help) by RX 500 shipping basically above RX 400 with some overhead remaining to be pushed further while not ballooning their power draw, they are already out of the gate in the top range of what current RX 400 can do(we shall see soon enough huh)
FYI 460 "china version" has the full 1024 shader 16/64 rop/tmu via 16CU core (slight clock speed boost)
standard 460 only has 896 shader 16/56 rop/tmu 14CU core.
SO, RX 560 is essentially the "full fat" version of RX 460 with further pushed clock speed (along with any other improvements that might have been done internally)
IMO another thing they should have done the "weak" 460 with 896 shader should have been LE version with only 2gb of memory(they could save $, made no fan versions by reduced clock speed or whatever to have on the shelf lowest cost possible for need budget GPU crowd) whereas the non LE would have had the higher clocks, more memory, more TMU, additional 2gb of memory to better justify the price they wanted/were selling at.
~10% better performance just with the small clock speed as well as extra TMU enabled non china RX 460 vs their RX 460.
Done now :P