The few WUs on this host that I see that ran for 12 hours, using the Rosetta v4.15
x86_64-pc-linux-gnu application, and were all ended by the watchdog. I see the host has 32 CPUs and 32GB of memory. How many active R@h threads are running at the same time on this host?

Of course my comment is only about choosing between Intel and AMD CPUs for CPU only workloads.

Just some added settings. PBO on in bios, memory at 3600mhz. No other changes, most current bios and chipset drivers.
Edit:Oh Ya, x570 motherboard.

As of today AMD CPUs are simply the best for distributed computing, no matter what, be it Ryzen, Threadripper or EPYC. Their total computing power is simply unmatched.

If it is CPU only.
For GPUs, AMD are best if Double Precision is required (unless you want to forkout for workstation hardware). For Single Precision, Nvidia.

It also depends on if you can run all the date being processed in the gpu in question. If not then "he" is right. If I understand the way Rosetti is approaching the processing using larger databases instead of discrete data packages then that design for data-analysis will not show the true benefits of the massive parallel processing offered by gpus.

Tom M

I am using a 3800x AMD with a 1080 video card, 16gb memory and a couple ssd's. I run Boinc at 100% with 8 core/16 threads but to keep it cool and use less power I am going into power and sleep, then under the power plan settings I set it to power saver, then under advanced settings I set maximum processor state to 90%. Temps change from 78c(its throttling at 80c)to 55-57c and the cpu power use drops from 80 watts to 32 watts with Boinc and firefox running. This is according to Ryzen Master.
It lets me run 24/7 without any heat issues and saves on power to run it at 3475 mhz(90%) instead of all out at 4100-4150 mhz on 8 cores. Keeps the cpu voltages down to just under 1 volt as a side benefit. I only use high performance or balanced in power settings for some games.
This is with air and a aftermarket brick.

As of today AMD CPUs are simply the best for distributed computing, no matter what, be it Ryzen, Threadripper or EPYC. Their total computing power is simply unmatched.

If it is CPU only.
For GPUs, AMD are best if Double Precision is required (unless you want to forkout for workstation hardware). For Single Precision, Nvidia.

The other CPU? More like the main one these days!

Between a 2600 and a 2700, the 2700 is simply better for distributed computing. More cores, more WUs.

Remember that the Threadripper has 4 memory channels, while a Ryzen is dual channel, for memory intensive project maybe it would help (no idea about how that really matters in Rosetta)

Now this is the question I have. How much (if any) difference is there with this project in terms of dual-channel vs quad-channel memory setups. I'm looking to build a few rigs for this project and would love to have such information to plan my builds.

Yes, I agree that the Ryzen 7 and 9 are one of the top contenders.

It's interesting how I first thought the Threadripper3 wouldn't be worth it, but seeing this review and plugging in the parameters into the formula convinced me that this one crunches well, too.

Based on FLOPS, this one should produce 100k RAC for ~$5k TCO ($8-$10/GFLOPS).

Although, we seem to have forgotten to consider that cores within a single motherboard also share memory bandwidth that could impact overall productivity. Not everything fits into the cache you know.

We also seem to have waved away the potential interest rate lost over the years on large upfront investments and the effect of inflation on resale value. Both of these favor a bit less energy efficient specimens with a bit higher power consumption. Hence we should refine the original formula measuring efficiency, I guess.

If you're looking for a new system, then a Ryzen 3000 is the only way to go if you plan to get it in the next 6 months or so. Then Ryzen 4000 for the desktop is due for release (rumours are late July, early Sept) & early leaks indicate it has a significant improvement in performance even over the Ryzen 3000 series.


At present- if money is tight then Ryzen9 or Ryzen 7.
For insane performance (and a price to match) Threadripper3

What motherboard chipset would you pair that with?

My Ryzen 2600 is on an A320 but if I replace my Piledriver set up with a Ryzen 4000 would I be better with a B450 or X570 or ???

[quote]This topic has reminded me of things I wanted to say over the last few weeks.

Dear Sid,
You are both on target and off target.

I had nothing but errors on both the i686 applications on my Ryzen. Gave up on Rosetta and moved to Einstein. Discovered later that you can set a flag in cc_config.xml to ignore alternate platforms.
<no_alt_platform>1</no_alt_platform>
That would have told the Rosetta scheduler to not send me x86 applications and just send me the x86_64 applications.

I decided to try that on my Windows 10 computer. Surprise - if Windows 10 even includes IE, it is very well hidden.

Due to how power saving and turbo boost is implemented in processors, it is much more efficient to run 50% of the time over every core than to use 50% of the cores.

Let me take my best shot at these, I cannot confirm the status of the original proposal.

I had a quick look at some slot directories after getting the 4.15 Rosetta app on a Raspberry Pi today.

It looks to me, and I could be wrong, the app unzips the database and other files into the slot directory when it first starts. Around 1.5 to 1.8GB of files, for each and every task start. This is going to be a killer on the Raspberry Pi’s and other SD card based systems. They don’t support trim and have minimal wear levelling, not to mention they are much smaller capacities (8 or 16GB cards are common) so have much less spare blocks than say a 256GB SSD. It’s also time consuming doing this for each task start.

If I recall from a couple of years ago GPUgrid used to do the same thing, except they were copying CUDA libs which were also huge. I believe BOINC has the ability to have static files in the project directory that can be sym-linked from the slot directory which is a heck of a lot more efficient than copying them every time. This of course is only suitable if the files are not being updated by the task. Maybe the app developers can look into this.