|
Post by ShrimpBrime on Jan 3, 2019 18:14:32 GMT -5
Do we have any threads with tutorials for tweaks for any of these benchmarks particularly GPUpi 3.3 100M and Y cruncher? I may have run one of these benchmarks once or twice....
Is GPUpi 3.3 for DL at HWBot? Or is their version different?
|
|
|
Post by georgekokovinis on Jan 3, 2019 19:49:39 GMT -5
For GPUPI there are some tips.
For Y-cruncher...TBH, I am surprised that Brad chose that piece of shit as part of our friendly competition. It is a disaster, an unfinished product, a problematic ( by design ) benchmark.
|
|
|
Post by ShrimpBrime on Jan 3, 2019 19:56:00 GMT -5
Haha, well then.
Can someone kindly make a GpuPi tweak thread and throw in some basics? Are there other software for tweaks like PiMod ect?
Don't recall running Y-cruncher. SO I have no feedback on it's issues. Perhaps a different benchmark if Aleslammer is up to changing it kindly?
|
|
|
Post by ozz on Jan 3, 2019 20:11:55 GMT -5
gpu as i understand it scotty, i know bugga about it, think it mighta been run a few times here with pete for whatever it was and that link is direct from hwbot lol
|
|
|
Post by ShrimpBrime on Jan 3, 2019 21:20:31 GMT -5
That's a different version? Version 3.3 not 3.3.2?
please pardon my ignorance, I really don't know if there are versions IE 3.3.1 or 3.3.0 ect...
|
|
|
Post by ozz on Jan 3, 2019 21:37:31 GMT -5
says hwbot rules, thats all bots got for dload for 100m i can see
GPUPI v3.3 GPUPI v3.3 - 1B Win 455 download website GPUPI v3.3 - 32B Win 122 download website
GPUPI GPUPI - 1B Win x x x 12660 download rules website GPUPI - 32B Win x 2339 download rules website
GPUPI GPUPI - 1B Win x x x 12660 download rules website GPUPI - 32B Win x 2339 download rules website
|
|
|
Post by Macsbeach98 on Jan 3, 2019 21:38:39 GMT -5
Do we have any threads with tutorials for tweaks for any of these benchmarks particularly GPUpi 3.3 100M and Y cruncher? I may have run one of these benchmarks once or twice.... Is GPUpi 3.3 for DL at HWBot? Or is their version different? It is downloadable from HWBot or at Shanes link its a multi threaded SPi that uses OpenCL dont download GPUPI 3.2 though its a lot slower make sure its 3.3.2 You have to install OpenCL 1.2 or 2.0 try both if you have got a radeon card OpenCL will be installed with the driver if not you have to get OpenCL from AMD or unpack a radeon driver vers 1.2 might be a bit hard to get from AMD so unpack a driver or one of us will put it up. once you have got it running play around with batch and reduction sizes. Ycruncher I have only run it once for a competition it demands system stability especially memory more stable than anything else I hae ever run it also uses AVX. Try it and see how you go.
|
|
|
Post by cbjaust on Jan 3, 2019 23:00:27 GMT -5
For GPUPI the biggest tweak is just getting OpenCL working and the drivers to recognise the hardware (CPU or GPU) as a target device for compute. Somewhere along the way AMD (catalyst 14 maybe or a later version) has dropped OpenCL support for CPU. I find that is a royal pain to get working, an experinece common to many of us I would think. The latest fail was trying to get GPUPI working for my HD 4870 X2 (hint I couldn't get it working).
|
|
|
Post by ShrimpBrime on Jan 3, 2019 23:48:28 GMT -5
OK Pete thanks. I've DL'ed it just about to give it a run with the 2700X. Got her running 4400Mhz. Assuming 0 Benchmark Pi is the selection I should type in.... lol. Here Goes......
There is a number of selections to make to run this.... no idea. going for option 1 25,000,000???
|
|
|
Post by ShrimpBrime on Jan 3, 2019 23:50:19 GMT -5
Is option one single thread what we are benchmarking here? Or are we supposed to run multithreading? I may have to change my hardware configuration a tad.... Total Compute time 7.617 seconds start to end wall time 7.882 seconds. Which of these do we submit?>
|
|
|
Post by Aleslammer on Jan 4, 2019 8:22:47 GMT -5
|
|
|
Post by Macsbeach98 on Jan 4, 2019 9:13:16 GMT -5
Is option one single thread what we are benchmarking here? Or are we supposed to run multithreading? I may have to change my hardware configuration a tad.... Total Compute time 7.617 seconds start to end wall time 7.882 seconds. Which of these do we submit?> Total compute time is what you submit here is a 1B sub of mine for the Teams Cup HWbot subs are supposed to have the HWbot submitter visable too you have to sub the Data file there. I dont know what Brads actual screen requirements are going to be though I am sure that will be included om the actual competition page. You will do a lot better running it on 8 cores 16 threads. But you have got it running Ok.
|
|
|
Post by dr4g00n on Jan 4, 2019 10:45:43 GMT -5
8700K/9700K/9900K are the ticket for y-cruncher. Ryzen lags behind by a decent amount because it just doesn't have the raw speed to crunch numbers like the Intel chips. It'll still do it of course but an 8-core will be in the 70-80sec range compared to 55-60sec for an 8700K at typical oc's.
|
|
|
Post by Macsbeach98 on Jan 4, 2019 15:40:05 GMT -5
Interesting Ryzen is back there with Haswell E we know that but I thought the extra 2 cores would of made up for the speed of 8700k.
|
|
|
Post by ShrimpBrime on Jan 4, 2019 16:31:47 GMT -5
Thanks Macs. I was running 4400Mhz on the Ryzen and that was a single threaded run not multi-threaded.
Also the version you have in that screen shot differs from mine.
Did not see an option for HWBot submitter. That looks to be a seperate exe to run eh?
I also see a list of runs I did. Most current at the top. That's the one we use for the submitter or delete the ones I'm not going to use correct??
lastly he has 100M for the bench but is not clear on single core or multi core benching with this Y-cruncher benchmark....
|
|
|
Post by ShrimpBrime on Jan 4, 2019 16:34:15 GMT -5
8700K/9700K/9900K are the ticket for y-cruncher. Ryzen lags behind by a decent amount because it just doesn't have the raw speed to crunch numbers like the Intel chips. It'll still do it of course but an 8-core will be in the 70-80sec range compared to 55-60sec for an 8700K at typical oc's. I'm not sure WHAT you are talking about.... If someone is using a Ryzen 2700X, they should be around the 6 second mark.... at stock. Where are we getting these numbers from?? Pardon me, I really have no clue what's going on........ ___________________ Dang bench mark is silly.... I had a Cpu utilization of 1456.95%. How the hell does this cpu able to use more than it's top 100% capability lol. (serious question though) And what exactly is Kernel overload and how can I get more or less overload depending on the target to have more or less, I know not of which is good. Having Kernel overload or not. __________________ Again, I am sorry for my lack of understanding here. I need a little more serious help.
|
|
|
Post by dr4g00n on Jan 4, 2019 16:53:38 GMT -5
8700K/9700K/9900K are the ticket for y-cruncher. Ryzen lags behind by a decent amount because it just doesn't have the raw speed to crunch numbers like the Intel chips. It'll still do it of course but an 8-core will be in the 70-80sec range compared to 55-60sec for an 8700K at typical oc's. I'm not sure WHAT you are talking about.... If someone is using a Ryzen 2700X, they should be around the 6 second mark.... at stock. Where are we getting these numbers from?? Pardon me, I really have no clue what's going on........ ___________________ Dang bench mark is silly.... I had a Cpu utilization of 1456.95%. How the hell does this cpu able to use more than it's top 100% capability lol. (serious question though) And what exactly is Kernel overload and how can I get more or less overload depending on the target to have more or less, I know not of which is good. Having Kernel overload or not. __________________ Again, I am sorry for my lack of understanding here. I need a little more serious help. I'm talking about 1B not 25M that's why. Use the Hwbot submitter.jar that is included in the y-cruncher folder to run the bench. The cpu utilization is relative to a single core/thread, so for your 2700X 1600% would be complete usage of the Cpu, but that's not possible because it has to run other things. My 1700 gets about 1420% usage for 25M.
|
|
|
Post by Macsbeach98 on Jan 4, 2019 17:54:18 GMT -5
HWBot submitter is in the folder for the bench if you downloaded it from HWBot have the folder on your desktop and open it and click on the submitter then if you save a datafile with the submitter it will save it in the folder as HWBot Datafile. Every time you do a run of the bench it will save a file in the benchmark folder that will show up in the submitter you can delete previous run files if you want before you do a run. The file that you submit to HWbot with the submitter is different to the file saved for each run. I dont know if you can make sense of that but I am sure you can work it out from there.
Your 1456.95% utilization is for all 16 threads added together if you look at my screenshot I have got 396.3% for 4 threads
|
|
|
Post by georgekokovinis on Jan 4, 2019 18:05:44 GMT -5
HWBot submitter is in the folder for the bench if you downloaded it from HWBot have the folder on your desktop and open it and click on the submitter then if you save a datafile with the submitter it will save it in the folder as HWBot Datafile. Every time you do a run of the bench it will save a file in the benchmark folder that will show up in the submitter you can delete previous run files if you want before you do a run. The file that you submit to HWbot with the submitter is different to the file saved for each run. I dont know if you can make sense of that but I am sure you can work it out from there.Your 1456.95% utilization is for all 16 threads added together if you look at my screenshot I have got 396.3% for 4 threads LOL, LOL, LOL
Irrelevant,
Havli said one of the wisest things I have read :
" And the last one, please shut down the servers and turn off the lights "
I dont know if you can make sense of that but I am sure you can work it out from there, haha
|
|
|
Post by ShrimpBrime on Jan 4, 2019 18:14:27 GMT -5
HWBot submitter is in the folder for the bench if you downloaded it from HWBot have the folder on your desktop and open it and click on the submitter then if you save a datafile with the submitter it will save it in the folder as HWBot Datafile. Every time you do a run of the bench it will save a file in the benchmark folder that will show up in the submitter you can delete previous run files if you want before you do a run. The file that you submit to HWbot with the submitter is different to the file saved for each run. I dont know if you can make sense of that but I am sure you can work it out from there.Your 1456.95% utilization is for all 16 threads added together if you look at my screenshot I have got 396.3% for 4 threads LOL, LOL, LOL
Irrelevant,
Havli said one of the wisest things I have read :
" And the last one, please shut down the servers and turn off the lights "
I dont know if you can make sense of that but I am sure you can work it out from there, haha
Your trying to say something.... Not sure what it is.... but ok.... I'll figure it out. One guys talking about a quad core. Another 1B not 100M like the comp states. Nobody can tell me if we are running a single thread or all of them as it doesn't state it in the rules.... And have yet to see an explanation about the Kernel overhead. lol.
|
|
|
Post by ozz on Jan 4, 2019 18:19:03 GMT -5
my opinion, i think it should be scrapped and start again, its been confusion from the get go on a few fronts, no offence meant ales
|
|
|
Post by ShrimpBrime on Jan 4, 2019 18:22:11 GMT -5
my opinion, i think it should be scrapped and start again, its been confusion from the get go on a few fronts, no offence meant ales I just want to seriously learn about the benchmark, but there are twists and turns I'm still trying to decipher if we are comparing an Intel to AMD at 1B. IDC about that. I just want to be sure I'm going in the right direction, Pete's been a big help so far. I get the submit to HWBot thing, but not really my primary goal, it was thrown into mention which is cool. I don't think it should be scrapped per say. Just a more detailed description of the end rules would be good though lol. Only because I have no idea about y-cruncher, it shouldn't be scrapped.
|
|
|
Post by ozz on Jan 4, 2019 18:26:35 GMT -5
my opinion, i think it should be scrapped and start again, its been confusion from the get go on a few fronts, no offence meant ales I just want to seriously learn about the benchmark, but there are twists and turns I'm still trying to decipher if we are comparing an Intel to AMD at 1B. IDC about that. I just want to be sure I'm going in the right direction, Pete's been a big help so far. I get the submit to HWBot thing, but not really my primary goal, it was thrown into mention which is cool. I don't think it should be scrapped per say. Just a more detailed description of the end rules would be good though lol. Only because I have no idea about y-cruncher, it shouldn't be scrapped. moreso i meant that ^^^^^ than scrapped, its just an aussie term used to start again
|
|
|
Post by ShrimpBrime on Jan 4, 2019 18:28:29 GMT -5
Oh, my bad Shane.
I have no more questions. I really just need to know single or multi threaded. Gonna keep googling around. OCN probably has some insite, I'll just go there.
|
|
|
Post by georgekokovinis on Jan 4, 2019 18:31:19 GMT -5
my opinion, i think it should be scrapped and start again, its been confusion from the get go on a few fronts, no offence meant ales I just want to seriously learn about the benchmark, but there are twists and turns I'm still trying to decipher if we are comparing an Intel to AMD at 1B. IDC about that. I just want to be sure I'm going in the right direction, Pete's been a big help so far. I get the submit to HWBot thing, but not really my primary goal, it was thrown into mention which is cool. I don't think it should be scrapped per say. Just a more detailed description of the end rules would be good though lol. Only because I have no idea about y-cruncher, it shouldn't be scrapped. Seriously now friends -
Although it is Brad's and only Brad's priviledge to decide, I insist that Y-cruncher is a piece of shit.
And here is the most simple explanation I can provide :
Two very high end systems, both at DEFAULT SETTINGS AND RAM AT XMP !
1) Asus Rampage VI Extreme edition + I9-7960X + 32gb Galax Hall of Fame. ALL STOCK. nothing overclocked. 2) Asus Apex X. I9-9900K + 16gb Galax Hall of Fame limited edition. ALL STOCK. nothing overclocked.
And the benchmark crashes. On a perfectly stable OS.
Come on now !!!
|
|
|
Post by ozz on Jan 4, 2019 18:31:54 GMT -5
Oh, my bad Shane. I have no more questions. I really just need to know single or multi threaded. Gonna keep googling around. OCN probably has some insite, I'll just go there. pete knows how every damn benchmark works, ask him
|
|
|
Post by Macsbeach98 on Jan 4, 2019 19:21:07 GMT -5
Jon my screenshot wasnt to compare 1B to 25M it was only to show an example screenshot. For HWBot its multithreaded. I would say it will be multi threaded but single thread might be more even that is Brads decision.
Yes George it is really demanding on memory I had to slow mine down a bit to get it to make it through the bench.
|
|
|
Post by Mr.Scott on Jan 4, 2019 19:23:21 GMT -5
I've never been able to get y-cruncher to work on anything. I have not a single sub on it. It's as difficult as GPUPI tho get going. I need to try harder. I know less the Jon about it.
|
|
|
Post by georgekokovinis on Jan 4, 2019 19:30:58 GMT -5
@ Pete - I am 110% sure Pete that you are correct. Question is why should someone spend even one minute of his life for a benchmark that gives no points.
@ Scotty.
GPUPI has never been an issue. One of the most straight forward benchmarks I have ever seen. The only points are : 1) Correct OpenCl driver installed ( Intel or AMD ) 2) HPET enabled for Win 10.
|
|
|
Post by Macsbeach98 on Jan 4, 2019 19:45:25 GMT -5
100% agree George I dont run benchmarks for no points the only time I have run ycruncher was for the Teams cup.
Scotty GPUPI 3.2 /3.3.2 works easy as George said you just have to have the right openCL its a lot easier to get running than 2.5.4 was you dont have to install the intel vcredist or anything just the openCL. 3.3.2 is about 40% more efficient than 3.2 and is why Alby stopped points and it has its own ranking.
|
|