Originally posted by Adarion
View Post
Announcement
Collapse
No announcement yet.
AMD A10-7850K vs. Intel/AMD CPU/APU Comparison
Collapse
X
-
Michael Larabel
https://www.michaellarabel.com/
-
Originally posted by chuckula View PostHave you actually ready AMD's own server roadmaps? The only "update" is that in 2015 AMD will relabel the exact same Piledriver based Opterons that it was selling last year. That's it. Oh, and Kaveri is being re-badged as a "server" part FWIW.
Btw, Intel is also still 1 or 2 generations behind on Xeon chips as well. Not sure why high end server chips has to be that much behind, but it seems to be common.Last edited by carewolf; 17 January 2014, 08:23 AM.
Comment
-
Originally posted by carewolf View PostI see 4 core streamroller based opterons coming out 1H 2014, what are you seeing? http://www.amd.com/us/press-releases...013june18.aspx
Btw, Intel is also still 1 or 2 generations behind on Xeon chips as well. Not sure why high end server chips has to be that much behind, but it seems to be common.
Comment
-
Originally posted by carewolf View PostI see 4 core streamroller based opterons coming out 1H 2014, what are you seeing?
Btw, Intel is also still 1 or 2 generations behind on Xeon chips as well. Not sure why high end server chips has to be that much behind, but it seems to be common.
As for Intel being "behind" their high-end server parts use Ivy Bridge cores which are behind the desktop Haswell cores. However, they give you a MASSIVE number of those cores and the power efficiency metrics are off the charts good. There's a reason that the new Mac Pro can pack a 12-core CPU into tiny wastebasket form factor.
Comment
-
Originally posted by chuckula View PostUh.. I said that AMD was rebadging Kaveri and calling it an Opteron in my earlier post. You just completely agreed with my point and then tried to make it seem like I'm wrong... you do realize that Kaveri is a "4 core" Steamroller part? (using AMD's rather loose definition of "core") As I said, and as you just agreed with me, AMD is not coming out with high-end Steamroller parts. They don't exist. A dual-core lower-priced lower-power usage Intel part will beat those chips in the large majority of CPU benches as Phoronix showed. In the server, the IGP will be even less useful for the vast majority of real-world server workloads.
As for Intel being "behind" their high-end server parts use Ivy Bridge cores which are behind the desktop Haswell cores. However, they give you a MASSIVE number of those cores and the power efficiency metrics are off the charts good. There's a reason that the new Mac Pro can pack a 12-core CPU into tiny wastebasket form factor.
Comment
-
Am I right in stating that it's going to take a while before software actually uses the real power of HSA?
There's a ludicrous amount of GFLOPS on tap in a Kaveri APU but it needs the software to leverage it?
I'm hoping one day I can rebuild my media server with an HSA enabled APU and get much faster transcoding than I can with traditional x86 CPUs.
Will it ever be possible for compilers to automatically optimise the binaries to use the best Compute Core for the job based on the source code?
Isn't that what LLVM is for?
Comment
-
Originally posted by RoboJ1M View PostAm I right in stating that it's going to take a while before software actually uses the real power of HSA?
There's a ludicrous amount of GFLOPS on tap in a Kaveri APU but it needs the software to leverage it?
I'm hoping one day I can rebuild my media server with an HSA enabled APU and get much faster transcoding than I can with traditional x86 CPUs.
Will it ever be possible for compilers to automatically optimise the binaries to use the best Compute Core for the job based on the source code?
Isn't that what LLVM is for?
Comment
-
Originally posted by RoboJ1M View PostAm I right in stating that it's going to take a while before software actually uses the real power of HSA?
There's a ludicrous amount of GFLOPS on tap in a Kaveri APU but it needs the software to leverage it?
I'm hoping one day I can rebuild my media server with an HSA enabled APU and get much faster transcoding than I can with traditional x86 CPUs.
Will it ever be possible for compilers to automatically optimise the binaries to use the best Compute Core for the job based on the source code?
Isn't that what LLVM is for?
Comment
Comment