Message boards :
Number crunching :
Come on admit it, you have an E5-2670 crunching on Boinc and/or Seti@Home. How is it doing?
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next
Author | Message |
---|---|
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
First of all the file name MUST be app_config.xml in plain text format to control the number of tasks to run. Also, did you select Astro Pulse in the Lunatics installer? If so you will have entries in app_info.xml for CPU/GPU astropulse applications. No need to edit, just check if they are there. |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
Yes, name is correct, corrected my post. It worked before this way, it also folds AP units and MB units, but waits till all MB units are done, i do have to set SETI to send no work so the cache is empty over time and then it starts to fold the AP units. Before it folded them at the same time, reserving a slot or something. So i had MB and AP folding at the same time. ( And i want it back again :P ) ahhh i let it fold the cache empty & reinstall. Edit the config file again and see what happens, maybe something somewhere somehow went wrong when i just ran the new latest build boinc installer over it. |
rob smith Send message Joined: 7 Mar 03 Posts: 22190 Credit: 416,307,556 RAC: 380 |
BOINC controls the processing sequence of tasks on a First in First out basis. AP tasks are very rare so we only get one or two at a time and you will have to wait for them to appear, then to be processed. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
I know that, the availability of AP units is not the problem. The problem is, i was folding 1 AP ( when available ) and 1 MB at the same time on my GPU. It's a 1080Ti it can do that easy. Now, since i installed the latest BOINC client it did not work anymore this way, which means, it first folds all MB units, then after its done AP units. I reinstalled all yesterday, same config file, see what happens. |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
from the config you posted it should be doing 3 tasks at a time (AP and MB mixture), is it not doing that? |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
2 questions I have, Is the computer waiting until there are no MB work units available to crunch then starting the AP? If so, then it sounds like the percentage of GPU is wrong somewhere as it's waiting for a full GPU availability before crunching the AP. If it not doing that, then have you been monitoring where in the cache the AP are? Boinc should do a first in, first out approach, so it you download 200 MBs and 1 AP, the AP will wait it's turn in the line before it crunches. The AP will not bypass all those other MBs in order to be crunched first. Let us know Z |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
2 questions I have, Is the computer waiting until there are no MB work units available to crunch then starting the AP? -----> Yes, if MB cache on my pc is empty, then it starts the AP if they are available. If it not doing that, then have you been monitoring where in the cache the AP are? -----> You mean the cache folder? They are in the same folder where the MB units are located Boinc should do a first in, first out approach, so it you download 200 MBs and 1 AP, the AP will wait it's turn in the line before it crunches. The AP will not bypass all those other MBs in order to be crunched first. I think that is happening now, i dont have any AP now, but will keep an eye out. But i saw previously AP units getting downloaded later on and folded next to the MB units at the same time, ( folding 2 MB at the same time, 1 was faster done then the other, then a AP unit was folded instead, not the next in line MB unit ) they did not sit in a waiting line. Maybe it sounds strange, sorry for that, maybe its just my computer :P, it all started when upgrading to the latest BOINC build. Yesterday i did a complete reinstall of BOINC latest build, ran Lunatics beta6 installer, configure the appinfo file for version r3584 SoG app, and created the appconfig file with above standing lines. Seems good now, but no AP yet, i will report later if i can with more details. |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
I have a feeling you have a (some) non-printable character in your app_config which can be caused by copy/pasting from the forum, etc. This could cause the astropulse section to be ignored and it is trying to run at the default <ngpus>1.00</ngpus> that would cause them not to run until the MB cache is empty (0.66 would be in use by MB), or the AP task changes to high priority because of the deadline. It wouldn't start because the would be no room if MB tasks are running. If it continues, try removing/re-creating (carefully) every space and carriage return in the file. Non-prinable stuff can look like a space, but can screw up the app when trying to read plain text. |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
I have a feeling you have a (some) non-printable character in your app_config which can be caused by copy/pasting from the forum, etc. Ok thanks, i did a copy paste from the forum indeed, it maybe was not smart. I will re-create the file and see what happens. |
Tom M Send message Joined: 28 Nov 02 Posts: 5124 Credit: 276,046,078 RAC: 462 |
This morning, I just noticed my E5 box wasn't running. My only excuse is I have been "busy". It looks like it may have been off for a month. :( The good news is it looks like only 5 tasks are past deadline. Unfortunately one of them is an AP. I don't have a clue why it shutdown. It is on a UPS and has a fan blowing in the front.... Gremlin's? Nah.... maybe it got lonely (see below). Oh, well. Teach me to assume all is well in the "server" room and not wonder in there daily. obtw, it was getting too hot in my home office and I needed to study, so I "moved" all my "desk" stuff to a "study nook" in another room. So I don't commonly get in here much. At least my little box kept running :) Tom A proud member of the OFA (Old Farts Association). |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
Well, after some testing and tuning, i have the system running 1 WU on GPU at a time, FiFo and 14 CPU cores ( AVX ) HT disabled. This results for my dual Xeon system the fastest crunching times per watt till now. Leaving this running for a time. On GPU i use these parameters: <app_config> <app_version> <app_name>setiathome_v8</app_name> <plan_class>opencl_nvidia_SoG</plan_class> <avg_ncpus>1</avg_ncpus> <ngpus>1</ngpus> <cmdline>-sbs 1024 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 512 -oclfft_tune_ls 512 -oclfft_tune_bn 256 -oclfft_tune_cw 256 -pref_wg_num_per_cu 12 -period_iterations_num 1 -hp -high_prec_timer -high_perf -tt 1500</cmdline> </app_version> <app_version> <app_name>astropulse_v7</app_name> <plan_class>opencl_nvidia_100</plan_class> <avg_ncpus>1</avg_ncpus> <ngpus>1</ngpus> <cmdline>-unroll 28 -oclFFT_plan 256 16 256 -ffa_block 16384 -ffa_block_fetch 8192 -tune 1 64 4 1 -tune 1 64 8 1 -tune 2 64 8 1 -hp</cmdline> </app_version> </app_config> |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
Two items of interest: 1) E5-2680 v2 (10 core, 20 thread) are now under $200 on ebay 2) E5-2670 v1 are > $100. So you can upgrade your dualie cruncher for around $200. The v2s are 100mhz faster than the v1s and will run 8 more threads and at the same wattage. Just a headsup, guys! |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
Thats interesting, gonna check it out today :) |
Marco Vandebergh ( SETI orphan ) Send message Joined: 27 Aug 10 Posts: 39 Credit: 12,630,994 RAC: 9 |
A little update from my side, ( if people are interested ) I have been playing around with this parameter: -pref_wg_num_per_cu First i left it folding for a while with -pref_wg_num_per_cu 1, after changing it to 4 my 1080Ti is folding a single guppy task from 4,05 minutes to approx 2,45 minutes. Setting this too high will result in crashes of display driver. Maybe worthwile a try for the experimental guys here. My current app_config.cfg: <app_config> <app_version> <app_name>setiathome_v8</app_name> <plan_class>opencl_nvidia_SoG</plan_class> <avg_ncpus>1</avg_ncpus> <ngpus>1</ngpus> <cmdline>-sbs 2048 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 512 -oclfft_tune_ls 512 -oclfft_tune_bn 256 -oclfft_tune_cw 256 -pref_wg_num_per_cu 4 -period_iterations_num 1 -hp -high_prec_timer -high_perf -tt 1500</cmdline> </app_version> <app_version> <app_name>astropulse_v7</app_name> <plan_class>opencl_nvidia_100</plan_class> <avg_ncpus>1</avg_ncpus> <ngpus>1</ngpus> <cmdline>-sbs 2048 -unroll 28 -oclFFT_plan 256 16 256 -ffa_block 16384 -ffa_block_fetch 8192 -tune 1 64 4 1 -tune 1 64 8 1 -tune 2 64 8 1 -hp</cmdline> </app_version> </app_config> |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
Two items of interest: I am now crunching with 2680 v2s, having bought 4 for $658 shipped on eBay. They run as expected, except that I am actually running about 20 watts less/2680 v2 vs 2670s (90-95 watts on 20 T vs the previous 110-115 watts on 16 T). However, this last is no doubt due to the completely BLC05 nature of the current workload, as they seem less "dense" computationally in some sense than the Arecibo WUs of yore. Just sayin' |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Hope all you Intel workstation motherboard users can help out a Intel newbie. I have no experience with any Intel processor since about 1990. I have a ASUS X99-E-10G WS motherboard that I acquired for free. It seems it is compatible only with Xeon E5 16xx/26xx V3/V4 processors. I have been told by eBay vendors that the motherboard is not compatible with either qualifying or engineering sample used/pulled/refurbished processors. So I have been scouring eBay for any V3/V4 used/pulled retail part and have not come up with any reasonable cost solution. I had read this thread back in the beginning and was amazed at some of the prices quoted for E5-26xx V2 processors. One was mentioned at $150. That is the kind of price range $150-$250 I would be willing to spend on eBay. But I have not found anything of that kind for any V3/V4 part. Is there that much difference between V2 and V3/V4? Are there any good eBay deals on Intel processors of this recent vintage? Hoping somebody can clue me in. TIA. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Kiska Send message Joined: 31 Mar 12 Posts: 302 Credit: 3,067,762 RAC: 0 |
Hope all you Intel workstation motherboard users can help out a Intel newbie. I have no experience with any Intel processor since about 1990. I have a ASUS X99-E-10G WS motherboard that I acquired for free. It seems it is compatible only with Xeon E5 16xx/26xx V3/V4 processors. V2 parts only support DDR3, V3/V4 parts support DDR4 so have fun finding some reasonably priced DDR4 modules. In regards to other differences E5-2670 V3: https://ark.intel.com/products/81709/Intel-Xeon-Processor-E5-2670-v3-30M-Cache-2_30-GHz Has 12 cores/24 threads E5-2670 V2: https://ark.intel.com/products/75275/Intel-Xeon-Processor-E5-2670-v2-25M-Cache-2_50-GHz Has 10 cores/20 threads Also with V2 parts only AVX is available on them, but with Haswell(V3/V4) parts AVX2 and FMA3 is available In terms of pricing for 2670 v3's you won't find them below $1k as most business haven't reached their End of Life/Upgrade cycle yet, only reason why v2's are cheap now is companies have upgraded to v3's or v4's. I would high recommend looking at the product stack and see if you can find some for cheap. Otherwise keep the board laying about for a year or 2, and by then these CPUs are gonna go on a heavy bulk discount |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Thanks for the reply. I was developing a suspicion that there would be no cheap solution. I am bidding on DDR4 RAM 4X4GB kits and the prices for 2133-2400 aren't too unreasonable for 4 sticks. At least compared to what I paid for Samsung B-Die DDR memory for my Ryzens. Don't need huge server quantities of RAM, this is just going to be a cruncher. My idea was to get at least a 8 core HT part so it could support 3 or 4 graphics cards. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Brent Norman Send message Joined: 1 Dec 99 Posts: 2786 Credit: 685,657,289 RAC: 835 |
As said v2 is DDR3, LGA2011R1, v3 is DDR4, LGA2011v3 - they are for different boards. One great resource is to look up a chip spec on the Intel site, then click on Product Family, which will give you a list of chips to go searching for. Intel-Xeon-Processor-E5-v3-Family So you can chose for slow and more threads, or fast and fewer threads .. or both for big $$ :D |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Yes, I have the ASUS CPU compatibility page bookmarked. Isn't available for download for some reason. I have been searching far and wide on eBay. But 99% of the V3/4 E5 chips are all Engineering sample ES chips. One of the vendors that I was interested in said the MB could only reasonable handle the retail chip or at minimum a QS chip with stepping of B0,M0 or R0. Anything like a L0 or step one ES chip would boot loop on Error 19 since the firmware can't handle the early chips. I have my heart set on minimum of 16 threads, spoiled with my Ryzens that have colored my perception I guess. The board supports 4 PCIeX16 double wide cards so it has lots of potential. Just have to find a chip that is compatible with the motherboard and my wallet ;-} This is my first Intel board in 20 years, so lots of learning to do. 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.