From FX to Ryzen

Message boards : Number crunching : From FX to Ryzen
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 14 · Next

AuthorMessage
Profile RueiKe Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 492
Credit: 378,512,430
RAC: 785
Taiwan
Message 1854993 - Posted: 12 Mar 2017, 8:45:04 UTC - in response to Message 1854893.  

I just posted a video on my experience with the R7 1700 so far. It includes benchmarks with the FX8370, i73930K, and i7-6950X. I have included the AVX MB app in my benchmarks. https://youtu.be/YpEOmxG3zV4

It looks like normal AR Arecibo tasks are running ~2hrs with the AVX app.
Is that at the 3.5GHz OC you mentioned?

I would have hoped for run times in the 1-1.5hr range at that speed. Given my dual E5-2670 system @ 3.0GHz takes ~2.5hrs when running 32 at a time.


The system is currently OC'ed to 3.5GHz. I took a look at the last 60 CPU tasks on both the R7-1700 and i7-6950X. Here is a summary:
           Avg Time    Avg Credit    Credit/s
i7-6950X    5732s       104.11        0.01816
R7-1700     5251s        73.39        0.01397

Unfortunately, they have a different mix of WU, so it is really hard to say how they compare from this data. As I posted earlier, running the same WU, the 1700 is faster, but fully loading all threads may be different.
GitHub: Ricks-Lab
Instagram: ricks_labs
ID: 1854993 · Report as offensive
Profile RueiKe Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 492
Credit: 378,512,430
RAC: 785
Taiwan
Message 1854995 - Posted: 12 Mar 2017, 8:49:04 UTC - in response to Message 1854982.  

Hi Rick, yes I watched your video and the numbers you came up for scaling looks mighty fine for Ryzen 1700. Are you still feeling out the system or have you started to tear down the 7-way block for your ATI cards? Curious how that turns out and whether the 1700 and CH6 board solves your bottlenecking on the GPU's.


I am going to make sure I have stable BIOS before I tear down my current system. I will be away from my computers next week, so it will be 2 weekends from now till I can tear down the current system. I need to give myself a whole weekend for that work. I am also hoping that it solves the current issue on my Penta-Nano setup, though my triple ProDuo system on the FX-8370 doesn't have the same issue. So not really sure what is going on with the current Penta-Nano system.
GitHub: Ricks-Lab
Instagram: ricks_labs
ID: 1854995 · Report as offensive
Profile RueiKe Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 492
Credit: 378,512,430
RAC: 785
Taiwan
Message 1855000 - Posted: 12 Mar 2017, 9:23:14 UTC

I posted a new video in which I show that PCIe 2.0x1 performs just as well as PCIe3.0x16 for MB tasks in my new RYZEN build. With this confirmation, my plans are on track to move my Penta-Nano GPUs to the new build. Link to video: https://youtu.be/lciBpXCE-60
GitHub: Ricks-Lab
Instagram: ricks_labs
ID: 1855000 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1855031 - Posted: 12 Mar 2017, 11:56:27 UTC - in response to Message 1854993.  

I just posted a video on my experience with the R7 1700 so far. It includes benchmarks with the FX8370, i73930K, and i7-6950X. I have included the AVX MB app in my benchmarks. https://youtu.be/YpEOmxG3zV4

It looks like normal AR Arecibo tasks are running ~2hrs with the AVX app.
Is that at the 3.5GHz OC you mentioned?

I would have hoped for run times in the 1-1.5hr range at that speed. Given my dual E5-2670 system @ 3.0GHz takes ~2.5hrs when running 32 at a time.


The system is currently OC'ed to 3.5GHz. I took a look at the last 60 CPU tasks on both the R7-1700 and i7-6950X. Here is a summary:
           Avg Time    Avg Credit    Credit/s
i7-6950X    5732s       104.11        0.01816
R7-1700     5251s        73.39        0.01397

Unfortunately, they have a different mix of WU, so it is really hard to say how they compare from this data. As I posted earlier, running the same WU, the 1700 is faster, but fully loading all threads may be different.


Um, so a ~$350 8core/16 thread 65 Watt part, achieving ~70% of a ~$1700 10 core/20 thread, 140Watt part ?
"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.
ID: 1855031 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855075 - Posted: 12 Mar 2017, 16:32:20 UTC - in response to Message 1855000.  

I posted a new video in which I show that PCIe 2.0x1 performs just as well as PCIe3.0x16 for MB tasks in my new RYZEN build. With this confirmation, my plans are on track to move my Penta-Nano GPUs to the new build. Link to video: https://youtu.be/lciBpXCE-60

Thanks again for a good video, Rick. The testing produced the results I expected. Good thermals and no impact at all running MB GPU tasks in PCIex1 or PCIex4 compared to PCIex8.

I know that you will likely wait for the "official" BIOS update from the "official" ASUS website. FYI, the "beta" 0902 BIOS that had the link published in the forums by "Elmor" is almost assuredly going to be the same 0902 BIOS published on the ASUS site because "Elmor" works for ASUS on their BIOS team and has been debugging the existing shipping BIOS and the BIOS updates on the ASUS site.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855075 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855077 - Posted: 12 Mar 2017, 16:35:51 UTC - in response to Message 1855031.  


Um, so a ~$350 8core/16 thread 65 Watt part, achieving ~70% of a ~$1700 10 core/20 thread, 140Watt part ?

Um, yeah. Nice, new and modern CPU design at 1/5 the cost of a similar Intel part and runs at half the power budget and has 70% of the productivity of the Intel equivalent part.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855077 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1855254 - Posted: 13 Mar 2017, 10:42:40 UTC - in response to Message 1853351.  


That would allow you to run Petri's Special Sauce CUDA app which smokes every other SETI app published so far. That definitely would increase your productivity. You'll find Petri's computers and many others in the top of the Top 100 Computers in the Statistics pages here on the website running his app.

A lot of us are waiting for Jason's fine tunings for Windows to appear.

Cheers.

Do you know if this is happening or is that just wishful thinking?


. . Jason has posted that he is looking into it, but I have heard no mention of progress.

Stephen

:(
ID: 1855254 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1855255 - Posted: 13 Mar 2017, 10:47:02 UTC - in response to Message 1853365.  


Do you know if this is happening or is that just wishful thinking?

I believe Jason has been working with the discoveries that Petri, Raistmer, TBar and the other volunteer developers have figured out with the latest CUDA drivers in the Linux camp and is applying that knowledge over in the Windows environment. That may just be my interpretation of reading between the lines of the forum posts. I have no direct knowledge.


. . Jason has said there is no issue in porting it to Windows, but rather in the Pulse find algorithm and he is trying to resolve that. If or when that problem is solved there may be an announcement we would all like to hear, until then I am trying to master Linux :)

Stephen

8^}
ID: 1855255 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1855259 - Posted: 13 Mar 2017, 11:08:59 UTC - in response to Message 1853610.  

The 1700 will arrive tomorrow but the MB is still BO and no sign of the Corsair AM4 bracket.
In the mean time I have started to play around with Linux. I have tried to install it alongside windows 10 with no success so I have installed it on another drive so I have to go to bios to select boot
drive but its OK. Next step is getting the basics and learning how to install programs. The first thing will be system monitoring tools as it appear there are no such things in Mint.


. . I have been unable to find any Linux monitoring tools that match those available for Windows, I am working with Linux myself and I miss my Afterburner et al. There are some tools though such as psensor which go part way to doing the job. As for solving the dual boot issue I found that if you have another drive to dedicate to Linux you can boot from a flash drive. Swapping between OS's is as simple as powering down and inserting or removing the flash drive. Since it is not something I would do more than once or twice a week (probably much less) I find that very workable. You make two partitions on the dedicated HDD, one for /root with your /home directory and the rest using EXT4 as your default format, a swap partition to make Linux happy, and then make the flash drive /boot. Set your BIOS to boot from USB first then whenever the flash drive is plugged in you boot into Linux, with it removed you boot into Windows and the two OS's never meet. Of course in Linux you make sure all your Windows drives are unmounted. In Windows the EXT4 format is meaningless (mind you NTFS is so to Linux as well).

. . I have been advised, and through an unwitting error learned the hard way, to NOT use the version of BOINC from the Linux repository but rather use the version from SETI. TBar has all the references you require in this thread for the app, "Linux CUDA 'Special' App finally available, featuring Low CPU use". Read through there and avoid the pain and mistakes I made.

Stephen

<shrug>
ID: 1855259 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1855260 - Posted: 13 Mar 2017, 11:15:48 UTC - in response to Message 1854033.  

My order is due to be picked tomorrow (10 March) - I'll let you know how I get on.

(One of the reasons for going for a 1700 rather than a 1700X was the inclusion of a stock cooler and the probable shortages of suitable 3rd party coolers for the latter.)


. . That's my take as well.

. . Good choice.

Stephen

:)
ID: 1855260 · Report as offensive
Profile RueiKe Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 14 Feb 16
Posts: 492
Credit: 378,512,430
RAC: 785
Taiwan
Message 1855261 - Posted: 13 Mar 2017, 11:23:38 UTC - in response to Message 1855075.  

I posted a new video in which I show that PCIe 2.0x1 performs just as well as PCIe3.0x16 for MB tasks in my new RYZEN build. With this confirmation, my plans are on track to move my Penta-Nano GPUs to the new build. Link to video: https://youtu.be/lciBpXCE-60

Thanks again for a good video, Rick. The testing produced the results I expected. Good thermals and no impact at all running MB GPU tasks in PCIex1 or PCIex4 compared to PCIex8.

I know that you will likely wait for the "official" BIOS update from the "official" ASUS website. FYI, the "beta" 0902 BIOS that had the link published in the forums by "Elmor" is almost assuredly going to be the same 0902 BIOS published on the ASUS site because "Elmor" works for ASUS on their BIOS team and has been debugging the existing shipping BIOS and the BIOS updates on the ASUS site.


If I was not about to go on a 5 day business trip, I would probably go for it. But since it will sit around until I get back, I won't rush into it.
GitHub: Ricks-Lab
Instagram: ricks_labs
ID: 1855261 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1855404 - Posted: 14 Mar 2017, 0:18:54 UTC - in response to Message 1855255.  


Do you know if this is happening or is that just wishful thinking?

I believe Jason has been working with the discoveries that Petri, Raistmer, TBar and the other volunteer developers have figured out with the latest CUDA drivers in the Linux camp and is applying that knowledge over in the Windows environment. That may just be my interpretation of reading between the lines of the forum posts. I have no direct knowledge.


. . Jason has said there is no issue in porting it to Windows, but rather in the Pulse find algorithm and he is trying to resolve that. If or when that problem is solved there may be an announcement we would all like to hear, until then I am trying to master Linux :)

Stephen

8^}


No update at the moment, as dealing with home issues. That situation is fundamentally correct though. In between other responsibilities, I am accumulating updated development tools and techniques, since the combination of Windows OS/Driver, GPU, and supporting technology changes does represent a 'perfect storm', more or less.

There are signs that situation IS finally starting to settle though. I'd like to dedicate far more time to software development in general than I've been able to, so will probably be considering options for creating some other non seti-specific supporting tools in the near future.

These will ultimately help guide making the hardware specific optimisations available for a more general audience. The limitations with Boinc server work issue itself, not being able to discriminate between mixed Cuda devices on the same host, are problematic from a stock integration perspective. It forces a large scale effort toward dynamic dispatch within the application itself, over targeted builds. While that doesn't affect the advanced user Linux+Mac Scenarios so much, it does rather heavily affect the dominant Windows group in terms of risks to the project.

Once the tools, techniques, and time resources are in place though, things can potentially move fairly quickly. For one thing I have managed to get Hardware passthrough activated on my Mac Pro through Ubuntu, which potentially means a Windows 10, Mac and Linux unified development system. Haven't checked if there is an OSX sierra pascal driver lately, but either way There are still the 680, 780 and 980 on hand should the automated cross platform build environment come together as planned.

The goal there is to automatically build, regression test, cross compare, and deploy successful builds to a website. Definitely that's been a mammoth undertaking to assemble, however it looks slightly closer to becoming a thing each week.
"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.
ID: 1855404 · Report as offensive
Profile betreger Project Donor
Avatar

Send message
Joined: 29 Jun 99
Posts: 11361
Credit: 29,581,041
RAC: 66
United States
Message 1855408 - Posted: 14 Mar 2017, 0:36:50 UTC - in response to Message 1855404.  

The goal there is to automatically build, regression test, cross compare, and deploy successful builds to a website. Definitely that's been a mammoth undertaking to assemble, however it looks slightly closer to becoming a thing each week.

Woooo Hoooooo
ID: 1855408 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855414 - Posted: 14 Mar 2017, 0:58:30 UTC - in response to Message 1855261.  



If I was not about to go on a 5 day business trip, I would probably go for it. But since it will sit around until I get back, I won't rush into it.

Probably smart. Rumor has ASUS pushing the 0902 BIOS release for the CH6 "golden" on their web support site on Wednesday. Should be there when you get back from your business trip.

HIGHLY recommended everyone using a ASUS Crosshair VI motherboard update to the 0902 BIOS to prevent "bricking" their motherboard if the motherboard decides to push the CPU_SOC Voltage above 1.2V in Auto on its own. That will most assuredly kill the EC (Embedded Controller) chip on the motherboard and brick it forcing an RMA.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855414 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855415 - Posted: 14 Mar 2017, 0:59:49 UTC - in response to Message 1855404.  

Outstanding update Jason. Just want to say thanks for being a great volunteer developer for SETI@Home all these years.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855415 · Report as offensive
Profile Verdonim
Avatar

Send message
Joined: 30 Sep 16
Posts: 170
Credit: 34,185,785
RAC: 0
Canada
Message 1855665 - Posted: 15 Mar 2017, 21:42:33 UTC

I have been very busy here did not had much time to follow all the posts but I will be off 4 days this week-end and I will get back into it. Still no news from Corsair bracket and its a joke.
I might change my water CPU cooler for another brand that is ready for AM4. My learning Linux project is also on the ice for now. Cheers
ID: 1855665 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1855673 - Posted: 15 Mar 2017, 23:17:11 UTC - in response to Message 1854224.  


Toms Hardware has been using an IR camera in their reviews for a while.
http://www.tomshardware.com/reviews/amd-ryzen-7-1800x-cpu,4951-11.html


. . I will wait until I am more awake to read that.

. . Still holding off before I leap in

Stephen

:)
ID: 1855673 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855740 - Posted: 16 Mar 2017, 10:44:06 UTC - in response to Message 1855665.  

I have been very busy here did not had much time to follow all the posts but I will be off 4 days this week-end and I will get back into it. Still no news from Corsair bracket and its a joke.
I might change my water CPU cooler for another brand that is ready for AM4. My learning Linux project is also on the ice for now. Cheers

Just got Numbskull put back together for crunching after putting in Ryzen and AM4. What a battle. I decided to not wait around for Corsair and put in a Noctua NH-D15-SE-AM4 air cooler temporarily until the bracket shows up. This thing runs hot, at least its reported temperature. Tctl temp is reporting 80° C. with the socket reporting 46° C. But that is a real temperature of 60° C. because the 1700X (mine) and the 1800X have a +20 degree offset put in by AMD. Air coming off the cooler is barely warm in reality. Still have the wrong memory for it and so far can only get it to run at 2400 Mhz instead of its XMP spec of 3200 Mhz. Have only got the CPU running at 3.775 Ghz so far with the Vcpu @1.36 with LL1. CPU utilization is currently only about 75% because I set the max tasks to 12 in app_config. I might push it all the way once I get a better handle on how it runs. Still very much in observation mode since the platform is so new.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855740 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1855804 - Posted: 16 Mar 2017, 17:49:09 UTC - in response to Message 1855740.  

I have been very busy here did not had much time to follow all the posts but I will be off 4 days this week-end and I will get back into it. Still no news from Corsair bracket and its a joke.
I might change my water CPU cooler for another brand that is ready for AM4. My learning Linux project is also on the ice for now. Cheers

I just got a reply from Corsair on a ticket I input about the AM4 bracket. Was told that I should expect my orders to ship around 3/22. So unhappy that it is taking so long to get them. But at least have some idea about when.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1855804 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22220
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1855832 - Posted: 16 Mar 2017, 21:28:22 UTC
Last modified: 16 Mar 2017, 21:50:12 UTC

Right, that was easy (apart from the motherboard mounting screws which decided to make a break for freedom).
Pulled the Phenom and its motherboard, cleaned all the case fans, replaced one that had died some time ago.
Mounted the Ryzen 7-1700 with its stock cooler, 8GB RAM, connected up the HDD & DVD, PSU (don't forget the 24 way....), plugged a monitor in & turned it on. A message about having a new CPU from the BIOS - well there's a surprise!, reboot, no rude messages, wait a few moments and the Mint boot screen came to life, the the desktop appeared. Launch System monitor - 16 "cores" reported. Resumed SETI, and a load more CPU tasks appeared, and look as if they are running. Time from suspending SETI to resuming - less than 2 hours, no re-installation of the OS, no configuration changes.
What's the system?
Ryzen 7-1700, stock cooler
Asus Prime X370 Pro motherboard
8GB DDR4, 2666MHz Corsair RAM
1 off GTX960 GPU

I'm going to let it run for a few days to see how it settles down before trying any sort of tuning.



Interesting to see how little RAM is being used - with 14 CPU tasks, and 1 GPU task 1.7GB (I'm running Petri's "magic" app). This compares with the 4GB being used on my last Windows PC - which is only running 4 CPU tasks and 4 GPU tasks (SoGs)
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1855832 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 14 · Next

Message boards : Number crunching : From FX to Ryzen


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.