Message boards :
Number crunching :
AMD's Upcoming Ryzen R9 "Threadripper" 16-core, 32-thread CPU
Message board moderation
Author | Message |
---|---|
ZenMercenary Send message Joined: 16 Mar 17 Posts: 15 Credit: 3,924,868 RAC: 0 |
Thanks to my new friend, Keith Myers, I have gotten my Ryzen 7 1800X build very well dialed-in and I am very excited about the increased contribution I've been able to make to this project as a result. The Ryzen 7-series CPUs really do a fantastic job here on SETI@Home. Earlier this week, however, AMD announced an even beefier consumer CPU on the X399 platform which they refer to as the "Threadripper" series and I believe will be called Ryzen 9. These CPUs have more cores than the Ryzen 7-series, headlined by a 16-core, 32-thread behemoth of a CPU that will likely retail in the $1,000-range. http://www.pcworld.com/article/3197147/components-processors/its-official-amds-threadripper-will-bring-a-16-core-32-thread-monster-to-the-desktop.html Does anyone have any plans to pursue this CPU next month when it starts shipping? I'm not sure if my wife would let me drop another $1,500 on another CPU and motherboard but I would love to live vicariously through somebody else :-) Maybe @RueiKe ? Please? Haha |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I saw the news about the new R9 processors too. I am thinking I might let the new motherboards and processors mature a bit before diving in again for my other two , older, FX system upgrades. I think I've had enough living on the bleeding edge and being an unpaid beta tester for the while. I have been so impressed and happy with the performance of my 1700x R7, that I think that the R9 processors can only hit it out of the park into the next county. So an R9 will definitely be on my short list. I've never had any system with quad channel memory before. That will be interesting to play around with too. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
DVDL Send message Joined: 5 Jul 12 Posts: 4 Credit: 1,893,682 RAC: 0 |
Why not just buy better GPU's? It so much more efficient right? |
Ghia Send message Joined: 7 Feb 17 Posts: 238 Credit: 28,911,438 RAC: 50 |
Why not just buy better GPU's? It so much more efficient right? Why not both... ? Well probably see to what extent the rumours are true at Computex very soon. The Threadripper series will come with an all-new HEDT platform. I'm not going to line up as guinea pig for that, seeing that the Ryzen 7 processors have been having their own problems and weaknesses. But if AMD can pull it off, and at a much lower price point than Intel's upcoming offerings (Kaby Lake-X and Skylake-X), we'll have a war for the mainstream processor throne. Since the market for consumer desktop computing power is mainly driven by gamers, I guess that field is where the outcome will be decided. And right now, very few games benefit from more cores/threads...but the cruncher community would probably jump on Threadripper ;-) Humans may rule the world...but bacteria run it... |
ZenMercenary Send message Joined: 16 Mar 17 Posts: 15 Credit: 3,924,868 RAC: 0 |
Why not just buy better GPU's? It so much more efficient right? I've only been doing this for a couple months, but in my experience the 1800X at 16 threads can complete *roughly* the same average number of tasks per-hour as both of my GTX 1070's combined. It's a little less because I allocate one thread for each of my four GPU tasks, but it's fairly close. I could upgrade to the GTX 1080Ti but for the same $1,500 investment (for two of them) I'm not sure they are superior enough to the 1070 for SETI to be worthwhile. I can be relatively sure, though, that $1,500 invested in a Ryzen 9-series would roughly double the amount of CPU tasks. That's where my head is at currently. But like Keith says above, I'm not sure I'm ready to sink a bunch of money into either right now. I might wait for the new AMD Vega and the new Nvidia Volta GPUs and just build an entirely new system in a year or so. I do think the ~70,000 RAC I'm achieving right now with an 1800X and two 1070's would be in the 100,000 range just with changing out my 1800X with the new 16-core 32-thread R9. Should be fairly substantial. |
ZenMercenary Send message Joined: 16 Mar 17 Posts: 15 Credit: 3,924,868 RAC: 0 |
Why not just buy better GPU's? It so much more efficient right? I'm far from an AMD fanboy (I'll buy an Intel CPU tomorrow if it makes sense) but I had been wanting to get into SETI@Home with a good rig for a long time, and objectively the Ryzen 7-series is just a great, great CPU for this purpose. I was actually looking into an i7-6950x for my build and I was seriously considering spending $1,650 on that CPU. When I heard about Ryzen around February, it blew my mind. Computing performance that's within a stone's throw of a 6950x for 1/3 of the price. I found out if I could just wait a couple more months, I could slash more than $1,000 off the cost of my build and still have a computer in the top 100 here on SETI@Home. I invested some of that saved money into dual 1070 GPUs and here we are. Now, don't get me wrong. I play games on this computer, too. Not heavily (maybe 3-4 hours per month), but enough that gaming performance at least needs to be sufficient. The rumor I've heard, though, is that Threadripper might be a better gamer than the 7-series just because they were able to fix some of the issues that initially plagued the R7's. That's just a rumor, of course. One thing is fairly certain: with the CPU industry shifting toward more cores, and the GPU industry beefing up for things like VR and AI, these BOINC projects should see enormous exponential gains over the next five years. We are approaching an entirely new era of computing, I think. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Why not just buy better GPU's? It so much more efficient right? Really depends on your definition of efficient, I guess. I've seen more improvement in RAC and power consumption from the move to Ryzen than the change from GTX 970's to GTX 1070's. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
I've seen more improvement in RAC and power consumption from the move to Ryzen than the change from GTX 970's to GTX 1070's. Yes, but you did move from Bulldozer based CPUs. They really were very, very poor when it came to power usage & performance. Grant Darwin NT |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
Earlier this week, however, AMD announced an even beefier consumer CPU on the X399 platform which they refer to as the "Threadripper" series and I believe will be called Ryzen 9. These CPUs have more cores than the Ryzen 7-series, headlined by a 16-core, 32-thread behemoth of a CPU that will likely retail in the $1,000-range. Sounds like a Naples CPU that didn't make it through validation. Shut down one half & you've got a super heavy weight desktop CPU. Grant Darwin NT |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
One thing is fairly certain: with the CPU industry shifting toward more cores, and the GPU industry beefing up for things like VR and AI, these BOINC projects should see enormous exponential gains over the next five years. We are approaching an entirely new era of computing, I think.As one among a few of us who have been blazing a trail in high core count crunching over the last year or 2, my thought on the gains in crunching capabilites are mixed. On one hand, it's great to have all this potential power out here to process massive numbers of tasks, but on the other hand, does the infrastruture have the ability to handle it? As is happening as I write this, we're expiencing another server issue, and one of my rigs is getting tasks only if I restart BOINC, another I noticed has nothing but a shrinking cache of GPU tasks, while 50 some cores sit idle. I'm not ripping on the project, indeed I am a huge supporter of it, but I just wonder of it's ability to handle the amount of tasks that could potentially be processed by ever more capable crunchers in the next year or 2, when it appears to be struggling with the current demands of existing hardware and technology. I know that they can aleviate this to some degree by just decreasing the number of tasks available to each user, but that could have a demoralizing effect on the most dedicated group of users. And yes, I understand that we do this of our own volition, and that SETI doesn't 'owe' us a certain quantity or even any tasks, but we are all here for the science, and (hopefully) the eventual WOW moment. At least that is why I am here, and why I care, and why I invest my resources as I have. So hopefully in the coming months (years? Gads!) this ship will be righted, and will be able to somehow handle all the additional capablities that advancing hardware (and software - thanks Petri!) will bring to the table. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
On one hand, it's great to have all this potential power out here to process massive numbers of tasks, but on the other hand, does the infrastruture have the ability to handle it? That was the question I asked when Eric first posted about needing more people to crunch the expected flood of Breakthrough Listen data. Considering that back in December the application preferences became broken & still haven't been fixed, and we have frequent issues with the web server and forums, and less frequent but still occurring random outages (such as the one that just finished) the answer at present is No. Seti needs more money for the staff and hardware needed just to meet present demands, let alone future ones. Grant Darwin NT |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
There are several ways to look at the servers. - They do have the potential to process large amounts of data (i.e 24 hours worth in the 12 hours after maintenance) so the potential is there after their weekly backups and database compression. - The fast computers would definitely pose the highest server load. - The slow computers definitely cause large database size. - The 8 week timeouts cause large database size. As I see it the servers would have no load problems at all if it were not for the large databases, but has been stated many times, changes to cache size or timeouts is not going to change. So we/they live with it. A possible solution may be a quick Friday morning database compression to improve performance between maintenance. All in all I think the servers do quite well with what we put them through, and asking for programing changes to a system that is getting by isn't going to happen with the planned restructuring and/or shutdown of the project in it's foreseeable future. OK, enough for off topic thoughts :) |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13751 Credit: 208,696,464 RAC: 304 |
A possible solution may be a quick Friday morning database compression to improve performance between maintenance. 2* 6-8 hour outages instead of the 1* 10-12hour weekly one? Still wouldn't have any effect on all the other random outages. An all flash storage system driven by appropriate Quad or Octa socket servers would reduce the weekly outage to under an hour. Anyone got $300,000+ to spare? Grant Darwin NT |
MarkJ Send message Joined: 17 Feb 08 Posts: 1139 Credit: 80,854,192 RAC: 5 |
For a no-cost solution reducing the deadline would help reduce the database bloat that needs frequent compression? While it may have an effect on the slower rigs they could run a smaller cache so they can meet the deadline. Currently over at Einstein they're doing a Gravity wave tuning run with a 5 day deadline. Now that's obviously not practical for normal use but it certainly clears the tasks a lot quicker. Maybe 10 to 21 days is a better number to use. BOINC blog |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
The problem with Gravity waves on the CPU is they are using almost 2 GB worth of Memory. 12 core would use more than 24 GB of memory (there seems to be some other activity going on that results in more than the 2 GB) Plus they want to use all the cores so it necessitates the use of an app_config.xml to reduce the number of cores being utilized. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
Earlier this week, however, AMD announced an even beefier consumer CPU on the X399 platform which they refer to as the "Threadripper" series and I believe will be called Ryzen 9. These CPUs have more cores than the Ryzen 7-series, headlined by a 16-core, 32-thread behemoth of a CPU that will likely retail in the $1,000-range. That was my impression as well. Naples is 32c/64t, and they said that it is a 409..4? pin socket, which is pretty obviously HUGE, and they wouldn't say what the socket would be named yet, but I suspect it is the same socket as Naples. One of the things AMD insists on though is that "Threadripper" (internal code name was "Whitehaven") will not interfere with the server market, whereas Intel's upcoming i9 (a 12-core Skylake, from what I read) is going to eat into the Xeon market. [source for this paragraph is: pcworld ] But yeah, it seems like Threadripper/Whitehaven is basically Naples CPUs that didn't pass the QC checks.. so turn half of it off, double the GHz, and you end up back at the same TDP as it was before when it was 32c and 1.7ghz. It sounds a lot like those Phenom 3-core CPUs, except hopefully they did actually sever the connection to the bad half so that it isn't still drawing power but never getting instructions. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
I will definitely get one as soon as I can get my hands on it. I hope the MB are in much better shape. I am still waiting on a bug fix from ASUS for C6H that is impacting my current Ryzen system. Thanks to my new friend, Keith Myers, I have gotten my Ryzen 7 1800X build very well dialed-in and I am very excited about the increased contribution I've been able to make to this project as a result. The Ryzen 7-series CPUs really do a fantastic job here on SETI@Home. GitHub: Ricks-Lab Instagram: ricks_labs |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
I will definitely get one as soon as I can get my hands on it. I hope the MB are in much better shape. I am still waiting on a bug fix from ASUS for C6H that is impacting my current Ryzen system. You give the new AGESA 1.0.0.6 Beta BIOS', 9943 and 9945 a try yet? Seems like there was lots of bug fixes in them, but mostly memory improvements. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
RueiKe Send message Joined: 14 Feb 16 Posts: 492 Credit: 378,512,430 RAC: 785 |
I will definitely get one as soon as I can get my hands on it. I hope the MB are in much better shape. I am still waiting on a bug fix from ASUS for C6H that is impacting my current Ryzen system. Just researching it now. Perhaps this evening I will give it a try. I hope they fixed the bug of the system forgetting the PCIe x4/x1 mode... GitHub: Ricks-Lab Instagram: ricks_labs |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
In all the bugs talked about and fixed ...... I never saw that mentioned in any of the threads. Crossing my fingers for you. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
©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.