Recent content by Dresdenboy

Reducing the L3 size could be one of the easier things. Indeed, there could be many possible reasons for this. And a full CCX is more like a topological building block, not a physical one. Shaving off some L3 in a single CCX design might come with a negligible perf loss if at all (no 2nd CCX to...

In contrast to SiSoft Sandra it is also good to look at some "real world" benchmarks: http://www.tomshardware.com/reviews/amd-ryzen-threadripper-1950x-cpu,5167-12.html Also the Intel Fortran compiler might even be a good choice for Ryzen (depending on code and problem size).

Moin moin, NUMA unawareness should only cause some inefficiencies due to an additional hop to reach a different socket's or die's physical mem range. Similar to Hypertransport.

That last posting by benbennett in AMD's forum is interesting. 46h (and counting) no crash with using a low latency kernel.

No. :)

To me it looks like there are just 256b AVX units. Multiple sources mentioned one 512b "port" with one fp port not being available during a 512b op, which means, those ops would be executed on two 256b units.

That's right and likely not worth the extra money for maybe 5% more fps. It'd be better to put that into a GPU then. But as I'm also doing deep learning and data processing, I got 2x16 for the R7. A TR buyer might also have other use cases in mind which benefit from more memory. Even a RAM disk...

I thought they would simply be slid (sp?) into the rails in the cover.

Some sites and magazines reported up to two digit points higher performance with dual rank mem (both apps and games). The software seems to benefit from some parallelism in the memory's inner workings. 2x the channels increase that even further. Points against that are the likely higher...

As a part time perfectionist I can understand the itch caused by a forced compromise like getting the highest ST performance with also a good MT performance esp. with 256b SIMD together with some limited I/O capabilities.

Typhoon Burner gives the most detailed info.

HWMonitor voltage is useless. Try HWinfo.

IIRC you were the first, who reported such problems. How about contacting AMD and getting some free test systems to help them out? ;)

Being 4 DIMMs this sould be fine! But you'll miss out on the dual rank performance gains, which are like 200-500 MHz higher RAM speed for several games and apps. ;) All those OC DIMMs are rated at JEDEC 2133 speeds or so, but come with an pretested headroom (the XMP setting). Why shouldn't ECC...

This bug is related to 16b real mode. There might be a few, but not many big players relying on such old software.