This table is making me very glad I upgraded from the X5650 lol. I'm also very surprised to see the 4790K beat the 6700K at a lower clock speed and with slower ram (DDR3 as well). What's up with that?
Code: Fra:1 Mem:574.60M (0.00M, Peak 830.52M) | Time:00:40.61 | Remaining:00:00.02 | Mem:260.74M, Peak:261.27M | Scene, RenderLayer | Path Tracing Tile 625/625, Sample 150/150 Fra:1 Mem:574.56M (0.00M, Peak 830.52M) | Time:00:40.71 | Mem:260.71M, Peak:261.27M | Scene, RenderLayer | Path Tracing Tile 625/625 Fra:1 Mem:574.56M (0.00M, Peak 830.52M) | Time:00:40.71 | Mem:260.71M, Peak:261.27M | Scene, RenderLayer | Finished Fra:1 Mem:310.79M (0.00M, Peak 830.52M) | Time:00:40.72 | Sce: Scene Ve:0 Fa:0 La:0 Saved: './ryzen0001.png' Time: 00:41.21 (Saving: 00:00.48) Intel Xeon E5-2630v3 stock, buffered @1866MHz RAM (if I remember correctly). Xen Hypervisor Tested in Debian SID x64 headless Vm (16 "CPUs", 10GB RAM). 8 other VMs basically idling.
2:22.35 here, on my burned and battered 8120 @ Stock. ... Iiii might be buying Ryzen when it releases. Ooh, hang on; It looks like the render settings were different for the AMD test, they had the samples set to 100, whereas it defaults to 150 in Blender on my end, which also seems to add a good chunk to the time.
Not bad for an i5, but unfortunately not a qualifier for the table! Edit: sorry for the double post. That's the spreadsheet updated - I've added a cores/threads column for quick comparison, and some of the CPU freq's I've had to guess because the new turbo boost behaviour can be very misleading (eg not boosting on all cores).
Ok, I gave it a crack with my 6600k clocked at 4.8GHz, memory running at 12-13-13-28-1T @ 2666Mhz (2x8Gb) and got the following result -
I'm going to assume that "6600K" is a typo - LOL Updated. Still baffled by the 6700K vs 4790K result.
Did you change the Samples? If I change the Samples, I get 40s or less without any problems. With default Settings I get 1min 16s+, no matter how hard I try
From the results I have seen with this benchmark around various forums RAM speed seems to be quite critical in achieving a good result. That is why my 5820k lags behind. A combination of clock speed, RAM and cores is key to getting top results. One thing AMD did not disclose was the ram speeds they were running. Or if they did I must have missed that part but yeah, RAM speed will really help. Probably because the RAM is being used like the very mischief in this test. The CPU can only work so fast, it needs feeding.
Did some more benching. same system as before (cpu turbos up to 3.2GHz, so maybe that should be its clock speed?) W7 x64 VM: CLI (over xenCenter console): 56.39CLI (over RDP): 57.26_________________________________________ Xeon x5650 @3.5GHz/ 24GB RAM @ 1800MHz W10 x64 (also host for virtualbox): CLI: 1.15GUI: 1.09CLI (over RDP): 1.26 Debian x64 SID VM ("12 cores"): CLI: 1.05.31GUI: 1.12.51 Win7 x64 ("12 cores"): GUI: 1.20.76CLI: 2.05.82 Would be interesting to run this "benchmark" on x5650 Linux host. Make your own conclusions.
LOL Vault-Tec... let's not get ridiculous - a difference of over half a minute has got nowt to do with RAM speed. We already have a 6600K result at 4500MHz and the score is 78.73s. I can elaborate: if I disable HT on my 5820K running 4600MHz I get this: If I reduce sample size to 100 I get this: So it's patently obvious that 23RO's 6600K was running 100 samples, whether he knew it or not.
Erm, no once again - Pre-render sampling settings as above... I simply click and render as previously stated.
You're not understanding me - what I'm saying is that no matter what settings you see in Blender, it's rendering at a lower sample rate. That's the only explanation I can think of for your score; it's categorically impossible for a 6600K at 4.8GHz to be achieving these results legitimately at a sample rate of 150. A cursory glance at the spreadsheet would tell you this.