Next-Gen Pascal GPU's.

Discuss any unrelated topics here.
Post Reply
User avatar
OldGnashburg
Posts: 91
Joined: Tue Nov 19, 2013 3:00 pm
Location: Sherwood Park, Alberta, Canada: A place with free universal healthcare, and lots and lots of oil.

Next-Gen Pascal GPU's.

Post by OldGnashburg » Thu Jul 23, 2015 6:02 pm

Hiya kids. I was just looking up some news on the Pascal GPU's that are going to be released next year, and I read that it is going to have unified memory, and what I got from the article I read (here is the link http://wccftech.com/nvidia-pascal-gpu-gtc-2015/) is that it eliminates the need for the CPU and GPU (in applications and other crap that use GPU hardware acceleration) to wait for each other to sync because they can instantaneously access each other's memory so there is no need for them to wait and sync. Anyways, it will be quite a bit of time before the Pascal GPU's are released but could these allow for us to emulate the RDP at fullspeed while also being both cycle accurate and pixel accurate?
Gnash, Gnash, Gnash...

User avatar
Nintendo Maniac 64
Posts: 185
Joined: Fri Oct 04, 2013 11:37 pm

Re: Next-Gen Pascal GPU's.

Post by Nintendo Maniac 64 » Wed Jul 29, 2015 12:27 am

Just an FYI, there's been a ton of way-too-early marketing and overhyping around Pascal, likely because AMD beat Nvidia to HBM and will also have exclusive first-dibs on HBM2, which is what Pascal will also be using. I mean, Pascal is expected Q2 to Q3 2016 - still a long ways away. For reference we haven't heard anything about AMD's HBM2 GPUs even though they're launching in Q1 to Q2 2016 - that's before Nvidia.

Regarding unified memory, I'd like to point out that this is what HSA and hUMA on AMD's APUs was designed for and that Kaveri, the first HSA-capable APU, has been available for 18 months now. For reference, MarathonMan has gone on record saying that even HSA and/or hUMA combined with the CPU and GPU on the same die would have too much latency for performance benefits; Cen64 is so latency-sensitive that, according to MarathonMan again, it actually is faster to use SMT ("hyperthreading" in Intel-lingo, but AMD's Zen will also have SMT) than to use actual separate CPU cores - something that I've never heard of being the case with any other software.
CEN64 Forum's resident straight-male kuutsundere
(just "tsundere" makes people think of "Shana clones" *shivers*)

CPU+iGPU: Pentium G3258 @ 4.6GHz/1.281v
dGPU: Radeon HD5870 1GB
RAM: Vengeance 1600 4x4GB
OS: Windows 7

User avatar
MarathonMan
Site Admin
Posts: 692
Joined: Fri Oct 04, 2013 4:49 pm

Re: Next-Gen Pascal GPU's.

Post by MarathonMan » Wed Jul 29, 2015 9:57 am

Nintendo Maniac 64 wrote:Cen64 is so latency-sensitive that, according to MarathonMan again, it actually is faster to use SMT ("hyperthreading" in Intel-lingo, but AMD's Zen will also have SMT) than to use actual separate CPU cores - something that I've never heard of being the case with any other software.
Just an addendum: HT/SMT's utility is a function of how well the CPU backend can cope with multiple hardware contexts spewing instructions from the frontend.

Sandy/Ivy Bridge and Haswell/Broadwell both have excellent HT/SMT implementations. Hopefully that's the case for Zen as well.

User avatar
OldGnashburg
Posts: 91
Joined: Tue Nov 19, 2013 3:00 pm
Location: Sherwood Park, Alberta, Canada: A place with free universal healthcare, and lots and lots of oil.

Re: Next-Gen Pascal GPU's.

Post by OldGnashburg » Wed Jul 29, 2015 1:20 pm

Is any hardware accelerated cycle accurate emulation possible? Or is latency just too much of a roadblock? I know this isn't going to be your design for CEN64, is there a feasible method of hardware acceleration out there that you have considered?
Gnash, Gnash, Gnash...

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest