Posts by Mr P Hucker

21) (Message 7690)
Posted 11 Jan 2023 by (banished: ID 389916)
Post:
Argh, that's the last time I use preview. I realised 0.1 seconds after closing the browser, I was looking at the preview and not my actual post. I'll type it again....

I have pendings back to Jan 1 which is the deadline horizon if these website side times are GMT. I feel like the amount of pendings i have outstanding relative to how much i have validated is rather low.
Yeah they'll get done eventually. Chances are if they time out the next guy will do them straight away. I for example return them in about a day. There's nothing dodgy going on here like in Rosetta where tasks get lost and not resent or something.

The available work behavior seems to be systematic, a large amount of work becomes available it gets worked out before new work is added, it runs out for a couple days and new work becomes available. I'm sure someone could better illustrate the work availability of timer.
I read on the main page they get data from all sorts of telescopes. I guess those send them here in batches every week. They need more sources to keep us busy!
22) (Message 7688)
Posted 11 Jan 2023 by (banished: ID 389916)
Post:
Hey Keith
You can validate my pending Android work that's been in the Que since Jan 03rd ;)

Cheers

G
Those are all allocated to mojdan in Czech who has 72 machines! None are android, so I guess a PC can validate a phone. Unfortunately for you he's set a buffer of about 5.5 days. That's if he doesn't run other projects aswell.
23) (Message 7683)
Posted 9 Jan 2023 by (banished: ID 389916)
Post:
Peter,

thanks for the good and clear reply.

BUT, since I am not running more than one CPU task at a time,
my "statement" is "correct".

I'm running other things on the remaining cores.

Thanks again and have a great day.

P.S. the "friendly repIies" did not pertain to you - but I think the message came over pretty clear - for the others ....
My point is your whole CPU can do more than your whole GPU, which is unheard of in Boinc projects.
24) (Message 7680)
Posted 8 Jan 2023 by (banished: ID 389916)
Post:
One more thing I've thought of. For anyone running a single Asteroids task at a time on their GPU:

How much RAM is used inside the GPU? Could you fit more tasks in at once?

If the GPU is not fully utilised (check in MSI Afterburner, GPU-Z, etc.) it may be waiting on the CPU. Load as many Asteroids tasks onto the GPU at once as will fit into it's RAM (although more than 4 is probably overkill). That way it can grab 4 CPU cores instead of just 1, and get on with work more quickly.

To achieve this, in C:\ProgramData\BOINC\projects\asteroidsathome.net_boinc (or similar folder if your system is different to mine), place a file called app_config.xml if it's not there already. Just a plain text file, containing:

<app_config>
    <app>
        <name>asteroids</name>
        <gpu_versions>
            <gpu_usage>0.25</gpu_usage>
            <cpu_usage>0.010000</cpu_usage>
        </gpu_versions>
    </app>
</app_config>
You will need to change "asteroids" to the correct term for the GPU version of the app, I don't know what that is since I have no Nvidia GPUs to get one. If you don't know, simply use the example above, then Boinc will produce an error like:

Your app_config.xml file refers to an unknown application 'asteroids'.  Known applications: 'something','somethingelse'
Where "something" and "somethingelse" are the possible app names.
25) (Message 7678)
Posted 8 Jan 2023 by (banished: ID 389916)
Post:
I do run Milyway and Eienstein as well as other projects and I make my GPU's available to those projects as well.
I'm just saying it's more sensible to allocate chips to what they're best at.
26) (Message 7676)
Posted 8 Jan 2023 by (banished: ID 389916)
Post:
Hi Peter,

I'm getting even more confused now.

My GPU tasks are using one core - same as my CPU tasks.

Friendly HELP needed ...

Cheers
(Conclusion at the bottom)

Ok, your Titan V can do one Asteroids task at a time, and takes 4 minutes to do so.
Your Xeon W 2245 can do 16 Asteroids tasks at a time, and takes 37 minutes to do so.

The Titan achieves one task every 4 minutes.
The Xeon achieves 16 tasks every 37 minutes, or one task every 2.3 minutes, so almost twice as fast as the Titan.

In other projects, my 10 year old 280X GPU will do a Milkyway seperation task in 35 seconds. My 6 year old i5 8600K CPU does a task in 45 minutes, but 6 at once, so one every 7.5 minutes, so the GPU, although much older than the CPU, is 13 times faster.

So your very fast GPU is half the speed of your good CPU on Asteroids.
But my ancient GPU is 13 times faster than my reasonable CPU on Milkyway.

GPUs are well suited to certain things, CPUs are well suited to others. The GPUs should go on projects where they excel, assuming you like projects other than Asteroids.

CONCLUSION:
Your Xeon is twice as fast as my i5, so it would churn out a Milkyway seperation task in 3.25 minutes. Your Titan would churn one out every 5 seconds, 39 times faster than your CPU, whereas here on Asteroids, it's half the speed!

If I had your computer, I would stick the CPU on Asteroids and the GPU on Milkyway.
27) (Message 7674)
Posted 8 Jan 2023 by (banished: ID 389916)
Post:
The CPU only uses 1 core for the task. So you need to divide that time by the number of cores, since it can run many at a time without slowing them down.
28) (Message 7670)
Posted 8 Jan 2023 by (banished: ID 389916)
Post:
Yes my GPU is slower than your CPU ... but it is faster than my CPU Dell Precision T5600 running
Intel(R) Xeon(R) CPU E5620 @ 2.40GHz [Family 6 Model 44 Stepping 2] (4 processors)

You may note that I have worked a couple credits for this project.
Just saying I've never seen such a slow project on GPU. Asteroids does not seem to be suited to GPU. You should put that GPU onto Einstein or Milkyway. Those run 30+ times faster than a CPU. If you look at the total flops in the applications page here, the CPUs are outdoing the GPUs, that's just absurd. https://asteroidsathome.net/boinc/apps.php
29) (Message 7667)
Posted 7 Jan 2023 by (banished: ID 389916)
Post:
Bill, how many do you run concurrently on your nvidia card? So I can get an idea of speed.
Bill here I believe that my NVIDA GX 1060 runs 4 concurrently and the Run and CPU times are a fraction of the time on my CPU only systems.
1647 seconds to do 4, so it churns out one every 412 seconds. My Ryzen 9 3900XT (using fractionally less power) does 24 in an hour, so one every 150 seconds. Your graphics card is slower than my CPU. That's just not right. Every other project, the GPU is way faster, not slower.
30) (Message 7661)
Posted 2 Jan 2023 by (banished: ID 389916)
Post:
What is pah project?
ROFL! Sorry, it does look like part of the list. It's an exclamation of mild annoyance, like "grrr..."

For example https://youtu.be/Th6ts_O7Jno?t=24
31) (Message 7656)
Posted 2 Jan 2023 by (banished: ID 389916)
Post:
Just wondering if the project is going to fire up the work generators again, soon?

Been getting no work for most of the day now with the RTS buffer at zero.
I guess it comes in batches. We need more data sources.

Pah, Rosetta and WCG are out of work too.

Look at it this way, it means us lot are keeping up with the demand of the scientists.
32) (Message 7646)
Posted 30 Dec 2022 by (banished: ID 389916)
Post:
It is not a GPU project. As you can see here Applications there is a lot of hardware supported.
I don't understand. It runs on Nvidia cards, so it's a GPU project.

But the tasks seem to run faster on CPUs than GPUs.
33) (Message 7641)
Posted 30 Dec 2022 by (banished: ID 389916)
Post:
Actually since there are Cuda application tasks this is a GPU project if I am correct?
I hadn't noticed them (I only have AMD cards). What's strange is the total Gflops shown for Cuda is much lower than for CPU. Are those tasks rare?

Bill, how many do you run concurrently on your nvidia card? So I can get an idea of speed.

And Georgi, any plans for OpenCL? My Tahitis are hungry. So little physics projects for GPUs.
34) (Message 7636)
Posted 29 Dec 2022 by (banished: ID 389916)
Post:
I see today there's version 102.14 appearing for everything except AVX. I've had 102.13 for everything up until now.
35) (Message 7635)
Posted 29 Dec 2022 by (banished: ID 389916)
Post:
I'm also curious about the choosing of the instruction set, because I have a Ryzen 3700X, and the only app version I am getting is sse2. Shouldn't I be getting sse3 or fma?
I fixed my Ryzen 9 3900XT. I reset the project, ran the Boinc benchmarks, and activated windows. One of those fixed it. I was stuck on sse3.
36) (Message 7634)
Posted 29 Dec 2022 by (banished: ID 389916)
Post:
This is why we encourage our contributors to upgrade their platform to the latest software platforms and driver packages if possible.
What in particular should we make sure is up to date? Since this is not a GPU project, that leaves that complication out. AFAIK there are no drivers to help CPUs behave better.
37) (Message 7633)
Posted 29 Dec 2022 by (banished: ID 389916)
Post:
Oh man, I hate those. I've seen what happened to Universe@Home when it was selected for a race. First people hoarding tasks and later, when the race started, overloading the servers to the point no real work could be done. I think in the long run those races do more harm than good, but that's just my opinion.
No, the server had recently been upgraded to SSDs and coped pretty good. They got three times the work done during the race.

However overall across all projects, I'm not sure if the competitions help. Maybe some people love competitions and this gets them more interested in Boinc?
38) (Message 7632)
Posted 29 Dec 2022 by (banished: ID 389916)
Post:
Well something worked. The offending machine is now trying AVX, FMA, SSE2, SSE3. I also notice version 102.14 (only on SSE2 and SSE3) which I've not seen on any other machines.
39) (Message 7631)
Posted 28 Dec 2022 by (banished: ID 389916)
Post:
It wouldn't have anything to do with the OS would it? Both are Windows 11 Pro, but the misbehaving one used to be Windows 11 Enterprise, HP branded installation. It still has the HP badge in system settings, but I "downgraded" it to Pro to try to fix something else. I also broke the activation, but it's activated again now. Will this help?
40) (Message 7629)
Posted 28 Dec 2022 by (banished: ID 389916)
Post:
I've reset the project, which gave me 129 sse2 tasks. I then ran Boinc's benchmark. I'll wait to see if the next batch gives me avx.

Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni ssse3 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 svm sse4a osvw ibs skinit wdt tce topx page1gb rdtscp fsgsbase bmi1 smep bmi2

The identical (apart from clock speed) CPU on the other machine gives precisely the same feature list.

I didn't mean for you to reset the whole server, I thought perhaps you could just delete the history of that host from it so it had to learn from scratch.


Previous 20 · Next 20