Did you try the latest beta BIOS? Latest from Elmor is supposed to be improved for memory performance. https://rog.asus.com/forum/showthread.php?91766-Crosshair-VI-Hero-new-UEFI-build-1001

Just put Numbskull back into service crunching BOINC. Replaced the dual rank memory with single rank memory. Boy, what a difference. Put in the DOCP profile for XMP 3200 and just like that, the machine boots into 3200 Mhz memory. No fighting the BIOS involved. Ran a couple of hours of Prime95, the latest version on Blend with the FMA2 settings and had to use Optimized settings for the phase converters to pass P95 without errors. For some reason, I guess the increase in the Infinity Fabric speed, from my previous best at 2400 Mhz to 3200 Mhz, increased the socket temperatures from 46-50° C to 79°C. It didn't make any difference in Auto or LLC1 in socket temps. The CPU temps are right back at 80-82° C. where they were before crunching on the previous memory and same 3.825 Ghz CPU core clocks at 100% utilization that they were before. Essentially running stock CPU, SoC and memory voltages and everything mostly in Auto in the BIOS, except for the Optimized phase inverters previously mentioned.

I think I will let this run for a day and see where the results end up. Hopeful there will be no Invalids or Errors. I don't think I will push any higher in clocks for the time being since 3.8 Ghz is where the 1700X is supposed to max out on core clocks. I wish I could get the socket temps back into the 50's. Things will change again I'm sure once I get the AIO water cooler back installed.

[/url]

---

GitHub: Ricks-Lab
Instagram: ricks_labs

Wrong motherboard. I don't have the Hero. I have the Prime X370. No new BIOS' for us yet. Still on the 0504 BIOS from February.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Wrong motherboard. I don't have the Hero. I have the Prime X370. No new BIOS' for us yet. Still on the 0504 BIOS from February.

That is the motherboard I wanted at first but it was not available at the time I placed the order. Could you put 3 Rog Strix 1070 on this board ?
That was my original plan but I turned to a dual PCI-E motherboard because it perfectly fit the esthetic of my build.
The motherboard is the ASRock X370 Killer SLI and I might keep it for a future upgrade of my 2nd machine and go for the ASUS Prime X370
if I can put my 3 1070s onto it. I would save on power consumption having them all on the same rig...
Never had ASRock motherboard before and I am not sure what to expect.

Yes, you could put 3 1070's in it. They would run at X8;X8;X4 which is plenty fast enough for crunching. I am thinking about redoing this rig myself. New case, bigger P/S to handle 3 cards and upgrading from a 240 to a 280 rad. This would make Numbskull the king of the hill among my crunchers.

Asrock is considered the second tier level under ASUS. But I have seen many comments that it has to be considered to be moved to first tier. Just pay attention to how many power phases the board has.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Talking about your crunchers I am looking at them right now and I am amaze with your RAC across various projects. You roughly have twice the crunching power I have but
your RAC is 4 times higher here at seti alone. 10 times higher RAC if we had the other projects that you are active !!! Ok I am not crunching 24/7 but still. It took me 3 days of full
throttle to go from 32,000 to 40,000 RAC. What is it that I am doing wrong ? I am running the opencl_nvidia_SoG (Boinc choose this app for me) maybe its the ''problem'' ?
Looking at your results I see ''anonymous platform'' as my results says ''windows_intelx86'' ? RAC is not important for me but it sure speaks for itself. Something is not optimize
here for sure but being new to boinc I dont know what. Perhaps I could increase my productivity without having to move to Linux ?
Thanks.

Some of that is tweaks to parameters and systems, and some of that system bloat that may or may not apply to you. In general though, RAC is a pretty sloppy measure of performance.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Yes, I use Anonymous platform with app_info and app_config. Just used the Lunatics installer for the app_info and moved all the tuning parameters for SoG app to app_config. With the SoG app you can tweak the hell out of the app thanks to Raistmer.

You can really boost the performance of the SoG app with parameters set for maximum performance. You just have to tune for whatever video card you are using and how much memory it has and how many compute units it has.

I really am seeing an improvement (obviously) on this cruncher Numbskull with the Ryzen 1700X. The output from the 970's haven't really changed but I am doing 4 times as much CPU units. And that is only with max 12 concurrent tasks running for SETI. I also crunch for MilkyWay and Einstein on the GPU's. I have noticed something that I wasn't aware of before with my stable of FX processors previously. It seems that BOINC understands what you have for CPU processing power and sends accordingly more CPU tasks assigned to the CPU. I always thought that tasks were tasks, irrespective of whether you were crunching them on the CPU or GPU. I don't know what exactly has changed in how BOINC is sending me work. It might have everything to do with using the rescheduler. But I am seeing usually 60-70 CPU Arecibo non-VLAR tasks assigned to the CPU that I can reschedule for optimum crunching on the GPU's. This has caused my tasks on board for the GPU's to sometimes exceed 300 tasks. I should only be able to have 200 tasks onboard for the GPU's according to standard BOINC allocation. Whatever the mechanism is ... it is welcome as I always thought the hard limit of 100 tasks per GPU was limiting, Especially on Tuesdays.

The tentative upgrade for this system would be to populate with 2 or 3 1070's. That is what I have in my first main crunching machines. This system was the slowest system and was to be the guinea pig for experiments with Ryzen.

Total success so far. I'm impressed enough to start planning to upgrade to Ryzen across all my crunchers. The improvement is massive relative to total upgrade costs.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Yes you could certainly improve your RAC on your 1070 machines with some tuning. It is a different story for Linux machines as they have access to Petri's special sauce CUDA app which really embarrasses the Windows SoG app.

Try putting this tuning line into an app_config.xml file for your machines. It will still use it even on standard SoG app platform

<app_config>
  <app_version>
    <app_name>setiathome_v8</app_name>
    <plan_class>opencl_nvidia_SoG</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <ngpus>0.5</ngpus>
    <cmdline>-sbs 2048 -period_iterations_num 1 -tt 1500 -high_perf -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 -high_prec_timer</cmdline>
  </app_version>
  <app_version>
    <app_name>astropulse_v7</app_name>
    <plan_class>opencl_nvidia_100</plan_class>
    <avg_ncpus>1</avg_ncpus>
    <ngpus>0.5</ngpus>
    <cmdline>-unroll 24 -oclFFT_plan 256 16 256 -ffa_block 16384 -ffa_block_fetch 8192 -tune 1 64 8 1 -tune 2 64 8 1</cmdline>
  </app_version>
<project_max_concurrent>7</project_max_concurrent>
</app_config>

This is what I have running on my FX-8350 and FX-8370 machines with 1070's

Cheers

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Total success so far. I'm impressed enough to start planning to upgrade to Ryzen across all my crunchers. The improvement is massive relative to total upgrade costs.

That is a very good news I feel more and more confident about my new built. As for app_info and app_config, Lunatics and Raistmer... I will have a look into that and put Linux on ice again. The only things I have done so far is an app_config file using the ''utility to run multiple GPU work units'' so I could run 2 tasks per GPU. I did not know I could do more than that. I wish there was a step by step guide named ''How to optimize your 1070 for newbies''....

Thank you so much Keith I will try that for sure ! It would have taken me years to come out with this ''tuning line''.
I go and try it right now. cheers.

Thank you so much Keith I will try that for sure ! It would have taken me years to come out with this ''tuning line''.
I go and try it right now. cheers.

When all else fails ... RTFM ;-}

Client Configuration

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Yes you are absolutely right. The infos are there somewhere to read but with the job, the business, the wife kids and dogs... I wish I had more times.
I am now crunching with your app config lines and it is indeed faster. Love it !

Glad that worked for you as a drag 'n drop into your app_config. If you are experiencing any keyboard input lag, you will have to lessen the aggressiveness of some of the parameters. Look at the ReadMe_MultiBeam_OpenCL_NV_SoG.txt file in your SETI directory for explanations of what each of the parameter tuning keywords does.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

. . Since AMD opted to implement their AVX at half the bit width that Intel use you might want to try the highest SSE version that is available for a comparison. Not sure of that is SSSE3.0 or SSE4.1. I know I would like to know which crunches faster on the R7.
. . please keep us informed.

Stephen
8-}

Unfortunately, we are stuck with SSE3 in the current apps. I've found over the past couple of years that the SSE3 app is not as fast as the AVX version. I miss Joe Segur's efforts.

. . Using Linux there is an SSSE3 app available and I have got to tell you, it knocks about half an hour off the runtimes compared to SSE3 on my Core2 Duo. I am also sad about Joe's departure and the hole his absence has left in the app development. But everything changes, and he has left a great legacy for the project. I know that on Intel SSE3 is much slower than AVX, but from what I have read about the AVX implementation on the Ryzen units, it would be interesting to see just what the runtimes are using each of those options. Just for the record like.

Stephen

:)

I had some time to play with Linux today while waiting for corsair AM4 bracket. Mint is installed and updated. BOINC is installed and I have ran 2 WU using CUDA60. When I look at ''computers on this account'' I see the same machine twice. One on WIN10 and the other one on Linux. Next step is to install system monitoring tools and then I might make a permanent switch to Mint. I have to say that it feels light and fast. On Windows 10 I am using all CPU and GPU meter as it is very visual so you know right away if your machine is crunching or not. Not quit as easy on Linux it seems. Once I have accomplished that I will try to install and run the special CUDA apps... But for now I am still learning all the //$$ sudo +++ /et/etc/etc/etc $$$ needed to use the terminal. Lots of work ahead. Learning Linux is not easy.

. . Amen to that brother. If you find any really good ( especially graph based ) monitoring apps for Linux please let me know. In the meantime you have the Nvidia Xserver that is provided with the driver package that gives you some basic fan control for cooling and some info on what is happening but it is not very dynamic nor graph based. For that you have psensor which is very basic but does give you a dynamic graph based display.

. . To use the fan controls in Nvidia xserver the command is "sudo nvidia-xconfig --thermal-configuration-check --cool-bits='n' --enable-all-gpus". The last part is if you have mulitple GPUs and 'n'=4 to enable fan control or 28 to give overclocking ability, but that is not my bag :)

. . I have to say I myself feel blind and a little lost running Linux. But with the patience of the existing users and some perseverance I am getting there.

Stephen

?

Talking about your crunchers I am looking at them right now and I am amaze with your RAC across various projects. You roughly have twice the crunching power I have but
your RAC is 4 times higher here at seti alone. 10 times higher RAC if we had the other projects that you are active !!! Ok I am not crunching 24/7 but still. It took me 3 days of full
throttle to go from 32,000 to 40,000 RAC. What is it that I am doing wrong ? I am running the opencl_nvidia_SoG (Boinc choose this app for me) maybe its the ''problem'' ?
Looking at your results I see ''anonymous platform'' as my results says ''windows_intelx86'' ? RAC is not important for me but it sure speaks for itself. Something is not optimize
here for sure but being new to boinc I dont know what. Perhaps I could increase my productivity without having to move to Linux ?
Thanks.

. . One word ... Lunatics. Then tweak the heck out of your chosen app (preferably SoG on a windows system) :)

Stephen

.

. . Since AMD opted to implement their AVX at half the bit width that Intel use you might want to try the highest SSE version that is available for a comparison. Not sure of that is SSSE3.0 or SSE4.1. I know I would like to know which crunches faster on the R7.
. . please keep us informed.

Stephen
8-}

Unfortunately, we are stuck with SSE3 in the current apps. I've found over the past couple of years that the SSE3 app is not as fast as the AVX version. I miss Joe Segur's efforts.

. . Using Linux there is an SSSE3 app available and I have got to tell you, it knocks about half an hour off the runtimes compared to SSE3 on my Core2 Duo. I am also sad about Joe's departure and the hole his absence has left in the app development. But everything changes, and he has left a great legacy for the project. I know that on Intel SSE3 is much slower than AVX, but from what I have read about the AVX implementation on the Ryzen units, it would be interesting to see just what the runtimes are using each of those options. Just for the record like.

Stephen

:)

Back in SETI V7 days, we had a choice to use the SSSE3 or SSE4.1 option for AMD CPU in the Lunatics installer. The SSE4.1 ran much faster on my FX's compared to SSE3 or SSSE3. I would assume that to be likely too of Ryzen. I see that all R7 CPU models show up in the Statistics page now. No way to know whether they are running on Windows or LInux though. Yes, it would be interesting to know how they run under Linux.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Back in SETI V7 days, we had a choice to use the SSSE3 or SSE4.1 option for AMD CPU in the Lunatics installer. The SSE4.1 ran much faster on my FX's compared to SSE3 or SSSE3. I would assume that to be likely too of Ryzen. I see that all R7 CPU models show up in the Statistics page now. No way to know whether they are running on Windows or LInux though. Yes, it would be interesting to know how they run under Linux.

Way back when I made the windows build under v7, the difference on Core2 45nM having SSE4.1, was small (a couple of percent) over ssse3, though consistent and reproducible on the core2Duos of the time. I never managed to afford an AVX capable system yet myself for development. Since the adoption of AVX, and Intel compiler becoming unusable, the situation changed a fair bit in that time (close to 10 years overall since I started), and Joe Segur hand coded AVX into stock. He's disappeared since then, and no-one has been able to precisely tell me what's happened, so we're probably looking at finding or making new CPU app resources from scratch, more or less.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Back in SETI V7 days, we had a choice to use the SSSE3 or SSE4.1 option for AMD CPU in the Lunatics installer. The SSE4.1 ran much faster on my FX's compared to SSE3 or SSSE3. I would assume that to be likely too of Ryzen. I see that all R7 CPU models show up in the Statistics page now. No way to know whether they are running on Windows or LInux though. Yes, it would be interesting to know how they run under Linux.

Those apps were hand optimised by Joe Segur.
Thats the reason they worked so well.

---

With each crime and every kindness we birth our future.

Back in SETI V7 days, we had a choice to use the SSSE3 or SSE4.1 option for AMD CPU in the Lunatics installer. The SSE4.1 ran much faster on my FX's compared to SSE3 or SSSE3. I would assume that to be likely too of Ryzen. I see that all R7 CPU models show up in the Statistics page now. No way to know whether they are running on Windows or LInux though. Yes, it would be interesting to know how they run under Linux.

. . Since the Core2 Duo supports both SSSE3 and SSE4.1 I have often wished those versions were available. Now I know that SSSE3.0 is a heck of a lot better than SSE3.0 so I can only imagine how much better SSE4.1 must be. A shame I will probably never get to find out. But I look forward to playing with those options when I eventually get a Ryzen, since Linux crunching apps include both the SSSE3.0 and AVX versions.

[edit] having read Jason's comment that SSE4.1 had only a small though consistent gain over SSSE3.0 I can feel better than what I am seeing is about as good as it gets with this processor.

Stephen

.