PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Number crunching : Intel GPU (Message 157927)
Posted 127 days ago by Profile Rafael
So you keep saying, but never tell us how!!!!

We already did. Multiple times: use app_info.xml. In other words, set this up: https://boinc.berkeley.edu/wiki/Anonymous_platform

A couple things before you do:
    1. As soon as you put that file, it will delete ALL PG jobs currently running, so make sure you empty your queue before you do.
    2. Only do it if you have a discrete ARC GPU, which is currently not natively supported. Eventually it will, but for the time being, it isn't.
    3. Again, in case you missed it: only for ARC GPUS. Do NOT set this up for intel iGPUs. It is NOT worth it. Belive me, we've benchmarked it. Multiple times. It is bad for you. In fact, that's why it isn't supported natively
    4. When running app_info, you will only get jobs from the apps you've set up properly. So if you intend to do CPU work, you have to set those apps as well.

2) Message boards : Number crunching : Intel GPU (Message 157860)
Posted 129 days ago by Profile Rafael
Why do the built in GPUs not work aswell? What is the difference?

They do work, and you could certainly run them through app_info.

But the reason PG doesn't support it by default is that is simply not worth it. Not only does it run very slowly, running on the iGPU also severely cripples the CPU performance because of memory bottleneck, so much so that you end up losing overall throughput all things considered. So the executive decision was made to stop people from trying to use intel iGPU in hopes of avoiding people getting baited by it.
3) Message boards : Problems and Help : CPU Affinity and Load Monitoring (Message 157294)
Posted 172 days ago by Profile Rafael
I see, thanks for your reply and help! I don't see any remaining time difference when cooking the task, I think affinity maybe does not work, so I won't bother setting them then.

Windows does a pretty good job of handling scheduling on Intel, so don't worry about it.

It would be a different story if it was AMD, though, there can be upwards of a ~10% performance lift switching to Linux / manually fixing affinity.
4) Message boards : Number crunching : Optimizations for 30x0 series gpus (Message 155841)
Posted 292 days ago by Profile Rafael
Hi all,

are there any optimizations that can be done to better utilize 30x0 series cards?

Thanks


Avoid the smaller GFN sizes since they use slower transforms and, depending on the model of the card and how high it is in the ladder, can't fully use the chip.

Besides that, if your CPU has Hyper Threading / SMT, enable it and set Boinc core usage to 50% so you get that extra thread to better feed the GPU.
5) Message boards : Number crunching : Intel GPU (Message 155296)
Posted 325 days ago by Profile Rafael
7 years ago Boinc began supporting OpenCL on Intel GPU, and a thread discussed you could force it to do some but not all Primegrid tasks on Intel GPU by devious use of config files.

What's the current status of doing such things? (Since I assume it's not supported automatically - I can't see an option on my settings on the site).

Back in the day, they would offer pitiful performance. You'd think it was better than nothing, but it was actually worse since it would steal RAM bandwidth from the CPU and cripple performance. And although this was before LLR2 and multithreading, this was the case even with SGS and PPSE, which are the smallest projects that fit within cache, so I don't think anything would have changed.

But if you actually want to try it out, you can download the apps and test manually yourself. Genefer works, AP27 has a vram requirement that makes it unusable unless you can set 2gb of vram in bios, and PPS Sieve I'm not sure.
6) Message boards : Number crunching : Badges - Full Sets! (Message 155104)
Posted 349 days ago by Profile Rafael
Finally all active to the worst color PG has to offer past poop brown, All Jades here.

Time to retire from badge hunting and get rack up SoB till I die.
7) Message boards : Number crunching : SOB, PSP and ESP at a glance (Message 154969)
Posted 363 days ago by Profile Rafael

If you start producing T-shirts, remember they can become obsolete if something is found in SoB, PSP, or ESP.

/JeppeSN

Make them out of verlcro straps so you can remove candiadates as they get proven :)
8) Message boards : Extended Sierpinski Problem : k = 202705 (Message 152454)
Posted 478 days ago by Profile Rafael
And I just switched to ESP 2 days ago!

Dayum, should have done that last week....
9) Message boards : Prime Sierpinski Problem : Optimizing PSP processing with LLR2 .... (Message 152276)
Posted 495 days ago by Profile Rafael
A question for the group: how should I allocate the six physical cores of my i7 8700k CPU when processing PSP units? Do I go after 3x2 cores, 2x3 cores or perhaps 1x4 / 1x5 / 1x6 cores?

I am running a task right now with 5 cores allocated to it on my i7 8700k and it seems that it will take about 27 hrs to complete. Similarly, another system (i3 6300) will require about 32 hrs with only two cores to complete, thus suggesting that the five cores with the i7 8700k are an overkill and I am wasting potential resources.

With LLR2 it is not vital that we rush to return a task and, as a result, I can easily afford a slightly slower return if it will mean greater throughput per unit of time.

Thanks!


Rule of Thumb #1: always have the sum of cache usage for all your tasks a bit below your CPU cache.

Looking at single one of your tasks, it's using 2880k * 8 = ~23MB of cache, which is above your CPU's 12MB limit. So you should run fewer tasks to reduce usage... but since you can't run fewer than one, just run that single task with as many cores as you have.
10) Message boards : Number crunching : "Game Ready" vs. "Studio" driver [Nvidia] (Message 151670)
Posted 541 days ago by Profile Rafael
I recently got a new machine and it came with software called GEFORCE (R) EXPERIENCE (TM) lol

There's an option to switch from the current default "Game Ready" driver to the "Studio" driver. Does anyone know whether however they differ matters for crunching? If it matters, I'm planning on mostly doing GFN >19 & DYFL. Thanks!

They really don't, that switch is just a way of saying "most recent release" and "has been around for a while now so probably there aren't any major bugs, but no new features either".


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2023 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.19, 1.39, 1.60
Generated 26 Mar 2023 | 11:20:11 UTC