Administrator ported boinc server to windows server.

I wish the Admin would go in and fix the venues!!! I want to run cpu tasks on some pc's and gpu tasks on other pc's and because of the way they setup the venues I can't

Like I told you already.
You don't have many hosts, not hard to figure out which one you probably were having issues with.

I doubt many folk would look through several hosts to help someone, they'd ask.

Because you don't listen the first time and think what you know is factual.

If it wasn't a fact I wouldn't think it was.

There you go again. Using useless metrics. Pick a CPU project, any CPU project. Attach your single 3900X to it and I will attach my single 64c/128t EPYC. I guarantee I do more than 2.4 times the points as you.

I've checked between my hosts, some are a lot faster than others, look at my 4 core ones.

I am assuming your hosts in another building are your Xeons and not your Ryzen which is probably your main computer you have inside your home.

I have two Ryzens. The X is in the garage, the XT is the house one. The garage also has two xeons and four misc 4 core machines.

EPYC motherboards have IPMI remote management. You can control everything on the host, including the BIOS settings, remotely using IPMI. Xeons have this ability too on some of the newer stuff. I don't think the older Xeons have IPMI and I'm not sure what year/generation they started using IPMI.

There's a remote thing in the BIOS of the xeons, I thought all servers did that. I don't bother using it as all I need access to is Windows or Boinc via Boinctasks. If it screws up so I can't get to it on the network, there's a keyboard, mouse, monitor in the garage I can use.

Like I said. You can use the one GPU for all the hosts to edit the BIOS the way you need it once. However, if you overclock those hosts some projects will put way more stress on that overclock than other projects where the overclock will fail. Its rare, but it does happen. Especially some of the Primegrid sub projects.

So you often change the overclock depending what project you run? I don't overclock, more hassle than it's worth. The only thing I turn above stock is the power limit on GPUs, and that's easy to adjust in MSI Afterburner in Windows.

Looking through the Universe@home hosts list I found a 128 thread EPYC here:

Shy about your own hosts I see.

https://stats.free-dc.org/host/uni/561422
It does around 1.3M PPD.
Here is a 3950X (little more powerful than your 3900X)
https://stats.free-dc.org/host/uni/602386
It does around 370K PPD.

370,000 * 2.4 = 888,000
370,000 / 1,300,000 = 3.5

In this example we can see that an EPYC is 3.5 times more productive than a 3700X. That means you'll need 4 3900X CPUs.

Not enough information. Link to hosts on the actual project, then we can see how long they take per task. You don't know if they're only doing Universe. Also is it hyperthreading or not? I have HT on.

Also having more PCs makes it easier to attach lots of GPUs.

Lets take a look at what full load on the 3900X does power wise.

Not interested in efficiency. I have a 21 year old car. I could buy a newer one and use less fuel, but I can't be bothered, and it would cost more to buy.

So we have to talk in simpler terms so your simple mind can understand?

Contrarywise, I assumed "capacitor" when he meant "limit".

Wasn't boasting. Just correcting someone wrong on the Internet.

You boast when you state your specs.

Hidden only hides your list. The host is still listed and can be found on the project host list. It just won't have a name attached to it. It will say anonymous.

Actually it just omits the name. Anonymous refers to a host without listing the owner. Not sure why it hides the name since I've said to not hide anything. It would be easier to say "the computer called Jenny is crashing tasks frequently" and everyone would know which one it was. Now I have to say which one by other means, like "the Xeon, no not that one, the other one, er.... the one with slightly less credits."

I don't have to work in IT to know a lot of about computers. Been playing on computers for over 20 years now.

Absence of denial of being a binman noted.

---

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

Administrator ported boinc server to windows server.

Are you telling me Boinc didn't create a Windows version? Ugh! It would be more popular if people didn't have to learn Linux to create a project.

---

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

I wish the Admin would go in and fix the venues!!! I want to run cpu tasks on some pc's and gpu tasks on other pc's and because of the way they setup the venues I can't

I guess you have to run two Boinc clients.

---

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

Two project accounts?

Two project accounts?

That's a good idea, but Mikey likes stats and probably wants credits from both machines to be together.

The app/cc configs could be used to prohibit GPU tasks from that project, not sure if it would stupidly download one then not to do it. Also, not sure if you can prohibit CPU tasks that way.

---

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

If it wasn't a fact I wouldn't think it was.

Gold Medal in Mental Gymnastics

Not enough information. Link to hosts on the actual project, then we can see how long they take per task. You don't know if they're only doing Universe. Also is it hyperthreading or not? I have HT on.

check mine.
https://universeathome.pl/universe/show_host_detail.php?hostid=664091

it's two 64-core CPUs. HT is on, so that's 128 threads each, and 256 threads total for the system.
the only project the system runs is Universe, and it crunches at 90% CPU utilization (roughly 230 threads) 24/7
~3.0M ppd total, ~1.5M ppd each CPU.

comparing cherry picked tasks from hosts is pointless since universe has varying task lengths. they are not all the same so it's easy to end up comparing two tasks that aren't comparable.

go nuts and cope.

---

If it wasn't a fact I wouldn't think it was.

Gold Medal in Mental Gymnastics

Not so, I only believe what I see to be true. As below, I believed cpubenchmark.net because it lined up with the speed of my own machines on most projects. I don't believe god exists, even though half the world tells me he does, because I see no evidence. Same for climate change.

check mine.

Ah, not as shy as Skillz then!

https://universeathome.pl/universe/show_host_detail.php?hostid=664091

it's two 64-core CPUs. HT is on, so that's 128 threads each, and 256 threads total for the system.
the only project the system runs is Universe, and it crunches at 90% CPU utilization (roughly 230 threads) 24/7
~3.0M ppd total, ~1.5M ppd each CPU.

comparing cherry picked tasks from hosts is pointless since universe has varying task lengths. they are not all the same so it's easy to end up comparing two tasks that aren't comparable.

go nuts and cope.

Two CPUs, cpubenchmark.net says they're about twice as fast (each) as my Ryzen 9 3900XT.

Your tasks take 3000+ seconds (and you do 115 at a time per CPU), equivalent to one completed every 26 seconds per CPU.

My tasks take 5400+ seconds (and I do 24 at a time per CPU), equivalent to one completed every 225 seconds per CPU.

So each of your CPUs are 10 times faster not 2.

Looks like cpubenchmark sux for big CPUs. It's very accurate for my lot (look how much slower my 4 core machines are) on almost every project.

So do you know a better place to get a rough speed idea? Because it's needed to choose new CPUs.

Looking here, comparing our CPUs, I find no benchmark giving a greater difference than x3: https://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_9_3900xt-vs-amd_epyc_7742

---

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

I doubt many folk would look through several hosts to help someone, they'd ask.

I don't know. Guess they just didn't need to know the host to fix the issue. Which most issues, you don't have to see the host's results anyway.

If it wasn't a fact I wouldn't think it was.

Ignorance is bliss.

I've checked between my hosts, some are a lot faster than others, look at my 4 core ones.

You can't be this stupid. OF COURSE the hosts are going to be faster/slower on a different CPU!!!! You're just trolling now or you really have even less of a clue than I originally thought.

I have two Ryzens. The X is in the garage, the XT is the house one. The garage also has two xeons and four misc 4 core machines.

I bet my single 64-core, 128 thread system can do more work than your entire fleet. Still. "But if you buy a lot of them, they'll do more than the EPYC. EPYC is a waste of money, blah, blah, blah" Sure, kiddo. All while using less electricity doing it. Then again, I pay for my power unlike people like you who steal power.

There's a remote thing in the BIOS of the xeons, I thought all servers did that. I don't bother using it as all I need access to is Windows or Boinc via Boinctasks. If it screws up so I can't get to it on the network, there's a keyboard, mouse, monitor in the garage I can use.

Older stuff requires propriety software to use, AFAIK. IPMI loads in a web browser. Super simple to use.

So you often change the overclock depending what project you run? I don't overclock, more hassle than it's worth. The only thing I turn above stock is the power limit on GPUs, and that's easy to adjust in MSI Afterburner in Windows.

If you don't overclock. Then you don't know. Worthless to discuss.

Some Primegrid projects would probably result in invalid results from overclocking GPUs, too. ;)

Shy about your own hosts I see.

My 128 thread EPYC isn't running Universe@home, so no point in showing you the stats from it as it'll show 0.

Not enough information. Link to hosts on the actual project, then we can see how long they take per task. You don't know if they're only doing Universe. Also is it hyperthreading or not? I have HT on.

You can enter the CPID from that link into the URL on the project site to go directly to the hosts link on the project. Probably too complicated for simple minds though.
Not sure what the argument is here. Sure, those hosts could be running other stuff, we will never know and looking at the project web site wont tell you otherwise. As noted previously, Universe@home tasks vary in their run times. Its nearly impossible to tell how many tasks a host can do in a single day.

If you look at those pages you can see it shows the credit the host earned each day on the last 28 days. It'll give you a good idea if the host is only running that project.

Also, going to that page you can see the CPID of the host which will show all the projects it's been working on. For example, the 3900X I linked is here:
https://stats.free-dc.org/host/uni/stats.php?page=hostbycpid&cpid=534cc09b43343501d435553480ad3d75

It only shows it's running Universe.

Meanwhile, the 64-core EPYC is running other projects.
https://stats.free-dc.org/host/uni/stats.php?page=hostbycpid&cpid=1c921357777e4c191d99a404a3373781
E@H on the GPUs I am sure.
It's also returning results for yoyo@home, tn-grid and asteroids (possible GPU), but still out performing the 3900X by a large margin.

So yeah. In this example I pointed to, the EPYC is doing multiple projects at once, while also being faster than the 3900X which is doing only one project. LOL

Also having more PCs makes it easier to attach lots of GPUs.

More of a hassle to deal with setting each one up. My hosts have 5+ GPUs each. One configuration configures 5 of them at once. No need to configure 20 hosts, each with 1 GPU when I can just configure 4 hosts with 5 GPUs each.

Not interested in efficiency. I have a 21 year old car. I could buy a newer one and use less fuel, but I can't be bothered, and it would cost more to buy.

You just like to argue to argue? This whole argument started on the fact that YOU claimed buying more than one Ryzen 9 setup is more cost efficient than running a single EPYC setup. So what argument are you even making now?




EPYC is more efficient than Ryzen 9 3000 Series CPUs by a good bit.

Actually it just omits the name. Anonymous refers to a host without listing the owner. Not sure why it hides the name since I've said to not hide anything. It would be easier to say "the computer called Jenny is crashing tasks frequently" and everyone would know which one it was. Now I have to say which one by other means, like "the Xeon, no not that one, the other one, er.... the one with slightly less credits."

Not what I am talking about AND I literally said it tells you the host is anonymous. Keep up with me, please.



It hides your LIST when you hide your computers. As shown in the screenshot above. The host, as I already said, can still be found on most projects in the hosts list. It just shows up as anonymous.

check mine.

Bloody showoff. First in total credit, but someone is running slightly faster than you! Can't see what computers he has though.

---

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

check mine.

Bloody showoff. First in total credit, but someone is running slightly faster than you! Can't see what computers he has though.

That user recently enabled his account to export stats.

So all the points he earned since he started his account shows up on one day.

You can see it better on free-dc.

That user recently enabled his account to export stats.

So all the points he earned since he started his account shows up on one day.

You can see it better on free-dc.

Ah, that stupid EU bullshit. They're ruining the world one byte at a time. What's more annoying than cookies? Asking you if you want a bloody cookie! Reminds me of Mrs Doyle.

---

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

yeaah, lets go E.T. :-D

it's two 64-core CPUs. HT is on, so that's 128 threads each, and 256 threads total for the system.
the only project the system runs is Universe, and it crunches at 90% CPU utilization (roughly 230 threads) 24/7
~3.0M ppd total, ~1.5M ppd each CPU.

comparing cherry picked tasks from hosts is pointless since universe has varying task lengths. they are not all the same so it's easy to end up comparing two tasks that aren't comparable.

go nuts and cope.

Perhaps.... your CPU is very fast at Universe since it's a simple program, even my 8 core Android runs it. Maybe your CPU has a tiny cache or something and can't cope with bigger programs like Cinebench, so falls flat on the floor for benchmarks. I'll stick to the tried and tested benchmark scores, not what you happened to have found it's good at. Maybe someone can recite the alphabet really fast, doesn't mean they're exceedingly clever.

---

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

it's two 64-core CPUs. HT is on, so that's 128 threads each, and 256 threads total for the system.
the only project the system runs is Universe, and it crunches at 90% CPU utilization (roughly 230 threads) 24/7
~3.0M ppd total, ~1.5M ppd each CPU.

comparing cherry picked tasks from hosts is pointless since universe has varying task lengths. they are not all the same so it's easy to end up comparing two tasks that aren't comparable.

go nuts and cope.

Perhaps.... your CPU is very fast at Universe since it's a simple program, even my 8 core Android runs it. Maybe your CPU has a tiny cache or something and can't cope with bigger programs like Cinebench, so falls flat on the floor for benchmarks. I'll stick to the tried and tested benchmark scores, not what you happened to have found it's good at. Maybe someone can recite the alphabet really fast, doesn't mean they're exceedingly clever.

cope harder.

it takes the tiniest bit of effort to google the specs. you don't have to theorize if an EPYC has "tiny" cache when you can go to the product page and see that it has 256MB L3 cache. not that Cinebench is very sensitive to cache size anyway.

your confusion is because you did exactly what I told you not to do. you compared two cherry picked tasks that are not comparable. if you look at systems that have been running continuously for a long time and have stabilized RAC, you'll see the 64-core EPYCs are about 4x, maybe 5x, faster than your 3900X/T, not 10x. Skillz was right, you need to be told everything twice.

---

it takes the tiniest bit of effort to google the specs. you don't have to theorize if an EPYC has "tiny" cache when you can go to the product page and see that it has 256MB L3 cache. not that Cinebench is very sensitive to cache size anyway.

Well something is making it rubbish at benchmarks. Benchmarks are just a program, which means you have a CPU good at certain things only.

your confusion is because you did exactly what I told you not to do. you compared two cherry picked tasks that are not comparable. if you look at systems that have been running continuously for a long time and have stabilized RAC, you'll see the 64-core EPYCs are about 4x, maybe 5x, faster than your 3900X/T, not 10x. Skillz was right, you need to be told everything twice.

I did not cherry pick anything, I looked at a few pages of tasks and picked the longest and shortest ones.

---

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

No point being pedantic about verbiage. You did the wrong thing. That’s what matters.

Universe tasks are not standardized. You cannot compare single tasks to each other without knowing they are the same size computationally. I told you this from the beginning and you went and immediately did exactly what I told you not to do.

Your other mistake is relying on a single benchmark. There is no “one-size-fits-all” benchmark. You need to either pick a benchmark that’s more similar to the intended workload (or just use the intended workload as a bench) or take a collection of relevant benchmarks and compare those.

---

Need to run same task standalone multiple times on different hardware.

No point being pedantic about verbiage. You did the wrong thing. That’s what matters.

Universe tasks are not standardized. You cannot compare single tasks to each other without knowing they are the same size computationally. I told you this from the beginning and you went and immediately did exactly what I told you not to do.

As I just told you, I looked at a range of times.

Your other mistake is relying on a single benchmark. There is no “one-size-fits-all” benchmark. You need to either pick a benchmark that’s more similar to the intended workload (or just use the intended workload as a bench) or take a collection of relevant benchmarks and compare those.

A good CPU should be fast at everything. I'm not going to buy something which is good at Universe only to find it's rubbish at other programs. Universe as I said already is a very small simple program.

A benchmark is designed to test all aspects of a CPU.

---

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

Well something is making it rubbish at benchmarks. Benchmarks are just a program, which means you have a CPU good at certain things only.

Perhaps.... your CPU is very fast at Universe since it's a simple program, even my 8 core Android runs it. Maybe your CPU has a tiny cache or something and can't cope with bigger programs like Cinebench, so falls flat on the floor for benchmarks. I'll stick to the tried and tested benchmark scores, not what you happened to have found it's good at. Maybe someone can recite the alphabet really fast, doesn't mean they're exceedingly clever.

Small cache? LMAO

Dude, Google is like... right here. www.google.com

3900X = 64MB L3 cache
3900XT = 64MB L3 cache
7742 (EPYC, 64-core, ROME) = 256MB L3 cache

As I have told you before and I'll say it again. Benchmarks are useless metrics. If you want to know how a CPU performs, then get real world data from what you actually intend to use the CPU for.

The EPYC is good on every BOINC project. Not just Universe@home. I only picked Universe@home because I knew GPUUG users have EPYCs that run it nearly 24/7. So I knew I'd be getting good results from their EPYC hosts.

Additionally, those benchmarks only work certain aspects of the CPUs. MOST of those benchmarks test things such as gaming, video/image production and compression which has a MUCH higher market of what people use their CPUs for vs the niche group of people who use it for BOINC where the programmers/admins of those projects build the projects with the language they know and are familiar with.

You are right. EPYCs are good at one thing. Number crunching which is what 99% of these BOINC projects need. What they are NOT good at is gaming which is what 99% of all benchmarks revolve around.