PC Perspective Podcast #461 – 08/03/17
Join us for AMD Ryzen 3, Threadripper, Logitech Powerplay, and more!
You can subscribe to us through iTunes and you can still access it directly through the RSS page HERE.
The URL for the podcast is: https://pcper.com/podcast – Share with your friends!
- iTunes – Subscribe to the podcast directly through the iTunes Store (audio only)
- Google Play – Subscribe to our audio podcast directly through Google Play!
- RSS – Subscribe through your regular RSS reader (audio only)
- MP3 – Direct download link to the MP3 file
Hosts: Ryan Shrout, Jeremy Hellstrom, Josh Walrath, Allyn Malventano
Peanut Gallery: Ken Addison, Alex Lustenberg
Program length: 1:38:20
Podcast topics of discussion:
-
Week in Review:
-
News items of interest:
-
-
1:00:45 Looks Like Vega Nano is GO!
-
-
Hardware/Software Picks of the Week
-
1:25:45 Ryan: Logitech G903
-
1:28:10 Jeremy: Old but still good, ASUS RT-N66U
-
1:29:25 Josh: Cool pathtracer for Quake 2
-
-
Closing/outro
Nobody will these card on
Nobody will these card on shelves and this is why:
https://www.overclock3d.net/news/gpu_displays/retail_staff_reports_high_levels_of_mining_performance_on_amd_s_rx_vega/1
Sorry you guys missed the Radeon Pro SSG for $7000. It is a killer card. It will make waves.
https://youtu.be/nuFGUVg0U9U
https://youtu.be/nuFGUVg0U9U
For those of you who missed
For those of you who missed the post-show, it went something like this.
Just some rambling…
For
Just some rambling…
For VEGA as with other older GPUs the most significat issue was “gaming” drivers. For example AMD shader compiler (in driver) was worse at optimizing shaders than equivalent for PS4 (NDA, so can’t share more details). Developers on PCs didn’t have tools to profile code and make it run fast and utilize 90%+ of the GPU. You have GPUView, Pix, Radeon GPU Profiler now, but not until quite recently. Compare it to Nvidia, where they have really nice shader compiler with nice optimization passes and they have a huge team of people that also “manually” tweak drivers for each new released title (shader fixup/even replacement under the hood). Their GPUs where utilized in like 95% in all optimized titles. AMDs in the 50-75% range, depending on title. Except maybe Vulkan inside Doom 2016 – go ahead, capture some traces with RGP, and see that it’s now utilized, that’s also why You see such big difference when comparing OpenGL and Vulkan on AMD cards. That’s also why all FLOPS comparisons are stupid. AMD should win perf on every price point since R2xx and maybe even earlier. If You don’t take % of GPU utilized during a test, FLOPS means nothing. AMDs cards are crazy fast, but until recently we were handicapped by tools / drivers. Now this is changing, which is mostly the result of AMD having provided CPU/GPUs for the consoles, Sony/Microsoft also had their input, and most of all devs shared their pain points multiple times. On Consoles You have a lot of tools to profile to use CPU/GPU fully, now we have same on PC, and that’s good, but it will take time. Why developers wanted tools and not “optimizing” behind the closed doors inside drivers? Because to push industry forward You have to know more, and more internals. Have standards for ALUs width, or at least know the ALUs width, swizzling, etc. Have common high-performance, highly parallel language for shaders, compute and more. For developers VEGA is more open than anything before. ISA docs are available for GCN3-4, they will be available for GCN5. As noted before profilers are here, pricing looks awesome for FLOPS/$, etc.
The last tidbit is power. For the FLOPS You get its on similar level with NVidia, but only if GPU will be utilized highly. Otherwise You can think of it as space heater, unless power-gating of CUs, or other things will get enabled in drivers (AMD told it was not for VEGA FE).
As for the release a year after 1080, when comparing VEGA64, I think it’s much faster than 1080 for similar/less money. I wished for it to be lunched at Q1 2017, then it would be the amazing, right now it’s ok.
In summary VEGA should bring competition back, and next round of GPUs will be interesting. NVidia will either have to open more or share more profiling tools for their cards and specs for devs (even under NDA, I don’t care, but do it) or keep their huge support / optimization teams in place. AMD will already have tools in place, drivers are getting better and better, and most of all have money from Ryzen/Threadripper/Epyc/Vega to either create a big support team or build better GPUs. Either way competition should be fierce next year and consumers should be happy. Until then we will soon have RX Vega to play with.
Cheers 🙂
Ps. Sorry for my bad English, not a native speaker…