Hi everyone,
Noticed that RAC varied a lot lately, while keeping the same video cards.
The 'waves' are pretty high, and I am curious why.
So, about 1 year ago, RAC was around 200.000. Then, went up to about 340.000. Now, it's down to around 160.000.
These numbers indicate a big change in something. So what has changed, that pushed the RAC up and down so much?
Thank you and happy crunching!
Copyright © 2024 Einstein@Home. All rights reserved.
With the configuration you
)
With the configuration you are running most of your credit comes from GPU tasks of which there are two primary flavors. Generally speaking, the Einstein gravity wave tasks award less credit per unit time then do the GammaRay pulsar tasks.
Also the gravity wave tasks are quite variable in computational characteristics and credit award rate even within the same day.
So your RAC will vary depending on what flavor of tasks you happen to receive from time to time. Before Gravity Wave work appeared for our GPUs the Gamma-Ray work had much less short-term variability.
John wrote: Noticed that RAC
)
In addition to what ARCHAE86 has said, I have some thoughts on the subject.
If you notice, the number of users in E@H has increased by ~500 since the halting of S@H, but I can only imagine that the individual "wingman" has not changed. This may cause the workload to increase and thereby delaying what we receive as credit. The RAC would be potentially lower because of the delays.
Just my thoughts...
Proud member of the Old Farts Association
OK, thank you, got it. Though
)
OK, thank you, got it. Though I never thought the variations can be so big.
Archae86, why do you have a double RAC on the 2nd comp, compared to the 1st? The cpus are quite similar, the gpus are similar, and yet, the RAC is double! Is it the same thing, about the 'flavour' that comp 1 and 2 get?
The cpus are NOT similar. i5
)
The cpus are NOT similar. i5 versus an i3. 16 cores versus 4 cores. I assume he runs multiples on the higher RAC computer because it has plenty of cpu cores to support the extra gpu work units running.
... I wonder what that [2]
)
... I wonder what that [2] on the i5 means ...
John wrote:Archae86, why do
)
I have three computers working for Einstein. Currently all three use Radeon RX 5700 GPUs, but there are important configuration or operating differences among them.
In case people wish to make comparisons elsewhere, I should also mention that all four GPUs are appreciably throttled down for power limitation preferences. But that is pretty similar among these three, so not the source of big differences.
The most productive machine 12260865 runs two 5700 GPU cards, exclusively on GPU Gamma Ray Pulsar work. While it reports itself to have 16 processors, it actually only has four. I taught it to report a falsely high number as a trick to get an adequate supply of GRP tasks to keep it busy. The CPU is no longer new, but still pretty fast.
The other two machines each run a single 5700 GPU. However the lower scoring one 10706295 has been running only Gravity Wave GPU tasks for a few weeks. I am still fiddling with it to run batches of tasks at 2x up through 5x to get extra throughput. When I tire of that game and just run it 2X all the time, it will coast down to a considerably lower RAC than the current report, which still has some memory of the good old days running GRP tasks. That one is honestly reporting a 6-CPU processor (no HT), but a rather slow though modern one.
The higher scoring of the single GPU machines 10659288 is running Gamma Ray GPU tasks only. It has a middle-aged two-core processor which is reported as 4 because it runs HT. This is by a big margin the least capable CPU among my three.
While my CPUs vary, that has little effect on current Einstein GRP GPU tasks. However the current GW GPU tasks are CPU hungry, so the capability of the host system CPU probably matters a lot more.
I don't run any CPU tasks. I don't run any distributed computing elsewhere than Einstein. I am very happy with my 5700 cards.
Thank you for the
)
Thank you for the details.
After comparing some more comps, it's still something strange. It doesn't really make sense that a machine with this CPU AuthenticAMD AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0]
(8 processors) and this GPU AMD Radeon RX 580 Series (4096MB) does 450.000 RAC and my comp, which is much stronger (both cpu + gpu) has 160.000 RAC. Just doesn't make sense.
Why do I have the feeling these cases shouldn't exist?
Another example: comp with 2 AMD R9 380 does about 1.000.000 RAC, my GPU (R9 Nano) does only 160.000 when it should be around 500.000 (or higher!). So please someone help me clear this, I don't like to waste the power (PC + electrical) due to some incorrect settings. Could this have come from some updates of Boinc soft?
Related to what Archae86 said, I've noticed that for quite a while I have only received Gravi wave search tasks (long names starting with "h1"). Before this, I had the other type of GPU tasks. And now I'm wondering why I don't get those tasks anymore. Is there something I can/should do, from the setting maybe? Or it's not up to the user?
Keith Myers wrote: The cpus
)
they are more similar than what is shown. he's spoofing the core count on the i5 for one reason or another.
his i3 chip is a dual core with HT (2c/4t)
his i5 chip is a quad core without HT (4c/4t)
the i5 is probably still close to 2x the performance of the i3 though with the clock speed advantage.
_________________________________________________________________________
John wrote:Related to what
)
You can influence what types of tasks are sent to you in general classes, and also select specific applications.
Go to the details page for your computer and determine what location (aka venue) it is assigned to among generic|home|work|school.
Go to the preferences section of your account.
Select the Project preferences page and scroll the Preference Set selection to match your computer's location.
Near the top you can choose to accept CPU or GPU tasks generally.
In the Applications section you can choose to allow only specific applications.
In the Beta section, if you choose to allow test applications, that may override an exclusion, so for maximum control you may wish to set that choice to No.
When you have the settings as you wish, click "save changes" at the bottom.
Sorry, really know very
)
Sorry, really know very little about Intel cpus and their nomenclatures and architectures.