Posts by petri33

1) Message boards : Number crunching : Would like to contribute. Need help. (Message 1939027)
Posted 10 days ago by Profile petri33
Post:
Hi,
Get a refurbished/used/old system and a new/newish GPU. On the NVIDIA side at least the Pascal (10x0) and on the AMD a GPU with the highest number of independent acceleration units...

... (since I'm expecting the OpenCL SW to catch up with the latest advancements regarding unused GPU cores ... grid(1, x, y). On NVIDIA the first grid parameter tells on how many SMXs to run (Imagine 1 of 80 on a VOLTA GPU). NVIDIA will not reveal their preference to schedule the GPU threads but my tests have revealed one bit.

Petri[/b]
2) Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread) (Message 1936847)
Posted 29 days ago by Profile petri33
Post:

...
CreditScrew just what it is, based on too many non-valid in real life assumptions.
The most hated (for me) part in it: it directly discourages any app optimization. I think it's counterproductive. Think Petry would agree...


Hi Raistmer!

I totally agree.

From what I have seen I think that the faster a WU is computed the less credit it is asking/awarded. My slowest cards get more points from same type of WU compared to the fastest card. I suspect that even if I had all 4 TITAN V's I'd see less credit per WU after any optimisation in the code.

I remember the time some years ago when I ran 2-8 at at time per GPU. The credit per WU was higher compared to running 1 at a time. And the machine did more WUs per hour. Double 'good'.
3) Message boards : Number crunching : Invalid Host Messaging (Message 1936504)
Posted 21 May 2018 by Profile petri33
Post:
My host is producing masses (more than before) of inconclusives that probably error out later. I'm testing CUDA 9.2. I'm at work at the moment so I can not stop the computer doing so.

petri33
4) Message boards : Number crunching : GPU FLOPS: Theory vs Reality (Message 1936333)
Posted 19 May 2018 by Profile petri33
Post:
I wonder what sort of performance this sucker would give?
GTX 1080Ti CUDA Cores 3584
Titan Xp CUDA Cores 3840
Titan V CUDA Cores 5120

For $3,000US you'd want it to kiss you good night & tuck you in as well as crunch like nothing else...


I have Compared the TITAN V to a GTX 1080. The Titan is 92 % (shorties) - 140 % (guppi & vlar) faster.
5) Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread) (Message 1936200)
Posted 18 May 2018 by Profile petri33
Post:
I thought CN basically ignored GPUs and went on CPU info to award credit for Seti.

Well, sot of, but if only it were that simple....
There is a scaling factor in the calculation that is supposed to cope with GPUs, but it is VERY crude, and doesn't cope with very fast GPUs properly. Also the logic that triggers its use is also not very robust and will make errors frequently, again when the GPU concerned is a very fast one (as Petri's are).
What one has to remember is that much of both the logic and the maths was laid down at about the time when GPUs were just coming onto the market, and nobody had any knowledge of just how fast they would get - they are a LONG way ahead of the curve when compared with CPUs. Back then it was assumed that GPUs would be about 10 times faster than a CPU, but today we are seeing many GPUs more than 100 times faster then CPUs - and some may be even further ahead.


Thank you for your comment and for the information content too.
I'd say 100x is the new normal.

-- -- there are places here on our planet where the Sun does not go down <3 They are the very same places where the Sun does not come up when there is the deepest of the winter
6) Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread) (Message 1936072)
Posted 17 May 2018 by Profile petri33
Post:
Hi,

I find it odd that my TITAN V is constantly awarded below 100 credits when it does an Arecobo vlar. It does them in 75 seconds and
that my 1080 and 1080Ti get 110-140 credits of the same type work when they use 125 - 170 seconds.

Is it possible that screwit new thinks that 75 seconds is too little time, a fake or something?

Petri
7) Message boards : Number crunching : Don't know where it should go? Stick it here! (Message 1936045)
Posted 17 May 2018 by Profile petri33
Post:
To get a grasp of what you can get with the layest HW you ccould check out this example.
http://setiathome.berkeley.edu/workunit.php?wuid=2976251655

How well does the Special app & the Titan V do with Arecibo VLARs?


Arecibo VLAR in 75 seconds!!
http://setiathome.berkeley.edu/workunit.php?wuid=2912450437

This is one that I found.

p.s. Yes there is a bug somewhere or it is the heat wave we're having here in Finland. My Titan V runs at 57C, Ti at 83C and the 1080s at 75C.
8) Message boards : Number crunching : Don't know where it should go? Stick it here! (Message 1935925)
Posted 16 May 2018 by Profile petri33
Post:
Hi, (thank You for this thread)

To get a grasp of what you can get with the layest HW you ccould check out this example.
http://setiathome.berkeley.edu/workunit.php?wuid=2976251655

This is a random pick. A TITAN V and a 1080 Ti. Different software. No way of knowing if any over clocking or under voltaging is applied.
I do not know if the other one is running one at a time or not.

My task is a plain vanilla one at a time with the latest software and with the overclock of the GPU suitable for Finnish summer.

p.s.
None of my GPUs has never been to a sauna but I just came from there so I'm probably in the need of a cooler more than my GPUs. Luckily I'm liquid cooled.

*i* sipping some liquid -- cooling

p.p.s. I'm kind of a cool guy for this age of a man anyway! Words of a nerd you might say unless you were not as nerd as I am.

p.p.p.s. Just admit it. Come out of your cooled computer closets/man caves/nerd holes/utility rooms/garages/family spaces/from under the stairs/ or move out of your mama.

Now *sipping coolant*
9) Message boards : Number crunching : Nvidia 397.31 driver available. (Message 1934801)
Posted 10 May 2018 by Profile petri33
Post:
latest driver and just in time compilation feature can give you a boost.


Petri, are you referring to how the current Nvidia drivers work? Or to your latest special app with the embedded CUDA 9.1?


NVIDIA apps in general. OpenCL too. New driver can have a new compiler and new optimisations. OpenCL (SoG) is compiled whenever a driver changes. CUDA 9.1 version has pre-compiled code for existing hardware and new hardware is supported through JIT compilation.

Petri
10) Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread) (Message 1934013)
Posted 6 May 2018 by Profile petri33
Post:
Hi,

*Laughing : Real time crunching* :)

If I remember correctly a task is a 108 second recording from the sky.

Most of the tasks can be done in under 108 seconds. Shorties take 16 seconds, 0.44 Arecibo tasks take 37 seconds, GBT vlars take 38 seconds and Arecibo vlars about 100 seconds on TITAN V. Those 1080s and a Ti take about twice the time to process them.

So you'd need as many TITAN Vs as there are simultaneous channels and polarisations recorded.

It was a couple of years ago when jason gee predicted that the real time analysis would some day be possible.

Petri
11) Message boards : Number crunching : Nvidia 397.31 driver available. (Message 1932917)
Posted 30 Apr 2018 by Profile petri33
Post:
If it's not broken...Don't fix it..

or in this case, if the driver runs the apps, don't update them... Most updates are only released to support new games, not scientific computations...

Abolutely correct. I'm still at 352.86, and I have no reason whatsoever to upgrade a fully working driver. The newer ones will not be faster for crunching.
on the contrary perhaps, since they have to support newer hardware too.

I never play games on my computer, so I'll stick with what I know works perfectly.


The latest will be faster for any new hardware generation after the SW was compiled. The software is built to run on existing HW and it is added an option to run on any future HW. So If your GPU is like the latest hot one then you may have a good luck that the latest driver and just in time compilation feature can give you a boost.

The other scenarios you can think ... You can always try.

..
Petri33
12) Message boards : Number crunching : New computer on the way (Message 1931076)
Posted 20 Apr 2018 by Profile petri33
Post:
Parts have been ordered for a new build.

The basics will be

Threadripper 1950x on an Asus Rog Zenith Extreme motherboard and 32GB 3200mhz Samsung B die memory

4 x Asus 1080ti rog strix oc cards watercooled.

Powered by a 2000W Platinum psu.

Cooling will be provided by 2 pumps one for graphics and another for processor, 3 rads and enough fans to generate a hurricane.

All mounted into a Lian-Li case.

Construction should start in about a weeks time.

Now its time to go to work:-(


A nice set-up!

I have not read the other replies to your original posting,
BUT from my experience you are going to have a problem with keeping the room cool enough. Your system will otherwise be fine.

I have a 16" tabletop 50W blower on the floor to supply fresh air in to the room. And then a duct blower aimed directly to the GPUs.

p.s. I have cut away the grills of the GPU back plates Those that face the rear end of the case with connectors for cables to the monitor(s).
13) Message boards : Number crunching : Panic Mode On (111) Server Problems? (Message 1929878)
Posted 14 Apr 2018 by Profile petri33
Post:
Volta does Arecibo tasks in 75 seconds and a 1080 does them in 150 seconds. Nice!

Half the time, similar clock speeds?


1080@1974MHz GPU / 10124MHz mem and
VOLTA@1575MHz GPU / 1944MHz mem
14) Message boards : Number crunching : Panic Mode On (111) Server Problems? (Message 1929724)
Posted 13 Apr 2018 by Profile petri33
Post:
Volta does Arecibo tasks in 75 seconds and a 1080 does them in 150 seconds. Nice!
VLAR work is totally ok. Credit per wu is bigger too. I'll wait till APR scales back up.
If it does not I'll find out what do I need to optimize next.

Petri
15) Message boards : Number crunching : Best bang -for-buck GPU 2018? (Message 1929044)
Posted 9 Apr 2018 by Profile petri33
Post:
I am working on getting you that 1050ti information. What I need to know is the query in smi for the current pwr usage?

nvidia-smi -l
16) Message boards : Number crunching : Best bang -for-buck GPU 2018? (Message 1929009)
Posted 9 Apr 2018 by Profile petri33
Post:
To keep Titan V average times comparable to your someone could calculate an average from my completed tasks. Those having around 39 seconds are Titan V blc tasks. The power draw is 133-140W during the task. Those around 68 seconds are 1080Ti (202-225W) and 90 second tasks are 1080 (141-154W).

Petri
17) Message boards : Number crunching : Best bang -for-buck GPU 2018? (Message 1928104)
Posted 5 Apr 2018 by Profile petri33
Post:
You can start with Shaggie76's often updated chart of GPU FLOPS: Theory vs Reality

I wish there were new cards released to motivate me to run another scan.

Could be interesting if you could run a scan with the Linux boxes who runs the latests CUDA builds.

I believe that could show if the 1070 realy has one of the best price x performance x power draw with them.


My titan v does twice the 1080 with less power. Titan v uses about 125-145W. 1080 uses about 160W and 1080Ti uses 225W.
18) Message boards : Number crunching : Cuda Error (Message 1927989)
Posted 4 Apr 2018 by Profile petri33
Post:
Something may have gone wrong with the driver or the GPU. A reboot may help.
19) Message boards : Number crunching : Anything relating to AstroPulse tasks (Message 1927988)
Posted 4 Apr 2018 by Profile petri33
Post:
I acquired this card just for the
a) fun,

Don't need to say anything else, havin fun is all.
Hope you have a lot of fun with it.

+1


The a) was the right choice and I'm certainly having fun.
With no WUs to process I have had time to test and optimize MB CUDA code.
20) Message boards : Number crunching : Panic Mode On (111) Server Problems? (Message 1927566)
Posted 31 Mar 2018 by Profile petri33
Post:
WUs are finally pouring in. Hundreds of them. Yeah!


Next 20


 
©2018 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.