The AMD Ryzen All In One Tread /Overclocking/Memory Speeds & Timings/Tweaking/Cooling

I have just finished a whole bunch of testing before installing windows patch and after. Gaming performance still the same , gpu benchmarks and cpu looks about the same but here come the big but ... I did noticed a big performance hit on my SSD writes and to a lesser extent reads too are affected. Anything IO demanding workload related seems it's going to be affected one way or another . Cleaned W10 , ran optimized tool , trimmed and same settings as always my Crucial MX300 1TB SSD it is performing much slower than before the patch :/ . Tested with Cristal Disk Bench , AS SSD , ATTO and Anvil. Those are my findings so far.

Please check your Storage performance and post your finding guys , it was said that this patch would affect high IO workloads , and i believe it did affect us Ryzen users on the storage department.

 

Thanks for sharing and crap!

Well, we can only hope further patches get released that allow AMD systems to operate normally.
Linux should get such patches at least I think.

 

Oh dear ... My old test on my SSD used to be Writes around 510~520Mb/s , and Reads used to be around 540~550Mb/s with my OS optimizations before the Patch. After the Patch i'm only hitting Max 450~460Mb/s Writes , and Reads are around 525~535Mb/s , that is a lot of performance degradation. Damn you Intel ...

 

As I posted in the front page news thread I used to get ~1000 score on the AS SSD benchmark for my OCZ 460A SSD on my Ryzen based system. After the patch I now get ~730 score on the same benchmark. I'm not too happy about that.....

 

Well,looks like micro$oft did it again.

Remember from march 2017:
"AMD Ryzen Performance Negatively Affected by Windows 10 Scheduler Bug",because of: "The Windows 10 scheduler is not able to correctly identify Ryzen’s principal core threads from virtual SMT threads. As a result, the OS doesn’t assign the tasks to a principal core thread, scheduling many of them to a virtual SMT thread."
micro$oft patched Windows,again,to correct that bug in their patch.

Now they need to patch the Windows again,because Ryzen is not affected by this hole mess named Meltdown.

Spoiler: Intel vs AMD flaw

 

Well this is odd. My Intel rig (i4790k) got the patch almost instantly when I checked.
I have checked several times on my Ryzen rig, but no update in sight.
How did you get your patch ? Through windows update or from the update catalogue ?

Benchmarking wise there is NO difference in performance running the AIDA benchmarks / cache tests on the i4790k. Read performance on SSD is about the same - but AIDA doesn't test 4K blocks.

 

I think I'll make that chart my new background image for desktop

Edit: btw where did you find it from?

 

Speeds on the SATA SSDs remained the same for reads and writes. IOPS took a huge hit from read of 93K down to 78K. The 960 EVO reads were roughly the same coming in at 3144 instead of 3210 and writes went from 1543 to 1485. IOPS took a huge hit as well, just like the SATA SSDs. Latency on the ram took a hit too. It is now up from 65.3 to 65.7. All else is the same.

 

On reddit Amd.

 

I'm hoping Samsung can claw back some of the lost IOPS performance with a new driver. All though I haven't noticed any performance difference om my 960 EVO except those you can measure. In daily use it feels the same.

 

^over the weekend was playing with the BCLK setting to see what I could gain back. Settled on a BCLK of 101.2 and a multiplier of 39. The ram is now functioning at 3508 (same timings as with 3466). Got the ram performance loss back with improved latency as well. The 960 EVO saw the IOPs come back to what they were before. The writes are even a little better at 1547. Reads are a lost cause at this point down to 232X from 3210. The 3144 seemed consistent having run the benchmark three times with 30 minutes between runs giving the same results. After a day or two, that reading plummeted to 212X. Like you say though, can only tell the performance loss by looking at the numbers.

 

Ryzen DRAM Calculator 0.9.9 v11

-> https://drive.google.com/file/d/1n-sILfAndZMJ1uVHbjzholbbJl_R8776/view

Spoiler

 

Interesting... still no KB4056892 patch via WU:



only Defender definitions, Flash player update and malware removal tool January 2018.

Checked update history too and not there either. Build ver 16299.125 still.

Did all of you who have the patch install it manually? Because it seems MS isn't even pushing it for Ryzen systems?

 

You can have it from here,if you want it so bad:
https://www.catalog.update.microsoft.com/Search.aspx?q=KB4056892

 

The patch came through WU last week, next day after all this hit the fan. Uninstalled both patches that installed. Confirmed it was the patch causing a few issues, and then had WU install it again. Difference was that WU did not install the other one second time around.

Just found this: http://news.softpedia.com/news/micr...tes-due-to-issues-on-amd-systems-519278.shtml

 

Exact opposite actually, quite happy it hasn't appeared. Was wondering how people here got the update.

 

So pulled on purpose, that explains it then. Didn't get it last week either though.

 

I installed it from there. Haven't had a single issue.
My i7 4790k got it from WU. No issues on that either.

 

Ok, good to know. I'll wait for WU rollout though.

 

Installed 3501. Skipped 3101.

Initial impressions is that it is a step in the right direction. Was able to post with bootstrap 3533. Will not make it into Windows. Not willing to play with CLDO_VDDP yet. Bootstrap 3600 is still a no go. Will not even post; memory hole must have moved as a result of certain default settings. Ram is trained a bit differently. Twr now defaults to 20. Default under 3008 was 26. All the rest stayed at the same defaults. Aside from that, previous settings booted and was stable at the same voltage as with 3008. Have not tried to see if less voltage is a possibility yet.

Do not know why but still anxiously waiting for 3600 to be a reality. Do not care about the whole 3600 at C16 is not as good as 3200 with tight timings. Fixated on achieving that speed for some reason.