I'd go for the card with bigger memory bandwidth as traditionally this is important in this project in the long run.
Regardless of what card I go with I will be taking a drop in memory bandwidth from my 3080. In regards to bandwidth AMD are winning the race by a slim margin about 16.8 GB/S
I would stop at the 535 driver level. The latest 545 series is wrought with many bugs and complaints from users.
I tried loading the 535 driver through apt and got an "Unable to locate package nvidia-driver-535" error. On looking through the Ubuntu 22.04LTS package list, I see the most recent package that would be available to me through apt is nvidia-driver-510. Will that be good enough to run compute functions on a RTX 4060Ti?
Ideas are not fixed, nor should they be; we live in model-dependent reality.
I would stop at the 535 driver level. The latest 545 series is wrought with many bugs and complaints from users.
I tried loading the 535 driver through apt and got an "Unable to locate package nvidia-driver-535" error. On looking through the Ubuntu 22.04LTS package list, I see the most recent package that would be available to me through apt is nvidia-driver-510. Will that be good enough to run compute functions on a RTX 4060Ti?
What kernel are you using? You may not be getting the 535 driver because of a lower/lessor kernel, so be sure to update that first. As I said, kernel 5.19 should work fine with NVIDIA driver 535.
Also, be sure to "purge" your AMD drivers first, and then reboot, and then load NVIDIA driver 535.
What kernel are you using? You may not be getting the 535 driver because of a lower/lessor kernel, so be sure to update that first. As I said, kernel 5.19 should work fine with NVIDIA driver 535.
Also, be sure to "purge" your AMD drivers first, and then reboot, and then load NVIDIA driver 535.
The auto-updater ran a couple of days ago, so I am running the 6.2 kernel.
Thanks for purge/reboot/load tip. AMDGPU has already been scrubbed from the system.
Ideas are not fixed, nor should they be; we live in model-dependent reality.
the GW tasks are very CPU biased/sensitive. about half, or more, of the total runtime is purely done on the CPU. so faster CPUs will finish this half faster. GW tasks will crunch on the GPU for 0-99% on BOINC progress reporting. and at that point you will notice that the task does not stop, but the GPU will be idle while the CPU keeps going on a single thread. when I tested this a while back (before they changed the app recently) it took about 8 mins to crunch the GPU, then another 8 mins for the final 1% on the CPU only. for a total of 16mins. if you keep the CPU the same, that last 1% will take the same amount of time no matter what GPU you have.
re: new cards and efficiency, the titan V does break the mold there. I don't know what magic is happening with the titan V and BRP7 with our optimized app, but on my systems, the titan V is a tad faster than a 3070Ti, and uses only 120W doing it. ppd/W was about on par with 4090s that I've seen. (about half the speed of a 4090, but half the power draw, and 1/3rd the cost). but that's specific to the BRP7 project. the 40-series cards seem to have a nice clock speed and efficiency advantage for several other projects like GPUGRID, PrimeGrid (especially PG!), and Asteroids.
I notice that is also takes some time to load the data into the VRAM. It takes about 2mins on my computer. What is my computer doing during that time?
And about the CPU-intense time afterwards, I heard that the CPU need to sort the candidates. Then there will be some trivial solutions to release that problem. By implementing your own sort function that create two threads to call something like std::sort in the divide-and-conquer process, you can have a roughly 2x speedup when you notice that the recursion depth, O(logn)>>1. But on Google it seems that Merge Sort benefits more from this. I don't know why. Besides, you can skip the final few recursions to get a roughly sorted result, which is still acceptible.
I currently have a 3080. I am starting to think about what my next upgrade will be sure to go for either a 4080/5080 if they are out sometime next year or should I change and go with a 7900 XT? (AMD)
I don't think it will make a difference but I am running Windows 11 pro
Is it worth waiting for the 4080 super card in early January or should I just get a 4080?
I notice that is also takes some time to load the data into the VRAM. It takes about 2mins on my computer. What is my computer doing during that time?
And about the CPU-intense time afterwards, I heard that the CPU need to sort the candidates. Then there will be some trivial solutions to release that problem. By implementing your own sort function that create two threads to call something like std::sort in the divide-and-conquer process, you can have a roughly 2x speedup when you notice that the recursion depth, O(logn)>>1. But on Google it seems that Merge Sort benefits more from this. I don't know why. Besides, you can skip the final few recursions to get a roughly sorted result, which is still acceptible.
I think you need to PM Bernd or one of the other Developers to bounce your ideas off them, most of us in here are just users of the software and not the ones who can actually make any changes to the Applications we run.
I notice that is also takes some time to load the data into the VRAM. It takes about 2mins on my computer. What is my computer doing during that time?
We would need to "see" your hidden computers to even speculate why it is taking 2 minutes to load the data into VRAM. Are you saying it takes 2 minutes before your system starts crunching the task?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
B.I.G wrote: I'd go for
)
Regardless of what card I go with I will be taking a drop in memory bandwidth from my 3080. In regards to bandwidth AMD are winning the race by a slim margin about 16.8 GB/S
Keith Myers wrote: I would
)
I tried loading the 535 driver through apt and got an "Unable to locate package nvidia-driver-535" error. On looking through the Ubuntu 22.04LTS package list, I see the most recent package that would be available to me through apt is nvidia-driver-510. Will that be good enough to run compute functions on a RTX 4060Ti?
Ideas are not fixed, nor should they be; we live in model-dependent reality.
Maybe you hadn't done a apt
)
Maybe you hadn't done a apt update in a long time? Or maybe you don't have the proprietary package sources selected in your sources.list?
$ apt show nvidia-driver-535
Package: nvidia-driver-535
Version: 535.129.03-0ubuntu0.22.04.1
Priority: optional
Section: restricted/libs
Source: nvidia-graphics-drivers-535
Origin: Ubuntu
Maintainer: Ubuntu Core Developers <ubuntu-devel-discuss@lists.ubuntu.com>
Bugs: https://bugs.launchpad.net/ubuntu/+filebug
Installed-Size: 1,734 kB
Download-Size: 487 kB
APT-Manual-Installed: yes
APT-Sources: http://security.ubuntu.com/ubuntu jammy-security/restricted amd64 Packages
Description: NVIDIA driver metapackage
To use the new Petri app you need at least the 525 drivers.
510 version is not new enough to recognize a RTX 4060 Ti.
If you can't get the drivers from Ubuntu you can also add the graphics-drivers PPA and get it from them.
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update
sudo apt install nvidia-driver-535
cecht wrote: Keith Myers
)
What kernel are you using? You may not be getting the 535 driver because of a lower/lessor kernel, so be sure to update that first. As I said, kernel 5.19 should work fine with NVIDIA driver 535.
Also, be sure to "purge" your AMD drivers first, and then reboot, and then load NVIDIA driver 535.
Proud member of the Old Farts Association
GWGeorge007 wrote: What
)
The auto-updater ran a couple of days ago, so I am running the 6.2 kernel.
Thanks for purge/reboot/load tip. AMDGPU has already been scrubbed from the system.
Ideas are not fixed, nor should they be; we live in model-dependent reality.
Ian&Steve C. wrote: the GW
)
I notice that is also takes some time to load the data into the VRAM. It takes about 2mins on my computer. What is my computer doing during that time?
And about the CPU-intense time afterwards, I heard that the CPU need to sort the candidates. Then there will be some trivial solutions to release that problem. By implementing your own sort function that create two threads to call something like std::sort in the divide-and-conquer process, you can have a roughly 2x speedup when you notice that the recursion depth, O(logn)>>1. But on Google it seems that Merge Sort benefits more from this. I don't know why. Besides, you can skip the final few recursions to get a roughly sorted result, which is still acceptible.
Speedy wrote: I currently
)
Is it worth waiting for the 4080 super card in early January or should I just get a 4080?
seewo wrote: I notice that
)
I think you need to PM Bernd or one of the other Developers to bounce your ideas off them, most of us in here are just users of the software and not the ones who can actually make any changes to the Applications we run.
seewo wrote: I notice that
)
We would need to "see" your hidden computers to even speculate why it is taking 2 minutes to load the data into VRAM. Are you saying it takes 2 minutes before your system starts crunching the task?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Speedy wrote: Is it worth
)
Nobody can answer that, if you can wait then wait and you'll see in January if it's worth the money or not.
But why not go for the AMD card? According to the hosts I posted earlier in this thread the 7900xtx is both: faster and cheaper than the 4080.