I'm getting a Quadro K620 for $50. It does have 2 gb ram. I hope this helps. The GTX 750 1gb will replace a GT 240 in another computer.

---

Yep,
Cuda error 'cufftPlan1d(&fft_analysis_plans[FftNum][0], FftLen, CUFFT_C2C, NumDataPoints / FftLen)' in file 'cuda/cudaAcc_fft.cu' in line 29 : invalid argument.
Is the error you get when you don't have enough vRam.
The thing to remember about vRam usage is generally, the newer the Code, the More vRam it uses. Newer OSes, and CUDA Apps use more.
Also, the Monitor is usually what pushes it over the edge, so, if possible don't connect a Monitor to a GPU with Low vRam. It might work if you add a 2 GB GPU to the machine and connect the Monitor to that 2 GB GPU. Also, you could try the CUDA 6.0 App. 6.0 uses the least vRam while 9.0 uses the most. As for the OS, Mint uses more vRam than say Lubuntu.

Today I'm getting 403
Forbidden
error when I click the download links

---

The CA site had a massive failure in the past few days. Arkayn has been struggling to rebuild it. He said the download links would likely be hosed. You should post a PM to him about that link to see if he can fix it.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I'm getting a Quadro K620 for $50. It does have 2 gb ram. I hope this helps. The GTX 750 1gb will replace a GT 240 in another computer.

. . HI, judging by the specs that is only an older version of the GT 730. I gave one of them away to the ex (GDDR5 VRAM). That will only support CUDA60 at best (is it CC 3.2? my 730, later model, was CC 3.5 and ran just OK with CUDA60). You would be better off sourcing some adaptors to run the 750 ...

Stephen

??

I'm getting a Quadro K620 for $50. It does have 2 gb ram. I hope this helps. The GTX 750 1gb will replace a GT 240 in another computer.

The Specs say the K620 is about Half the performance of a 750Ti. Right now there are 750 Ti on eBay for around $60. I just bought one for $59.
I'd stay away from CUDA Special 6.0 if possible, on my 750 Ti 6.0 is quite a bit slower than 9.0. The Memory savings are from using an older driver, there isn't much usage difference when both versions of zi3v are run with the 9.0 driver. Everything Newer uses More of Everything.

The problems at C.A. are slowly improving. Today I was able to Log In, yesterday I couldn't. Hopefully the links will be fixed soon.

BTW, anytime you see a GPU with DDR3 ram, back away as fast as possible...

I'm getting a Quadro K620 for $50. It does have 2 gb ram. I hope this helps. The GTX 750 1gb will replace a GT 240 in another computer.

The Specs say the K620 is about Half the performance of a 750Ti. Right now there are 750 Ti on eBay for around $60. I just bought one for $59.
I'd stay away from CUDA Special 6.0 if possible, on my 750 Ti 6.0 is quite a bit slower than 9.0. The Memory savings are from using an older driver, there isn't much usage difference when both versions of zi3v are run with the 9.0 driver. Everything Newer uses More of Everything.

The problems at C.A. are slowly improving. Today I was able to Log In, yesterday I couldn't. Hopefully the links will be fixed soon.

BTW, anytime you see a GPU with DDR3 ram, back away as fast as possible...

. . Ditto!

Stephen

..

First task completed with the k620 https://setiathome.berkeley.edu/result.php?resultid=6780072989

---

Hmmm, 3 Compute Units...not much there. It would be a little faster if you used the other Two though. You still have it set to Using unroll = 1 from command line args A 2 GB card is usually good up to Unroll 6, it starts having trouble around unroll 8. The best thing to do is remove the unroll entry and just use <cmdline>-nobs</cmdline> That will cause it to use a full CPU and make it as fast as it's going to be. Being about half a 750 Ti it should complete the BLC 16 in about 16 minutes. That CPU App you're using doesn't look very impressive either. It looks as though it's from the Seti-Boinc folder instead of the Optimized folder. That CPU has AVX, it might work better using the AVX Optimized App here, http://lunatics.kwsn.info/index.php?action=downloads;sa=view;down=467 That machine has 3 CPUs? At 80 watts a piece? Best make the best use possible from them. A 750 Ti burns around 30 watts, and finishes the BLC 16s in around 8 minutes.

First task completed with the k620 https://setiathome.berkeley.edu/result.php?resultid=6780072989

. . For the sake of reference. The Quadro card has some enhancements over the GT730 variant. It has 3 compute units compared to the 2 of the later GPU but it's OS is CC=5.0 (compared to CC=3.5), that is a big thing and will allow you to make use of the Cuda 8 or Cuda 9 successfully (as you have found out). You are still stuck with DDR3 ram but hey, if it works ...

. . FWIW, that run time is about on par with what the GT730 got running Cuda 6. As TBar says, try unroll = 3 or unroll autotune. Not too sure about trying anything higher but if TBar thinks it is worth a try with unroll = 6 then I'd give it a go.

Stephen

:)

... Even if you have 3 SATA ports to use and enough adaptors, the extra one only takes you up to a 230W GPU. So NO Titan V's :) ...

   1  TITAN V             On   | 00000000:06:00.0 Off |                  N/A | 
|100%   72C    P2   168W / 300W |   7326MiB / 12066MiB |     97%      Default |

That is the highest I've seen a TITAN V use. Things may change. This is with BLC 16 2bit guppi that run in 34 seconds.

Petri

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

... Even if you have 3 SATA ports to use and enough adaptors, the extra one only takes you up to a 230W GPU. So NO Titan V's :) ...

   1  TITAN V             On   | 00000000:06:00.0 Off |                  N/A | 
|100%   72C    P2   168W / 300W |   7326MiB / 12066MiB |     97%      Default |

That is the highest I've seen a TITAN V use. Things may change. This is with BLC 16 2bit guppi that run in 34 seconds.

Petri

. . Hi Petri,

. . Wow! So many issues in one small paragraph.

. . First, 168W / 300W was what I was referring to ... :)

. . Second, it was a joke as the other person was looking for a budget solution not a very, very pricey one. (He bought a used Quadro K620)

. . Third, I was pretty sure that all the tasks we are getting now are 4 bit (over 700KB in size not the older 350KB).

. . Fourth, again Wow! Are the Titan V's really that efficient? My 970s use almost that much power (150/180 and 137/180) and take 3.4 mins to crunch a Blc16 task not 0.6 mins. OK now I have to save my pennies furiously to buy a pair of Titan Vs. Apart from the prohibitive purchase price I could increase productivity over fivefold for a negligible increase in power consumption So what do they cost these days? $2,000 USD each? :)

. . Except of course that I don't want to put my credit into negative figures. I was feeling happy about the new Blc01 tapes being mounted as they take far less time to process and will greatly increase my gear's productivity. On the 1050ti they are 4.8 mins compared to 5.8 for the Blc16 WUs and on the 970s they are 2.8 compared to 3.4. But predictably CreditScrew has responded in it's usual b*ll busting fashion by dropping credit from about 62/task to about 50. So with a less than 20% drop in run time there is a greater than 20% drop in credit.. So with Titan V's run times would drop at least fivefold therefore credit would drop by 5 x 12 points which would mean I would be paying them credit at about 10 to 15 per task I processed .... :( { 8^) } That could wipe out my total credit completely in a relatively short time :)

Stephen

:)

...That CPU App you're using doesn't look very impressive either. It looks as though it's from the Seti-Boinc folder instead of the Optimized folder. That CPU has AVX, it might work better using the AVX Optimized App here, http://lunatics.kwsn.info/index.php?action=downloads;sa=view;down=467

That machine has 3 CPUs?

One 6 core with hyperthreading.


I'm still very new to linux. I think I trashed all my cpu workunits trying to install the avx app. The app info had two listing of versions, 8,04 and 8.05

<app_info>
<app>
<name>setiathome_v8</name>
</app>
<file_info>
<name>MBv8_8.05r3345_avx_linux64</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_v8</app_name>
<version_num>804</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<cmdline></cmdline>
<file_ref>
<file_name>MBv8_8.05r3345_avx_linux64</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<version_num>805</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<cmdline></cmdline>
<file_ref>
<file_name>MBv8_8.05r3345_avx_linux64</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>

I was just going to copy and paste into my current app_info, but I cut the 8.04 reference out

I set no new tasks. I'm going to return the portion I pasted over and run them out in the next day or 2

---

Wow, I'm confused. I thought you were talking about the Linux host with the K620 gpu. I looked at that host and see it running the TBar SSE41 r3711 cpu app. No sign of the r3345 AVX app. I find the r3711 SSE41 app faster than the r3345 AVX app. I mostly have Ryzen hosts with better SSE41 support than AVX, but I do run the r3711 SSE41 app even on my host with the Intel i7-6850K as benchmarking showed it still faster than the AVX app in most cases.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Yes, it's the K620 gpu computer, I just got the AVX app installed, no workunits completed yet. At least I now know I can revert this change

---

OK, got it. Will be curious of your runtimes with the r3345 app versus the r3711 app. Juan always found the r3345 app the fastest on his 6850K.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

That machine has 3 CPUs?

One 6 core with hyperthreading.
I'm still very new to linux. I think I trashed all my cpu workunits trying to install the avx app. The app info had two listing of versions, 8,04 and 8.05

. . I may be wrong but I think that confusion was my reference to the K620 which has 3 CU's compared to the GT730 version which has only 2 CUs

. . As for trashing WUs, that is an occupational hazard when one toys with app_info.xml. The good thing is that if you are so inclined you can recover them by initiating resends if you have a mind to.

I was just going to copy and paste into my current app_info, but I cut the 8.04 reference out
I set no new tasks. I'm going to return the portion I pasted over and run them out in the next day or 2

. . Not much point in putting it back once the WU's are ghosted. If you trigger resends they will come down as the new version, same as any new work.

. . If you don't know the process just ask ...

Stephen

168/300W
That is the highest I've seen a TITAN V use. Things may change. This is with BLC 16 2bit guppi that run in 34 seconds.
Petri

. . OK, for a bit of fun I looked up the prices ... no worries I'll get 2 tomorrow ... <joking>

. . I wasn't even close on price ... list is $2,999 USD, here they are $4699 AUD and cannot be bought individually, the only local supplier I found so far only sells them with a new system.

. . I would say there are many things apart from power requirements that will stop most people putting them in their crunchers ... :)

Stephen

:)

. . Hi ppl

. . I found an interesting wingman today ...

. . Ironically this host has a GT730 running CUDA 4.2 (stock host), personally I would have loaded Lunatics and gone for CUDA 5.0.

http://setiathome.berkeley.edu/result.php?resultid=6868448450

. . on this machine ...

http://setiathome.berkeley.edu/show_host_detail.php?hostid=7957802

. . It is a standard sort of GT730 running the stock GPU clock of 901 MHz, and I am guessing the "normal" DDR3 ram. I could never find many cards with the GDDR5 ram when I was shopping for mine. Which BTW is factory overclocked to 1002MHz with 2GB of GDDR5 ram. Of course I was not surprised that the run time on this host even using Cuda 4.2 was only 46 mins. LOL But I guess I am just a liar ...

Stephen

:)

. . For BOINC he recommends this version as opposed to the one in the Linux Software repository. Having accidentally used the other one I have to agree with his recommendation. This one allows you to install BOINC into /home/<username>/BOINC where you can work with it freely while the other version installs under identity ROOT in /var/lib/boinc-client and creates hassles with access and permissions. That is best avoided.

boinc_7.2.42_x86_64-pc-linux-gnu.sh

. . He has a link to the BOINC version in his thread, I will see if I can copy it to here.
Stephen

:)

This one? http://www.bc-team.org/downloads.php?view=detail&df_id=15&sid=c9a35920f823db9c27e994901cee4881

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

. . Here is his link to the CUDA60/80 Special app:-

http://www.arkayn.us/forum/index.php?topic=197.msg4499#msg4499

Is NVIDIA-Linux-x86_64-375.39.run still most suitable driver?
Found CUDA 6/CUDA9 but no CUDA8. Is it obsolette?

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.