New AMD Radeon RX Vega Details Surface In Linux Patch - 4096 Shader procs

Published by

Click here to post a comment for New AMD Radeon RX Vega Details Surface In Linux Patch - 4096 Shader procs on our message forum
#5427625
data/avatar/default/avatar02.webp
Judging by the specs it should knock the f*ck out the 1080ti. If that's not the case then i ****ing pity AMD with their GPU lineup. Why would you invest this time and money into a GPU that's as fast as a dual rx580.:bang::bang:
If it is really performing as dual rx580 its great it will be faster than 1080 and near/or equal 1080ti/titan xp. I can't find it wrong. U mad bro? :bang:
#5427627
https://forums.guru3d.com/data/avatars/m/250/250066.jpg
Looking at that spec sheet, it is like a sli 570 XD Might cry a little if this ends up being the case, i highly doubt it. but it would be funny for us to wait so long for a duo card
Reading a guy post that is saying "sli 570" make me feel bad for guru3d forums unless you are speaking of gtx 570 sli :3eyes:
#5427629
https://forums.guru3d.com/data/avatars/m/130/130124.jpg
If it is really performing as dual rx580 its great it will be faster than 1080 and near/or equal 1080ti/titan xp. I can't find it wrong. U mad bro? :bang:
yeah my bad, corrected my statement.
#5427630
data/avatar/default/avatar22.webp
Just because it has a same arrangement of shaders, TMU and ROP units it doesn't mean that VEGA = FIJI It's also not just Fiji with higher clock and smaller node. Vega has a new scheduler, new pixel and vertex engine, tile based rasterizer, new more flexible compute unit, new memory controller... That's about the same absurd as saying Volkswagen Up! with 1.0 engine is the same as Mitsubishi Lancer Evo X with a 2.0 Turbo engine, just because they both have 4 cylinders.
Stop making sense. You will derail the echo chamber in this thread.
#5427631
data/avatar/default/avatar24.webp
rx570 cf equals easily or surpasses 1070/1080gtx rx580 cf equals easily or surpasses 1080ti/titan xp let us see finally what Vega is capable of it's either faster than 1080 or faster than 1080ti/titan xp That's my opinion. End of speculations. I await official benchmarks.
#5427633
data/avatar/default/avatar21.webp
Comparing AMD teraflops to Nvidia teraflops is idiotic. Rx480 has 5.8 tflops and the 1060 has 3.8 tflops yet they perform roughly the same. This has been the case for many years. AMD's 12.5 tflops is equal to 1080 performance if not a little less. And if Fury is any indication then forget about overclocking.
#5427635
https://forums.guru3d.com/data/avatars/m/115/115462.jpg
While I doubt that it'll be faster than a 1080 Ti, I sure hope that it'll at least be as fast as a 1080. Either way though, it's 1 year late to the party.
#5427636
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
That doesnt look like its going to beat a 1080ti sadly 🙁 I still hope it can but these specs feel underwhelming at first glance 🙁
Of course it wont. We have seen it on level with 1080 so that would be a huge jump in performance.
Ppl need to understand that the only reason why Nvidia can clocks the bejesus out of the Pascal is the low leakage and optimized perf/Watt onchip operations. That is the reason behind the "clock gap".
Not entirely true. The tech is different: Nvidia uses different sized pipelines to what AMD uses on their GPU's. I think Nvidia uses smaller (that do less work each clock cycle) so it can clock higher, AMD uses bigger (that do more work each clock cycle).
#5427637
data/avatar/default/avatar18.webp
Let's calculate something. Considering the pure similarities at the specs and then by looking at guru3d reviews/benchmarks. Now then, if you double all the frames/scores of RX570. You have a card equal and/or faster and/or slower than TitanXP/1080ti. If Big Vega has the full strength of 2x570 at least, it's king. 🙂 BUT, If it equals rx570x2 it will trade blows with titanxp/1080ti as expected :3eyes:
#5427638
https://forums.guru3d.com/data/avatars/m/80/80129.jpg
Comparing AMD teraflops to Nvidia teraflops is idiotic. Rx480 has 5.8 tflops and the 1060 has 3.8 tflops yet they perform roughly the same. This has been the case for many years. AMD's 12.5 tflops is equal to 1080 performance if not a little less.
1060 is 3.8tflops at stock clocks. At average boost clocks it's 4.4tflops and at maximum stock boost it's 4.8tflops (1911mhz). That's not to mention that the RX480 pulls away in most DX12 titles where it's "true efficiency" comes into play nor is it to mention that there are multiple architectural changes designed to improve utilization in Vega. So it should perform closer to it's theoretical tflop range. Also most people here know what you posted anyway - they use tflops to assess general performance, not nail it to a specific spot.
#5427645
data/avatar/default/avatar08.webp
It's really triangle throughput and single-threaded submitted draw calls that are bottlenecking AMD cards. Polaris has almost fixed the triangle part, Vega will do even more for that. My personal question is about the new memory controller and command processor. If it manages to feed the more flexible SIMDs properly, we might even have an efficiency surprise in our hands. I still believe that the initial benchmarks will disappoint, but this will end up to be the card to get, as is tradition.
I cant say that I know what bottlenecks GCN, but as far as I can remember from looking in perf/watt while gaming (because older cards did not had power limiters like cards these days) everything bigger than pitcairn comes with decreased perf/watt. nV cards obviously does not have this issue nad you can clearly tell in which SM iteration card belongs just by looking at perf/watt. I really hope they will improve GCN with vega even if Im not going to buy or recommend to anyone any AMD GPU in near future
#5427646
data/avatar/default/avatar20.webp
Not entirely true. The tech is different: Nvidia uses different sized pipelines to what AMD uses on their GPU's. I think Nvidia uses smaller (that do less work each clock cycle) so it can clock higher, AMD uses bigger (that do more work each clock cycle).
Where you're pullin this "deeper pipeline" for AMD from? The opposite looks to be true. If anything its Nvidia's GPUs who are more clock efficient, and doing more work per flops, ie per (shader*clock). And NV shaders certainly look "bigger" and doing more work in absolute terms: RX480: 2304 shaders GTX: 1060 - 1280 shaders
#5427650
data/avatar/default/avatar30.webp
Problem is whether Vega can actually reach the clocks needed to make it 2x570 actually and I highly doubt it they can without having ridiculous power consumption and heat. It's not like they did some magic there - Vega is most likely build on basis of Polaris and Polaris is not your most eco-friendly solution when clocks are pushed as we see from 580. I actually doubt it will be able to do anything more than 1300MHz on air with that monstrous die, if that.
#5427652
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
Where you're pullin this "deeper pipeline" for AMD from? The opposite looks to be true. If anything its Nvidia's GPUs who are more clock efficient, and doing more work per flops, ie per (shader*clock). And NV shaders certainly look "bigger" and doing more work in absolute terms: RX480: 2304 shaders GTX: 1060 - 1280 shaders
Please watch this: [spoiler]https://youtu.be/c6sDyJzu6m8?t=1m1s[/spoiler]
#5427655
https://forums.guru3d.com/data/avatars/m/164/164033.jpg
Problem is whether Vega can actually reach the clocks needed to make it 2x570 actually and I highly doubt it they can without having ridiculous power consumption and heat. It's not like they did some magic there - Vega is most likely build on basis of Polaris and Polaris is not your most eco-friendly solution when clocks are pushed as we see from 580. I actually doubt it will be able to do anything more than 1300MHz on air with that monstrous die, if that.
They did some changes that will improve their perf/w a lot. Same as nvidia did with their maxwell cards. It's a bigger change in their core design since hawaii?
#5427657
data/avatar/default/avatar07.webp
Problem is whether Vega can actually reach the clocks needed to make it 2x570 actually and I highly doubt it they can without having ridiculous power consumption and heat. It's not like they did some magic there - Vega is most likely build on basis of Polaris and Polaris is not your most eco-friendly solution when clocks are pushed as we see from 580. I actually doubt it will be able to do anything more than 1300MHz on air with that monstrous die, if that.
The Vega based MI25 is around 1525MHz, and that is a passively cooled professional card. (12.5TFLOPS=4096shaders*2*1525MHz) The Asus 1080Ti is average ~60% faster than a Fury X according to Techpowerup. Fury X has 1050MHz clocks, a Vega AIB i think could be closer to 1600MHz based on the speed of the MI25, that's already 52% more clock speeds. Also as we seen with Polaris 480 the efficiency is improved. It keeps up with the R9-390X while has 512 less shaders, and half the ROPS. In theory this means that the bottlenecks which hampered the Fury X (it cannot make use of the lot of shaders)are resolved. So in theory it can be on par with the 1080Ti without counting the changes above the Polaris chips. (primitive shaders, variable length pipeline, etc)
#5427681
data/avatar/default/avatar06.webp
Please watch this: [spoiler]https://youtu.be/c6sDyJzu6m8?t=1m1s[/spoiler]
AMD GPUs are doing more work per clock Nvidias GPUs are doing more work per flops Nvidia shaders are doing more work per clock Are you really arguing against these very basic facts?
#5427684
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
Speculating Since we are speculating on speculations any chance these specs are not for the largest Vega? Depending on clocks these specs could land Vega anywhere from GTX 1080 to Titan XP. Here is to hopping they knock this out of the park as Nvidia has been as bad as Intel with HEDT pricing, this last generation of GTX10xx cards. :banana:

Previous Page

Next Page