PDA

View Full Version : Core i7


science man
01-17-09, 01:44 PM
I know this question probably came up before, but just my sake, can the smp client fill that cpu to 100% stress?

Schlotkins
01-19-09, 10:39 PM
I have the same question. I loaded the SMP version and I'm showing 51% usage on the 8 cores.

Thanks,
Chris

klear
01-19-09, 11:08 PM
if you guys are talking about winSMP then it sounds like you need to run 2 WinSMP clients to fill the 8 threads, i think.

Adak
01-20-09, 04:05 PM
Exactly right, klear. The smp client is hard coded at present for 4 cores. If you have more than that, you need to double down and run two instances of it, with different machine ID numbers (don't forget that!).

You may find, as I did with my 8 core server, that it works better still with 3-5 instances, depending on the system, it's RAM, and the type of WU that Stanford is sending out. I'm currently running 4 smp instances.

Is_907
01-20-09, 06:42 PM
Wouldn't it be better, though, to just run two or three VMs with the Linux SMP client? ;)

science man
01-20-09, 06:53 PM
Wouldn't it be better, though, to just run two or three VMs with the Linux SMP client? ;)

VM=?

science man
01-20-09, 06:53 PM
Exactly right, klear. The smp client is hard coded at present for 4 cores. If you have more than that, you need to double down and run two instances of it, with different machine ID numbers (don't forget that!).

You may find, as I did with my 8 core server, that it works better still with 3-5 instances, depending on the system, it's RAM, and the type of WU that Stanford is sending out. I'm currently running 4 smp instances.

3-5 smps on an 8 core!? Dude you would be (and are since your running 4) killing that thing. It would kill my machine to run 2 smps. I'll tell you one thing, running 3+ smps on an 8 core, idk how it makes the deadlines.

Adak
01-20-09, 11:40 PM
VM's are Virtual Machine instances which have been set up by some free software, to run autonomous regions of memory, like a separate machine (computer). With a VM instance, you can have a Windows rig, but the VM can be running Linux. Check out the sticky's for all the details on it.

The 8 core server already runs Ubuntu linux, so VM's would only slow it down. Each of the 2 quad Xeon cpu's runs at 2.5 GHz, and has a 12 Meg cache. System memory is 8 Gigs, which Ubuntu can all recognize and use, if needed (but it never uses even half of that).

The current work units are much less generous than the previous month's, that's for sure! They used to always finish in less than 24 hours, but now it's more like twice that long. The only deadlines it misses are WU's that "hang", for some reason or other.

science man
01-21-09, 09:32 AM
You are overloading it. I can tell because my quad only takes a day and a half when I'm running a gpu at the same time, but when only running smp it only takes about a day, maybe a bit more.

Adak
01-21-09, 12:44 PM
You are overloading it. I can tell because my quad only takes a day and a half when I'm running a gpu at the same time, but when only running smp it only takes about a day, maybe a bit more.

Define "overloading", please.

The cores are running at less than 100%, but above 90%. Their temps are normal, their speed is default. This is a supermicro server board and has no overclocking options that will allow folding to be done.

This is not a "gaming" board or mobo, or RAM, or cpu. It's designed for low power needs, and reliability, rather than speed. Comparing this system with a regular desktop system is like comparing a team bus with a Corvette.

FAH gets the WU's returned with 80% - 87% of time remaining, the mobo temps are fine, the OS reports all is well, so I base my belief that it's not overloaded, upon those facts.

science man
01-21-09, 11:00 PM
How is folding a low power need? When I fold 24/7, which is what I do, i can tell folding isn't a low power need because of the big increase I get in the electric bill. And when I said overloading I didn't realize that your system isn't running utilized at 100% which I find very shocking for two reasons. 1. Running 4 smp clients should overload that thing unless it has 16 cores or more. 2. According to what you're saying it's slower than an average desktop. Btw what does mobo mean and what's VMware and Ubuntu?

Adak
01-22-09, 02:45 AM
Mobo = motherboard, VM refers to VMWare, which creates on the system, a virtual machine of it's own. This allows you to use another operating system OS, "on top" of your normal operating system, inside the virtual machine. Ubuntu is a favorite (free), Linux version, and means "All of us together" in an African dialect. The founder of Ubuntu is from The Union of South Africa, but Ubuntu is headquartered in Great Britain.

My server is slower than a gaming rig, certainly. It was purchased for server work, not folding or gaming. It also is a server with low power requirements - lower than a gaming rig would have. It folds OK however, and will continue folding until I finish the programming for it's real project to begin.

I don't believe folding is a low power need. It's just a need. You decide what to give it. It's like giving to any other charity - you give something you are comfortable with, you don't give two pennies, that would be useless; you don't give your whole paycheck either, that would be very unwise.

CPU's are a bit like a cup - you can't "overload" them. You can fill the cup to the very brim, but you can't somehow cram more water into the cup. When it's full, 100% full, it will simply hold no more.

Same with your cpu. You can't somehow force the OS to give the cpu an "overload" of work. It can't be done.

science man
01-22-09, 05:55 PM
Here's an e.g. of an overload.

Running 2 or more smp clients on one quad. When I say "overload" I mean when you give the cpu so much work, it slows down at 100%. Like running a new app on an old computer.

Adak
01-22-09, 08:38 PM
There's a big difference between running a gaming quad on Windows XP, and running a server rig on Linux.

The former will only recognize 3.2Gigs of RAM, Linux recognizes all 8 Gigs of my server's RAM, and Linux is more stable, and more efficient at handling multiple tasks at the same time, than Windows (any version).

When a system is returning SMP WU's with 75% to 87% percent of their time still remaining before their deadline expires, that's not an overloaded system. When that number decreases to 20% or less, then you have a real concern.

Your opinion, and my opinion, are mostly hot air and hand waving, if we have incontestable facts available to us. That's the essence of the scientific method isn't it? We form a hypothesis, and then we make an experiment to test it - to get the *facts*. Without those tests, and those facts - we could believe anything that came along and had a pretty song, couldn't we?

Bouncing a basketball a lot will make you grow way taller - sure!
Global warming - you bet!
Evolution - oh yes!

On the radio last night some author was talking about the new perpetual motion machines we'll see soon! :screwy::screwy::screwy:

Truth is, the earth has been cooling for the last 10 years, and we've regained ice sheets the size of Texas near the North Pole. Evolution is a theory, not a fact. Science has never seen evolution, only trivial adaptation, by any species. That doesn't mean it's not true - that doesn't mean it's not an important theory in science - it just means that it's not a proven *fact*, yet.

As an objective and rational human, you have to cling to the facts, and see past the "spins", the pseudo facts, that everyone wants to "sell" you, as the truth.

Fortunately for us, FAH gives us the facts of our folding power, with every WU we return, right in our logs. (You may need to use the -verbosity 9 start up flag to see all of them).

So what we believe is good loading or overloading doesn't matter - we have facts - blessed facts, to guide us. :soda:

cw823
01-22-09, 11:03 PM
Here's an e.g. of an overload.

Running 2 or more smp clients on one quad. When I say "overload" I mean when you give the cpu so much work, it slows down at 100%. Like running a new app on an old computer.

It's not an overload, I think you must mean something different. If it were in fact overloading, it'd shut off.

It's like moore's law, "if some's good, more's better". If he gets higher PPD with 3 clients than 2, it's a no-brainer

science man
01-22-09, 11:11 PM
There's a big difference between running a gaming quad on Windows XP, and running a server rig on Linux.

The former will only recognize 3.2Gigs of RAM, Linux recognizes all 8 Gigs of my server's RAM, and Linux is more stable, and more efficient at handling multiple tasks at the same time, than Windows (any version).

When a system is returning SMP WU's with 75% to 87% percent of their time still remaining before their deadline expires, that's not an overloaded system. When that number decreases to 20% or less, then you have a real concern.

Your opinion, and my opinion, are mostly hot air and hand waving, if we have incontestable facts available to us. That's the essence of the scientific method isn't it? We form a hypothesis, and then we make an experiment to test it - to get the *facts*. Without those tests, and those facts - we could believe anything that came along and had a pretty song, couldn't we?

Bouncing a basketball a lot will make you grow way taller - sure!
Global warming - you bet!
Evolution - oh yes!

On the radio last night some author was talking about the new perpetual motion machines we'll see soon! :screwy::screwy::screwy:

Truth is, the earth has been cooling for the last 10 years, and we've regained ice sheets the size of Texas near the North Pole. Evolution is a theory, not a fact. Science has never seen evolution, only trivial adaptation, by any species. That doesn't mean it's not true - that doesn't mean it's not an important theory in science - it just means that it's not a proven *fact*, yet.

As an objective and rational human, you have to cling to the facts, and see past the "spins", the pseudo facts, that everyone wants to "sell" you, as the truth.

Fortunately for us, FAH gives us the facts of our folding power, with every WU we return, right in our logs. (You may need to use the -verbosity 9 start up flag to see all of them).

So what we believe is good loading or overloading doesn't matter - we have facts - blessed facts, to guide us. :soda:

Hey, how do you tell how much time you had left before the deadline?
Btw I agree with you except for one thing but I'll get to that later. I agree with what you say about facts and the fact I'm using is that when a computer is slowing down because its cpu has to much to process aka it is overloaded. If you know of a fact that contradicts that, then link me to the source of that fact please.

Ok, the thing I don't agree with you on is, why do you think that global warmming isn't happening and that the oppisite is? Hasen't it been proven that the ozone layer is thining out and that, that is a sign of global warmming?

science man
01-22-09, 11:11 PM
It's not an overload, I think you must mean something different. If it were in fact overloading, it'd shut off.

It's like moore's law, "if some's good, more's better". If he gets higher PPD with 3 clients than 2, it's a no-brainer

Oh well its quad that's being given the work of an oct.

Adak
01-23-09, 12:37 AM
Hey, how do you tell how much time you had left before the deadline?
Btw I agree with you except for one thing but I'll get to that later. I agree with what you say about facts and the fact I'm using is that when a computer is slowing down because its cpu has to much to process aka it is overloaded. If you know of a fact that contradicts that, then link me to the source of that fact please.

Ok, the thing I don't agree with you on is, why do you think that global warmming isn't happening and that the oppisite is? Hasen't it been proven that the ozone layer is thining out and that, that is a sign of global warmming?

I use FahMon to help monitor the progress of most WU's. It's a free utility, available from www.fahmon.net/

I don't need FahMon for the server, because I know how it works - and how it should fold. I just look at the log file if there is any problem.

The cpu doesn't slow down unless it's overheating, and has built in protection from same (modern cpu's do have this). Otherwise, the cpu will slowdown for things like "inactivity" if it is set up that way, by the BIOS and or operating system. Both Linux and Windows have this feature.

Other than the above, which are controlled by the BIOS and OS, the cpu itself will never slow down except for a thermal overload. It runs on an oscillating circuit that sweeps through the system many times each second.

It can't "slow down". It's hardwired to run at the speed set by the oscillating spikes of electricity. It's like a surfer riding the waves - it doesn't make the waves, or tell them how fast or how big to be - it just rides them.

If the OS is backed up trying to organize a queue of things to be handled by the cpu, then it appears the cpu is running "slow", but it's not the cpu - that's your OS. The cpu is still riding those waves - same speed as before.

Global warming has happened - many times. So have Ice Ages. Man had nothing to do with those, and may not be influencing the current trend, either.

Data from satellites and weather researchers has shown that we've been in a cooling trend for the last 10 years or so. Before that, we were in a warming trend. The basis for man made global warming rests on whether our release of more carbon dioxide, etc., particularly from fossil fuels, will cause a "greenhouse" effect.

The factual answer is: we don't know. There has been no credible experiment made, to test that hypothesis.

We do know that the increase in temperature in the Arctic Ocean, has been partially caused by increased underwater volcanic activity, off the coast of Greenland. I have seen the video of this but don't have the url. In any case, the video can't be verified by me.

Currently, Europe is experiencing the coldest winter in decades, as is most of the U.S. Will it continue to cool down? Again, we don't know.

What I do know, is that Al Gore (who's already a millionaire), will be a billionaire if we accept man made global warming as a fact, and take steps to try and stop it. He's heavily invested in companies to make that happen.

I also know that no scientist will debate global warming with facts, because they have none. They speak of "trends", of "associations", and show dramatic pictures of the Mendenhall Glacier retreating from the global warming, the snows retreating up Mt. Kilimanjaro in Africa, etc.

And bouncing a basketball will make you taller, too! :D :p :p :p

Here's a fact for you: We know that Alaska was once a heavily vegetated region - quite lush, in fact.

Do you think all that vegetation grew in the frigid arctic weather Alaska has now? Not a chance. The oil in Alaska was made from the compressed vegetation under great pressure and heat. Have you seen pics of the oil fields on the North slope? Hell could freeze over with room to spare, trust me. :p

We also know that huge ice sheets once stretched across North America and most of Europe and Asia. We're talking ice two to three miles thick. Man didn't make that ice form, and man didn't make it disappear, either.

An objective person has to discern fact from fiction, and science from spin silliness. Now it's all PC to be a big believer in *global warming*. It sells papers, T shirts, TV time, cars, air conditioners, heaters, you name it.

Truth remains - we don't know what causes global warming, or what causes global ice ages, either. One guess - it's caused by cycles in our Sun.

Google for "John Coleman Weather" and you'll have links to more facts on global warming. John's a very old time weatherman who was one of the first to say "Hey, wait a minute - this isn't science", when the man made global warming boat started picking up steam.

eddyg519
01-23-09, 10:36 AM
hes using server grade equipment (supermicro) on a rock solid OS and doesn't have cores @ 100%

his systems fine and its more then likely laughing at this thread now lol

Adak how much PPD does that machine produce from CPU folding alone? I always wondered what sort of scores a server rig like that would produce compared to a budget GPU2 folding rig. Of course I understand comparing the two is silly because they are totally different but what PPD does it get?

Adak
01-23-09, 12:53 PM
Right now, it's not able to handle more than 4 cpu SMP WU's at a time, and they're buggers because they take nearly 30 minutes each to upload the results! PPD averages out to about 4,500. With more favorable WU's it can handle 6 SMP instances at a time, and make about 9k per day.

The budget GPU (any GPU) rig is vastly better. Two good 260's or so, and you're beating the server, easily. Even a good gaming type quad will easily beat one quad's worth (1/2) of my server, until you factor in the power it used to do that. :)

In my case, power isn't the problem, but heat is. I'm near the desert outside San Diego, and anything that makes more heat than absolutely necessary, is very unwelcome, since that triggers the extra expense of Air Conditioning to keep the home livable. So you pay twice - once to power the folding, and once more to cool down the air you just heated up, while folding. :argue:

I wouldn't recommend servers for folding, but this one has been a very nice fill in folder, until and unless it gets started in a new DC project that I'm currently investigating. In fact, it's the only rig I have folding, right now. The other 3 have all developed problems of some kind or other, that I have to sort out.

ChasR
01-24-09, 12:17 PM
THe Core i7 has 8 logical cores, but it only has 4 physical cores. As windows always has, it lies about cpu utilization on Hyperthreaded cpus. A 4 threaded app run across 8 logical cores is reported to use 50% of the cpu even though it may be using 100% of the execution circuitry of all four physical cores.

Running A2 core WUs in native Linux the Core i7 makes about 2689 ppd/GHz with the -smp 8 flag set and about 2715 ppd/GHz with 2 clients running the -smp 4 flag (THis comparisson is between two different machines that likely have differing ram configs). So it's not really worth it to run two clients on A2 WUs on the i7 (a single A2 core WU nearly fills the execution circuitry), same as on the rest of the quads.