To check your score please check the spreadsheet Lenny has provided. Thanks ! OK, so after a few weeks of letting the Ryzen launch digest I decided to do some digging this morning to see just how good it is. The method of this is simple. First you need to go and download Blender, making sure you get the correct version for your OS (X86 32 bit or X64). https://www.blender.org/download/ It is completely free. Then you need the AMD Ryzen test file. http://download.amd.com/demo/RyzenGraphic_27.blend OK, so go ahead and install Blender. Then open the file (the .blend one) Then click on render, and Render Image (or F12). Once the benchmark has concluded look for your time. And a close up of mine, I got 50.25 seconds. Now when you compare with Ryzen the magic number is between 36 and 37 seconds for a full render. Let's see how other CPUs compare. I am running a Xeon E5 2680 V2 8c 16t @ 3.3ghz with 16gb RAM and a Fury X.
Ooh this looks like fun. Very easy to DL and install Blender, in case anybody's wondering. It might be worth putting a table in the OP so we can quickly reference the performance stats based on hardware & OS. This is running on a fresh install of Win7 SP1 using my new X99 setup, all stock settings (3.3GHz, 32GB DDR4 @ 2400MHz).
Here's mine for comparison - Intel Xeon X5650 @ 3.6Ghz on Windows 10 x64. Interesting to see there's not much difference (time was 1.01:58) - so just over 60 seconds
That's quite a big difference in Blender. 5820k @ 4.5ghz. Crappy ram speed holding me back I would guess.
Got a GSOD 2:30 into the render for me... and a reminder of what a mess blender is [imo]. Extrapolating out how long it took up til that point i'd estimate a render time of ~4:40 for my G3258 though i'd wager my HDD held the render back, blender was fair thrashing it throughout. EDIT: Wasn't that far out in my estimation... manage to get a completed render out of my PC in 4:30 EDIT 2: To see how consistent the render times were i ran it a couple more times... Results were as follows: run 0: bugcheck after 2m30s run 1: 4m30s run 2: 4m14s run 3: 4m35s run 4: 4m07s so it can vary quite a bit. EDIT 3: for the sake of completeness, the system: Pentium G3258 [2c/2t] @ 3.2GHz 8GB [2x4GB] 2133MHz DDR3 [@ 11-11-11-31] GTX 650ti @ 928/1350 w/ 1GB GDDR5 Samsung HD103UJ HDD EDIT 4: Ran it again using the 32-bit version... for the lulz and because i'm bored. Run 1: 3:36.97 Run 2: 3:53.63 Run 3: 3:27.82 So yeah... Not especially consistent...
@Vault-Tec, you ain't lying about your RAM - I'm hot on the heels of Ryzen if I overclock to 4.5GHz. I did virtually no tweaking to get this score, other than manually set ram to 2400MHz in case it wanted to change of its own accord. I changed CPU multi to 45, vcore to 1.275 and VCCIN to 1.9. Bish-bash-bosh, 39 seconds. @oasked, the 5650 is quite a bit behind the 5820K in multi-threaded apps (I just did some tests to this effect as I still have my X5650 rig). In Cinebench R15, the 5850K at stock trumps the X5650 at 4.4GHz. The biggest failing of X58 is the memory throughput which is extremely weak compared to the newer architectures. I never thought upgrading from 6C/12T to 6C/12T could be so nuts.
VT - prepared to collect the results + spec of everyone in the first post? Or shall we create a new thread to collect? Could be great to have a whole pile of results ready for when Ryzen drops. Me: 37.74 seconds at 150 samples, 25.58 at 100 samples. My seriously unoptimized machine: 5960X at stock 3.5G and 2133MHz DDR4 set to bog slow SDP in dual channel, not quad (semi-passive cooler too big). 16 threads of Haswell-E do the business. There's a nice spreadsheet here: https://docs.google.com/spreadsheets/d/1O2Or6XOETZr3a4gLAuCuYEx8m5X_lv23LdwZiCZpBcM/edit#gid=0 Reference Ryzen result: 100 samples 25s, 150 samples 36s. The test ^^ runs at 150 samples and you can change it on the right column settings. My 5960X at 3.5 matches a Ryzen ES (which is also dual channel). Anyone know the clock speed that AMD benchmarked the chip at?
Mine runs it in 52.39 I need to start OC'ing it I think Changed my system to a 7700K and now seeing 39.06, second pass 38.02
I'm not so good with things like that due to memory/meds. If some one else wants to then feel free The Ryzen chip was 3.4ghz no boost IIRC. They mentioned later they were still working on tweaking the boost clocks Yeah ram speed matters in Blender apparently. I'm still running 2133 from Haswell E's launch with DDR4. That may be because you "only" have 8gb ram and the 64 bit OS will allow Blender to use all of it and more. Ram speed/amount really can affect the scores apparently. I went from Sandy E @ 4.9ghz (3970x) to my 5820k @ 4.4 daily. I scored around the same in most benchmarks, but anything that used ram was faster. I don't regret it either, because to maintain those clocks on Sandy meant my motherboard only lasted me a year. I did get a RIVE as a replacement, but bottled it and went for warranty instead, given how bad companies are for holding replacement stock these days (glares at MSI). I hardly even bother running mine overclocked now.
So I have been told that basically in Blender Compute is enabled for AMD but disabled for Intel. If you enable it for Intel a 4790k scores 34 seconds. Just trying to figure out how you enable it now.
@ Bindi, I can put up a spreadsheet on google docs if VT is happy to link it in the OP. @VT, if that's true it's quite the revelation, but there's also the Handbrake result where Ryzen actually beat a 6900K, which leads one to conclude that it isn't down to software optimisation.
Hang on.. Just did some reading and "compute" in Blender is to enable your GPU on the action. So that would be cheating, unless AMD did it. But if they did that they would then have had to leave the 6900k "vanilla" and would get caught out. So it's unlikely they did that. Off the bat the demo uses no GPU rendering. You can enable OpenCL and CUDA with a hack though apparently. So yeah, could just be an Intel fanboy with a not so fresh bunch of grapes.
Sounds like it. I've put together a quick and simple spreadsheet here if you want to link it in the OP. I'm only going to include high-end processors in the sheet as it's already a stretch comparing hyperthreaded quad cores to a hyperthreaded octa core.