mfakto: an OpenCL program for Mersenne prefactoring

[tului]

Going to give it a shot on the iGPU on an AMD Ryzen 9 4900HS and see how it works.

edit:
for Factor=N/A,98302559,75,76 it is saying 14h47m to complete at the first screen update at 0.2%. 126GHz-d/day.
Has anyone got any tweaks for AMD's latest iGPUs?

[kriesel]

Originally Posted by tului

for Factor=N/A,98302559,75,76 it is saying 14h47m to complete at the first screen update at 0.2%. 126GHz-d/day.
Has anyone got any tweaks for AMD's latest iGPUs?

That's a good rate for an igp. Recent Intel igps are typically 18-20GhzD/day.
Try the generic ini file tuning, similar to https://www.mersenneforum.org/showpo...99&postcount=8

[DrobinsonPE]

For the last almost 2 weeks I have been running the windows compiled version of mfakto 0.15pre7 on the iGPU of a Ryzen 3 3200G. it passed multiple runs of mfacto -st so I started feeding it a few GPU72 TF assignments and a few PrimeNet manual assignments.

Depending on the exponent I have seen between 126 and 132 GHz-d/day. I have not tried tuning any of the ini file settings yet. I also installed the windows compiled version of gpuowl and got it working but it is slow so I stopped it. I was seeing 18.1 ms/itt for a 111M exponent.

The above numbers are with the CPU idling. I just loaded Prime 95 and started running it but do not have any firm numbers yet.

[kriesel]

Originally Posted by DrobinsonPE

I also installed the windows compiled version of gpuowl and got it working but it is slow so I stopped it. I was seeing 18.1 ms/itt for a 111M exponent.

The above numbers are with the CPU idling. I just loaded Prime 95 and started running it but do not have any firm numbers yet.

That's about 21GhzD/day in gpuowl on the iGPU, pretty good, giving completion of a PRP in ~23 days. (What version gpuowl-win was that?) Please benchmark prime95 alone, mfakto alone, gpuowl alone, prime95 with mfakto, and prime95 with gpuowl and post results. In some cases I've seen prime95 throughput go up when mfakto is run on an igp. See https://www.mersenneforum.org/showth...284#post550284 for the details.

[DrobinsonPE]

I have mfakto 0.15pre7 running 100390691 from 74 to 76 at about 107GhzD/day. At the same time prime95 30.2 is running LL-D on 56923663 at about 6.3 ms/itt. I will work on the other configurations to get data. The computer is pulling 94W at the plug while doing the above.

[kriesel]

Originally Posted by DrobinsonPE

I have mfakto 0.15pre7 running 100390691 from 74 to 76 at about 107GhzD/day. At the same time prime95 30.2 is running LL-D on 56923663 at about 6.3 ms/it. I will work on the other configurations to get data. The computer is pulling 94W at the plug while doing the above.

Sounds good.

https://www.mersenne.ca/credit.php?w...tton=Calculate gives 120.65 GhzD for the LLDC. 56923663 * .0063sec/iter /3600 / 24 = 4.15 days; so prime95 ~29.07GhzD/day there. (Really 56923663-2, not that it matters at 12.6msec difference.) So, about 136 GhzD/day combined.

V30.3b2 prime95 is out and includes some bug fixes. https://www.mersenneforum.org/showthread.php?t=25823

[aheeffer]

Originally Posted by DrobinsonPE

For the last almost 2 weeks I have been running the windows compiled version of mfakto 0.15pre7 on the iGPU of a Ryzen 3 3200G. it passed multiple runs of mfacto -st so I started feeding it a few GPU72 TF assignments and a few PrimeNet manual assignments.

Depending on the exponent I have seen between 126 and 132 GHz-d/day. I have not tried tuning any of the ini file settings yet. I also installed the windows compiled version of gpuowl and got it working but it is slow so I stopped it. I was seeing 18.1 ms/itt for a 111M exponent.

The above numbers are with the CPU idling. I just loaded Prime 95 and started running it but do not have any firm numbers yet.

Thanks! Your post made me realize I could make some use of the idle APU of my Ryzen5 3400g box which is running with two Radeon VII cards. To my great surprise Mfacto 015pre7 runs between 420 and 480 Ghz-d/day (depending on the factor size) on the Vega 11 APU of the 3400G! Not bad for a 140 € cpu.

[DrobinsonPE]

Originally Posted by kriesel

That's about 21GhzD/day in gpuowl on the iGPU, pretty good, giving completion of a PRP in ~23 days. (What version gpuowl-win was that?) Please benchmark prime95 alone, mfakto alone, gpuowl alone, prime95 with mfakto, and prime95 with gpuowl and post results. In some cases I've seen prime95 throughput go up when mfakto is run on an igp. See https://www.mersenneforum.org/showth...284#post550284 for the details.

The results are somewhat off topic for this thread so feel free to move this post but here are my results. I added the benchmarks you requested, filtered my spreadsheet for the 3200G, and printed it to PDF (see attached).

[DrobinsonPE]

My other windows computer that I have been running mfakto 0.15pre7 on is a Celeron J4150. It does about 7.5 Ghz-d/day with Prime 95 while also running mfakto 0.15pre7. This is not a heavy lifting machine; more of a science project to see what I can do with it. It is also the computer I use for my colab notebook.

My problem is that it does not fully pass the -st.
Selftest statistics
number of tests 34026
successful tests 33288
no factor found 738

It does run exponents though:
Started a simple selftest ...
Selftest statistics
number of tests 30
successful tests 30

selftest PASSED!

got assignment: exp=207727591 bit_min=73 bit_max=74 (9.21 GHz-days)
Starting trial factoring M207727591 from 2^73 to 2^74 (9.21 GHz-days)
Using GPU kernel "cl_barrett32_76_gs_2"

Found a valid checkpoint file.
last finished class was: 2016

Date Time | class Pct | time ETA | GHz-d/day Sieve Wait
Aug 13 19:31 | 3968 85.9% | 226.15 8h28m | 3.66 81206 0.00%

To get mfakto to run I changed GPUType=INTEL in the ini file. I have updated the graphics driver, downloaded GPU-Z and opencl-z to compare the settings and driver information. I am now far beyond my knowledge base.

I do not trust the results so I have rerun all of the exponents through my 3200G.

[kriesel]

Originally Posted by DrobinsonPE

My other windows computer that I have been running mfakto 0.15pre7 on is a Celeron J4150. ...
My problem is that it does not fully pass the -st.
Selftest statistics
number of tests 34026
successful tests 33288
no factor found 738

Is there a pattern to which are not found? Certain bit levels?
I had one with issues above ~81 bits. It's usable lower.

[DrobinsonPE]

Originally Posted by kriesel

Is there a pattern to which are not found? Certain bit levels?
I had one with issues above ~81 bits. It's usable lower.

I have not been logging the results to a file and the command prompt window only goes back so far and stops. I saw in a thread today where you used mfakto -st -d xy >>selftest.txt. I assume that the >>selftest.txt sends the output to a text file so I will try that and see what it shows. It takes about 8-9 hours to run the self test so I will start it tonight and see if it is done in the morning.