Posts by jason_gee

1) Message boards : Number crunching : What GPU to use? (Message 1834265)
Posted 9 hours ago by Profile jason_gee
Yes, the App should run on the same GPUs that are currently running the Petri 'Special' App and all others with CC 3.2 and above. That includes some low end 700 GPUs that AFAIK have never tried the 'Special' App. It will be interesting to see how it runs on the Low End GPUs. It runs fine on my 750Ti & 950 cards. As an added Bonus the CUDA 6.0 Libraries are already at Beta and Main and can be downloaded from there as soon as someone figures out the download links. I'm just waiting for the source code to be posted in the repository before posting the App.


and updated over the same alpha subfolder svn location, for easy comparison to the prior sources. I have a list of issues that need addressing for the Windows build at least. Which ones may or may not be applicable on Linux/Mac I'm unclear about at this point. My extra Mac Pro PCIe cable has arrived. will likely be seeing if I can wedge in the 1050ti at some stage (presuming the web drivers will work, el capitan for now). Waiting on some SSD caddies, then it'll get Ubuntu and Win10.
2) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1834155)
Posted 21 hours ago by Profile jason_gee
Another main exceptional area to watch is the 'too many triplets' fallback processing, which at least in my own build appears to be broken, and falling into a 'funk'. That's an extremely rare situation, but has happened here at least once in recent days, with the probability going up with noisy data.

What I'll probably do is update the svn alpha code with the zi+a submission as supplied by Petri, but continue focussing on the clean OOA&D 'long game' which is turning up some interesting opportunities/holes in the old app design(*). Am likely to have more time from the end of this week (fingers crossed).

[ (*), namely the baseline, alternate optimised, Cuda internal, and exceptional case paths are behaving polymorphically already in an incomplete way, which opens up a number of streamlining/simplification options]
3) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1833786)
Posted 2 days ago by Profile jason_gee
Yes x41+a did well here to some point (before the described showstoppers). It'll be interesting to see if you come up with something the same or different :)
4) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1833764)
Posted 2 days ago by Profile jason_gee
True that. Since I installed the 1050ti, I can hear the Corsair PSU and Noctua CPU fan.... also mice crawling in the walls :/
5) Message boards : Number crunching : I've Built a Couple OSX CUDA Apps... (Message 1833763)
Posted 2 days ago by Profile jason_gee
I was just looking at the repository and noticed the last CUDA Special App there is zi3f , https://setisvn.ssl.berkeley.edu/trac/browser/branches/sah_v7_opt/Xbranch/client/alpha/PetriR_raw3/analyzeFuncs.cpp#L715
It's possible the zi+ code may work reasonably well when compiled in ToolKit 6.0 and run with a CUDA 8 driver. I'm still looking at it.
Anyway the zi+ code can be added to the repository just in case this build works out? Since early on the 30th the two builds look to be promising.


x41zi+a is looking promising, but has a number of showstopping warts I'm trying to clarify the source of prior to committing. Those are namely intermittent failure with 'unknown error' in the powerspectrum/summax pipeline, and some very odd validation concerns that didn't appear at first. I had suspected my own system + GTX980 as being a possible issue related to what I was seeing, but the picture hasn't really improved with brand new 1050ti, and very thorough maintenance & system checks. In that light I switched back to prepping x42 infrastructure to receive all of baseline, Petri, and internal regression code. The situation has basically become complex enough that the full range of issues won't be completely isolated until the tools exist to find the problems :)
6) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1833138)
Posted 6 days ago by Profile jason_gee
Having installed the GTX 1050 on the Windows 10 PC, dated 2014, I would like to install the GTX 750 OC on my SUN workstation dated 2008 and still working. But I think its BIOS is legacy, not UEFI. Can I install the 750 on it? It is running SuSE Linux 42.1 and has now a HD 7770 GPU board.
Tullio


Yes. The 1050ti I run is on a legacy G33 chipset earlier than that, with only PCIe1.1, so 750 should feel right at home.
7) Message boards : Number crunching : SETI@home v8.19 Windows GPU applications support thread (Message 1832933)
Posted 7 days ago by Profile jason_gee
... But I see they gave me a separate installation of Vulkan libraries without asking me...
The Vulkan Runtime/loader I'd expect since, for gaming, ultimately it's intended to become ubiquitous. [Compute also, but less mature]. LunarG supply the Windows SDK, loader (ICD) and runtime with the latter two meant to be tacked onto drivers and/or application installers just like that. The current GFE and m$ situations being mentioned around seem like a good idea to avoid to me. I see no good coming from those (but that's a topic for another time/thread)
8) Message boards : Number crunching : SETI@home v8.19 Windows GPU applications support thread (Message 1832860)
Posted 8 days ago by Profile jason_gee
particularly odd, maybe maybe not, I run 375.95. Mind you I do use a very old modified version of Boinc. perhaps something went off the rails since now and 5 years or so before the 1050 came out, lmao. Have noticed reading isn't Boinc developers' strong point [Edit: or listening].
9) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1832859)
Posted 8 days ago by Profile jason_gee
huh, the plot thickens :)
10) Message boards : Number crunching : SETI@home v8.19 Windows GPU applications support thread (Message 1832846)
Posted 8 days ago by Profile jason_gee
Can support the view that the newer variants of Geforce Experience, and any supporting telemetry, are poisonous to what we want to achieve here.
11) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1832771)
Posted 8 days ago by Profile jason_gee
I have substituted a GTX 750 OC with a GTX 1050 Ti on my Windows 10 PC and it is much faster. But why does its fan run and make noise even when not crunching GPU tasks? The 750 fan was almost never heard even when crunching GPU tasks.
Tullio


Don't know, but by default my Gigabyte G1 gaming 1050ti variant turns the fan off completely ( 0rpm) for extended periods until some threshold, then runs at inaudible ~1000rpm under high loads for a bit. I have the reverse situation to yourself, whereby I would rather a steady low fan speed, so I set 'aggressive' fan curve in (evga) precisionxoc. Since that runs ~95% load at 39C, most likely you could set to a quiet profile comfortably, depending on the cooler yours has. [..and choice of software... still coming to grips with the increased configurability in this gen.]

[Edit:] I've also yet to work out why I see mine boost above the rated boost clock (to ~1866MHz) periodically also, but only while playing youtube videos at the same time as crunching. I'm guessing there's some driver threadiness going on that throws a switch, but fingers crossed I might have lucked out in the silicon lottery finally...
12) Message boards : Number crunching : Best Driver for 750ti? (Message 1832329)
Posted 10 days ago by Profile jason_gee
Vulkan and dx12 are still evolving, so where applicable this will be a factor. This is a positive for development (where Cuda and OpenCL rely on these), but a transitional burden. Naturally in the case of dx12 you're looking at win10 only, while the vulkan infrastructure (especially compute) seems to be receiving a lot of attention from all parties other than m$.
13) Message boards : Number crunching : GPU FLOPS: Theory vs Reality (Message 1832319)
Posted 10 days ago by Profile jason_gee
Nice!, can vouch my house is substantially cooler switching for development to a single 1050ti. Having been digging, I'd pin the more or less equivalent Credit/Wh of the top four or five, on the appropriate graph, as intentional. The wider variance on the 1050ti in particular I would tend to blame on the recent driver bloat/crud, which requires manual intervention to deal with.
14) Message boards : Number crunching : Looking for a program to deal with drive usage.... (Message 1832151)
Posted 11 days ago by Profile jason_gee
Cleansweep ?
15) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1831974)
Posted 12 days ago by Profile jason_gee
Yeah fascinating times ahead as the changes kick in. The compute portion of Vulkan is starting to look quite viable now, which is a good thing. Will eventually take some of the edge off comparing apples with oranges at some point.
16) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1831892)
Posted 13 days ago by Profile jason_gee
Nice, new toy, now you could build a new 10x0 target app. LOL


Built a new 10x0 target app months ago. Possibly just turns out my crap hardware is too hammered. Will see what the inconclusives do with the new board.
17) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1831884)
Posted 13 days ago by Profile jason_gee
looking at what I'm seeing so far, something about half a GTX 980 in throughput for about a quarter the power. The fans are alternating between 0rpm (off) and ~1000rpm and remarkably quiet. Driving 2 1080p displays while crunching and watching twitch streams no problems (win7, old core2duo host).

I recently got a Skylake-based host to help Raistmer test out the iGPU problems, and once we'd sorted that out, I dropped in a 1050Ti - 'Asus Expedition', designed for non-stop use (dual ball-bearing fans). It's running GPUGrid cuda 8.0 jobs at the moment, until there are SETI apps which can make use of 4 GB of VRAM - but the CPU is knocking through BLC VLARs in well under an hour. Host 8121358

The whole system is typically drawing around 120W from the wall at full load - sometimes up to 130W if GPUGrid sends out a particularly intensive task. And the whole rig is almost completely silent - I have to put it in a room with no other noise source and shut the door to be able to tell it's running by sound alone.


Well I'm certainly marvelling at how quickly things have changed. It's going to be interesting trying to palm off some of my old hardware to make room for the niceness. lol.
18) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1831877)
Posted 13 days ago by Profile jason_gee
looking at what I'm seeing so far, something about half a GTX 980 in throughput for about a quarter the power. The fans are alternating between 0rpm (off) and ~1000rpm and remarkably quiet. Driving 2 1080p displays while crunching and watching twitch streams no problems (win7, old core2duo host).
19) Message boards : Number crunching : GPU Wars 2016:  Pascal vs Polaris (Message 1831862)
Posted 13 days ago by Profile jason_gee
Can confirm as of a short time ago, the 1050ti is 'really nice' for the older system upgrade, or budget cruncher. At least the Gigabyte G1 gaming version I received this morning. Looks like I'll be running a bit cooler (and quieter) for the coming aussie summer.
20) Message boards : Number crunching : Too many errors Ubuntu 16.04 nvidia (Message 1831843)
Posted 13 days ago by Profile jason_gee
I'm seriously considering wasting money on some Windows licenses :(


There's as much flux in the Windows world (for NV at least), and just as much hairpulling :)


Next 20


 
©2016 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.