I do wonder which would be more efficient: a bunch of small (2T) Nbody tasks, or a smaller number of larger (4T) tasks, or a single very large (8T) task. Right now I've got mine set up as 2 avg cpus, 2 threads per task for the command line, and 4 max concurrent tasks for 4 tasks, 8 threads taken up by MW@H. Guess I'll have to experiment and see how my RAC changes.

I use a quicker method, I look at a CPU usage graph (task manager, MSI Afterburner, whatever you prefer). I use Afterburner because it shows GPU aswell. I try to make both be very busy.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Update... For people looking for Astrophysics related projects for their GPU's ...

Asteroids@home has NVIDIA app's for the following system groups

Microsoft Windows (98 or later) running on an Intel x86-compatible CPU
Microsoft Windows running on an AMD x86_64 or Intel EM64T CPU
Linux running on an AMD x86_64 or Intel EM64T CPU

Enjoy
Bill F

They're no faster than CPUs, so pointless.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

And yet I play fine in HD with an old 8000GFlops Nano.

What games? What display resolution?

I thought that was obvious. Two cards, twice the computing power, should achieve twice the calculations and render twice as many frames in the game, just like a twice as fast card.

Both cards are not running to their full potential.

So why can't I use those rates to estimate how fast they are at games?

Don't argue with me. Go look at gaming benchmarks with the 4090 and 7900XTX comparisons. You'll see the 4090 has better frames in nearly every, single game.

I think it's more to do with things getting so much more complex, programmers divide things into chunks. Just like building a massive building, everyone works on a different part, then finds something the plumber wants to do ruins the structural integrity.

Not exactly. Building games for more developers is like having a building already built (the game engine) and then populating that building with your assists (the content in the game). Unless the game is built with SLI/Crossfire in mind, you wont see any performance gains. Nvidia doesn't even included SLI bridges on most of their new cards anymore because it wasn't worth the little gains.

Because you said you hated Nvidia. And since AMDs are cheaper the three of them wouldn't set you back much more.

Not all games support Crossfire. Even the games that do may not optimize it. I've done the whole multiple GPU gaming setups before in the past. It's not worth the headache for the little gains you get in the games that support it.

A larger screen requires a higher resolution to be sharp.

Well no kidding. But you said, and I quote "a 100 inch 8k display" where you specifically stated the resolution of the display. 100 inch is irrelevant. As I said before. Just saying "an 8k display" would be the same as what you said. Because it doesn't matter if it's an 10in 8k display or a 100in 8k display. It's 8k and takes orders of magnitude more processing power to render than 1080P.

Ah, you live in another world, one where people have an extra zero on their salary.

Regardless of how many zeros I have. AMDs cards can not process their 1s and 0s as fast as Nvidia's flagships. They have been getting extremely close and that closeness has become even tighter with AMDs release of the 7000 series, but they're not there yet. Maybe in a few years AMD will have something that I will give their GPU drivers another shot at on whatever rig I am gaming on at that time.

For now, 7950X and 4090 is the best setup for me. (Note that I also run BOINC on the computer and other production stuff, so having the extra cores over the "real" gaming CPU with less cores from AMD is a better solution for me.

And yet I play fine in HD with an old 8000GFlops Nano.

What games? What display resolution?

Any games, they are all pretty similar. I already specified 3D shooting.

I thought that was obvious. Two cards, twice the computing power, should achieve twice the calculations and render twice as many frames in the game, just like a twice as fast card.

Both cards are not running to their full potential.

I see no reason they shouldn't.

So why can't I use those rates to estimate how fast they are at games?

Don't argue with me. Go look at gaming benchmarks with the 4090 and 7900XTX comparisons. You'll see the 4090 has better frames in nearly every, single game.

So you don't know.

I think it's more to do with things getting so much more complex, programmers divide things into chunks. Just like building a massive building, everyone works on a different part, then finds something the plumber wants to do ruins the structural integrity.

Not exactly. Building games for more developers is like having a building already built (the game engine) and then populating that building with your assists (the content in the game). Unless the game is built with SLI/Crossfire in mind, you wont see any performance gains. Nvidia doesn't even included SLI bridges on most of their new cards anymore because it wasn't worth the little gains.

We were discussing why programmers make mistakes.

Because you said you hated Nvidia. And since AMDs are cheaper the three of them wouldn't set you back much more.

Not all games support Crossfire. Even the games that do may not optimize it. I've done the whole multiple GPU gaming setups before in the past. It's not worth the headache for the little gains you get in the games that support it.

I'm looking for an answer to why there isn't 100% gain.

A larger screen requires a higher resolution to be sharp.

Well no kidding. But you said, and I quote "a 100 inch 8k display" where you specifically stated the resolution of the display. 100 inch is irrelevant.

Tautology, it helps in arguments.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Gpu asteroids tasks finish in approximately 20 minutes minutes for me.
Maybe your boinc client incorrectly estimated time?

Gpu asteroids tasks finish in approximately 20 minutes minutes for me.
Maybe your boinc client incorrectly estimated time?

I don't have and never will have a Nvidia. I'm going by what others have reported as times.

Looking at your tasks, this one https://asteroidsathome.net/boinc/workunit.php?wuid=168597146 was run on your GTX 1650 in 1293 seconds and also on your wingman's Xeon W3565 CPU in 7891 seconds.

Assuming you're running one at a time on the GPU....

His CPU has 8 threads, so can actually do one every 986 seconds. So your £90 GPU is slightly slower than his £7 CPU.

Your CPU is 3.5 times faster than his, see https://www.cpubenchmark.net/cpu_list.php so you could do one every 300 seconds on your CPU, over 5 times faster than your GPU. GPUs should be 50 times faster than CPUs. Pick another project for the GPU and run Asteroids on your CPU.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Maybe they want to run N-Body with the CPU and keep the idle GPU running on Asteroids

Maybe they want to run N-Body with the CPU and keep the idle GPU running on Asteroids

Possibly, but I'd prefer the GPU on Einstein where it can do work colossally fast. You don't help to shift a million tonnes of goods with a car. Sure, you're helping, but not much. Use the car for carrying passengers, what it's intended for.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Gpu asteroids tasks finish in approximately 20 minutes minutes for me.
Maybe your boinc client incorrectly estimated time?

I generally do a Asteroid CUDA gpu task every 2-3 minutes on all my hosts.
Run at 0.5 gpu_usage.

---

No it's for any of my multi-core pc's that are also running tasks from other projects and I'm limiting it to 6 cpu cores for Milky Way to use. I prefer the 3 double cpu tasks because they take longer for me to crunch and they also don't get hung up like the 16 cpu core tasks do.

So it's specific to your requirements. I have no problem with any Nbody right up to 16, they never get stuck. I also don't see the point in making them take longer to crunch.

and there's no point in crunching them faster unless you are in it for the credits, which I'm not.[/quote]

No I want to run tasks efficiently and without cranking up the heat, an 8 core task produces more heat than 4 dual core tasks. Now obviously that;s not possible at every project but where I can I cut the crunching to fewer cores per task.

I never run both cpu and gpu tasks from the same project on my pc's

Why not? I do if they're different things. Pointfull here, pointless at Einstein (where they can make any of their work run on a GPU).

Because Boinc is absolutely terrible at managing the cache and considers each project as a whole and not each app separately. ie the gpu will run out of work while the cpu still has dozens of tasks to run and Boinc is sitting there saying 'no new work is needed cache is full'. Even if I increase the cache size I get more cpu tasks and still no gpu tasks. Mostly, not always but mostly I run gpu tasks at project that have them and cpu tasks at projects that don't have gpu tasks, but with projects like Milky Way having a different app for the gpu's then I will run MW cpu tasks on this pc and MW gpu tasks on that pc.

No I want to run tasks efficiently and without cranking up the heat, an 8 core task produces more heat than 4 dual core tasks. Now obviously that;s not possible at every project but where I can I cut the crunching to fewer cores per task.

Isn't heat a good indicator of how much work you're achieving?

Because Boinc is absolutely terrible at managing the cache and considers each project as a whole and not each app separately. ie the gpu will run out of work while the cpu still has dozens of tasks to run and Boinc is sitting there saying 'no new work is needed cache is full'. Even if I increase the cache size I get more cpu tasks and still no gpu tasks. Mostly, not always but mostly I run gpu tasks at project that have them and cpu tasks at projects that don't have gpu tasks, but with projects like Milky Way having a different app for the gpu's then I will run MW cpu tasks on this pc and MW gpu tasks on that pc.

I've never seen Boinc be that stupid, it maintains a seperate cache for CPU and GPU. The only thing it gets wrong is the estimated time, so it can download far too much of one. But it doesn't stop the other getting work.

If I look in the messages tab, I see things like "Not requesting work, GPU cache full", indicating there is a seperate cache.

The only weird project is Gerasim, where the project refuses to give you work for GPU unless you take some CPU at the same time. But that's not my Boinc client doing that.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

We would appreciate your input on this because we expect that it will probably take some time for GPU-oriented users to swap that hardware over to different projects. How long would you like us to wait before we shut down Separation?

It takes 10 seconds to switch to another project.
Just do it and don't let people who do not read the forum waste their resources!

The only weird project is Gerasim, where the project refuses to give you work for GPU unless you take some CPU at the same time. But that's not my Boinc client doing that.

I have the cpu tasks unchecked and DO still get gpu tasks when they have them

The only weird project is Gerasim, where the project refuses to give you work for GPU unless you take some CPU at the same time. But that's not my Boinc client doing that.

I have the cpu tasks unchecked and DO still get gpu tasks when they have them

Ok maybe you have to do some CPUs first. I started with GPU and it just said 0 tasks. When I switched on CPU aswell it gave me 16 of each.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

No I want to run tasks efficiently and without cranking up the heat, an 8 core task produces more heat than 4 dual core tasks. Now obviously that;s not possible at every project but where I can I cut the crunching to fewer cores per task.

Isn't heat a good indicator of how much work you're achieving?

Maybe but when the electric bill gets too high I have to stop down crunching, this way I do the same thing but keep on crunching.

I know people who are using liquid cooled system who have lower pc temps than I do and are crunching MUCH faster, one guy's system is running cooler than ambient temp while still crunching flat out.

Maybe but when the electric bill gets too high I have to stop down crunching, this way I do the same thing but keep on crunching.

Why not just change the number of cores used? There are many easier methods of lowering the power bill.

I know people who are using liquid cooled system who have lower pc temps than I do and are crunching MUCH faster, one guy's system is running cooler than ambient temp while still crunching flat out.

Does he overclock?

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Maybe but when the electric bill gets too high I have to stop down crunching, this way I do the same thing but keep on crunching.

Why not just change the number of cores used? There are many easier methods of lowering the power bill.

I know people who are using liquid cooled system who have lower pc temps than I do and are crunching MUCH faster, one guy's system is running cooler than ambient temp while still crunching flat out.

Does he overclock?

I don't know, I do know he optimizes his bios and system settings

Does he overclock?

I don't know, I do know he optimizes his bios and system settings

I used to do those and overclock, then gave up as it caused too many crashes. It's hard enough keeping things stable at the normal speed. Some new RAM I bought only runs at 85% speed in one machine, but 100% in another. Different combinations of hardware don't like each other.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.

Does he overclock?

I don't know, I do know he optimizes his bios and system settings

I used to do those and overclock, then gave up as it caused too many crashes. It's hard enough keeping things stable at the normal speed. Some new RAM I bought only runs at 85% speed in one machine, but 100% in another. Different combinations of hardware don't like each other.

Yes I know, I have some new ram to put into a pc because it's got a combination of ram speeds in it now and the new stuff is faster than the slowest stuff but not as fast as the fastest stuff, and yes I know about them all running at the slowest speed that's why I bought the new stuff. I will spread the old ram out between my other pc's trying to keep speed in mind.

Yes I know, I have some new ram to put into a pc because it's got a combination of ram speeds in it now and the new stuff is faster than the slowest stuff but not as fast as the fastest stuff, and yes I know about them all running at the slowest speed that's why I bought the new stuff. I will spread the old ram out between my other pc's trying to keep speed in mind.

That's what I was trying to do, along with the problem of two RAM types - some go fast due to Intel XMP (something to do with raising the voltage for more speed), some are just fast and don't do XMP. But if you get a 3200MHz XMP and a 3200MHz normal and mix them, they don't go 3200MHz. The MB switches off XMP because one of them can't do it, and the XMP one can't go full speed without XMP, so they now both go about 2600MHz. Anyway my strategy didn't work to swap RAM about, because I'm limited to which computers like which RAM! My best machine crashes with the new RAM, yet it works fine on another, so my best machine now has slower RAM in it, but everything has lots of RAM which is the main thing for Boinc.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.