New PC build PSU

mikey
mikey
Joined: 22 Jan 05
Posts: 11888
Credit: 1828024173
RAC: 207692

RE: RE: ... The most

Quote:
Quote:
... The most annoying event for you and others like you who have made such a hefty investment would occur when the guy down the street running a single Pi finds one first. :>)

I don't care who 'makes the find' - it's sufficient for me simply to be involved. I derive my satisfaction from the contribution itself.

Hey me too!! As I see it there a number of tasks to crunch, as long as they get crunched, by whoever or even whatever, the goal is met. If I do one task and you do 1 million, that's okay with me, if I do a million tasks and you only do one, that's okay with me too! Now that being said I do have put a personal limit of one billion credits per project on myself for right now. That may change as I get more and more of the gpu projects, that I will crunch, up to that number. I have gotten two there so far, and have left both at that point. I don't ever see me getting a cpu only project up to that point, it just takes too long.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109376976218
RAC: 35984897

RE: ... In light of your

Quote:
... In light of your knowlegde ...


What knowledge :-).

I don't have any particular knowledge of what might be coming down the track a year or two from now and how it might perform. I tend to wait until others try out the new stuff and write about it. Once I commit to something, I try to make it last for as long as possible, so quite often I'm well behind the latest and greatest.

Cheers,
Gary.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109376976218
RAC: 35984897

RE: ... Now that being said

Quote:
... Now that being said I do have put a personal limit of one billion credits per project on myself for right now.


Unfortunately, there seem to be one or two projects out there that offer ridiculous rates of credit compared to the majority. Quite off-putting if you want to compare your 'whole-of-BOINC' standing against others. I haven't done that for at least five years or more.

Apart from using RAC as a measure of the benefit ( or otherwise) of certain performance tweaks within the fleet, I largely ignore credit these days. I choose projects based on how interested I am in the science the project is doing. Once I join something, I tend to stay with it, unless something really significant changes. I don't have any 'limits' of the type you mention.

Cheers,
Gary.

mikey
mikey
Joined: 22 Jan 05
Posts: 11888
Credit: 1828024173
RAC: 207692

RE: RE: ... Now that

Quote:
Quote:
... Now that being said I do have put a personal limit of one billion credits per project on myself for right now.

Unfortunately, there seem to be one or two projects out there that offer ridiculous rates of credit compared to the majority. Quite off-putting if you want to compare your 'whole-of-BOINC' standing against others. I haven't done that for at least five years or more.

Apart from using RAC as a measure of the benefit ( or otherwise) of certain performance tweaks within the fleet, I largely ignore credit these days. I choose projects based on how interested I am in the science the project is doing. Once I join something, I tend to stay with it, unless something really significant changes. I don't have any 'limits' of the type you mention.

EXACTLY....I reached one billion on one project in 11 months!!! RIDICULOUS!!! The 2nd one I did because I could. These 'Science' ones are a bit harder though as they are a bit more careful with how many credits they give out. I am NOT complaining, I am just saying that lots of credits does NOT equal lots of science at those other places!!

tbret
tbret
Joined: 12 Mar 05
Posts: 2115
Credit: 4812613666
RAC: 79502

Phil - One thing you asked

Phil -

One thing you asked about here or somewhere that I don't think ever got addressed.

I don't crunch on CPUs and I don't buy the most expensive GPUs. Since that's true, I've never looked past what's available for my AMD GPUs in terms of number and speed of PCIe slots.

The fastest thing I could find (on average) for AMD was a board that took each of two x16 slots and divided them into four x8 slots if they are all populated.

The question is, at what point in the line-up do cards process data quickly enough to become data-starved at x8?

I hope it is obvious to you that the more work units you crunch simultaneously is going to change the answer, as well as the raw processing power of the card.

Let's say a GTX 690 is "starved" on a x16 slot. A GTX 690, elsewhere, "seems" to have about the same ability to crunch numbers as two overclocked GTX 670s. That's not a scientific raw score of chip-power. That's the practical "truth" if you look at real-world results in some other projects.

Keep in mind the owner of those cards was trying to process multiple work units on each of the two GPU-chips on a single GTX 690. If he had cut that back to one work unit per card, would the result have been the same?

I don't know and it doesn't much matter, because you don't crunch on powerful cards like that and deal with the heat and power consumption to process one work unit at a time.

Based on people I know's results, I would say that the GTX 690 is capable of over-taxing the bandwidth available on a single x16 slot. Is the 670 also over-taxing the slot? It doesn't appear so. Do they over-tax a x8 slot?

Well, I just don't know the answer to that question either. The reason I can't tell is because I don't know if any change from two cards in two x16 slots to four cards in x8 slots is caused by the bandwidth available, or because the RAM and CPU are so much busier tending to four cards that every card is having to wait just a fraction of a second longer every once-in-a-while.

I'm not the only one that experiences this. It seems to be fairly universally true that running one card is really quick. Two cards are pretty quick. Three cards are "fairly quick" and four, if you've got enough processor, fast enough RAM and you're running at everything at x8,... four are quick but measurably slower than one the same card in the same machine by itself.

I came back to tell you that only so you'll know that ten machines with one card each in them will be *slightly* faster than five machines with two cards (typically).

Some people are devoted in ways I am not. They'll run a $1,000 processor and really expensive RAM, overclock everything including the PCIe bus, cool their machines in exotic ways using esoteric motherboards and RAM coolers and apply a deep understanding of over-clocking everything together - and I've noticed some really exotic expansion cabinets being used to double-up on GPUs.

They might be able to not experience this "consumer-grade" level of performance deterioration.

You'll notice I did not let this keep me from putting-together machines with four GPUs in them. That's still the way to get the most crunch in the least amount of floor-space even if it does mean a lot of fans and warm rooms.

But then, I don't have anything on the level of a GTX 780 to push until it starts screaming in pain.

I do feel ignorant about that, but it is rational ignorance since at some point my devotion to any given machine only goes so-far. I've got too many computers to worry about to put any of them on the jagged edge.

Phil
Phil
Joined: 8 Jun 14
Posts: 579
Credit: 228493502
RAC: 0

Hi Tbret, Your timing is

Hi Tbret,

Your timing is impeccable with your post. I had just decided that I was probably over-reaching considering my inexperience and needed to slow down a bunch.

Thank you for the info on the PCI buss. You are right. Somehow that got lost in the shuffle.

After much thought, I decided to really back off on the really fast computers for now till I get some more experience with Linux and crunching in general. You know the old cliche, gotta walk before you can run. I was in too much of a hurry without enough knowledge.

So I'll start off with some less expensive machines and post the builds, the fine tuning, and the results, for those interested in building their own crunchers.

I thank you for all the time you put into several of your postings, and your willingness to nudge (or drag kicking and screaming) a new guy in the right direction.

Phil

robl
robl
Joined: 2 Jan 13
Posts: 1709
Credit: 1454479721
RAC: 8921

RE: I would strongly

Quote:

I would strongly encourage you to read this thread regarding PCIe slot performance. Read the entire thread or scroll down looking for "message 127974 (written by "ExtraTerrestrial Apes")and read from that point. I learned a lot. It was not how I thought things worked.

No sense is loading a PC with high end GPUs if your not going to get the performance you expect.

Again you might read the "this" link above regarding PCIe slot performance.

Phil
Phil
Joined: 8 Jun 14
Posts: 579
Credit: 228493502
RAC: 0

Interesting read robl. Thanks

Interesting read robl. Thanks for the link. Lots of good info in there.

Phil

Phil
Phil
Joined: 8 Jun 14
Posts: 579
Credit: 228493502
RAC: 0

Well, the decision was kinda

Well, the decision was kinda made for me. I recently came into some 19 inch equipment racks from a defunct radio station. Guess what, my rack mount cases are too large (front to rear) to fit. But free is always better than several hundred dollars each for new ones. Thankfully I had only purchased one rack mount case for testing so didn't waste a bunch of money.

Soooo, I've gone completely the other direction. My current computers are sitting on shelves in the rack with the sides off and sometime in the future I will strip the mobo trays out and mount everything in the open air.

Tbret wins our friendly debate over cooling issues by default. Not that he cares about winning, he was just trying to keep me from making some costly mistakes. Big smile there tbret, lol. Free Scooby snacks for Christmas dude.

I know a whole bunch more than I did back in June when I started this journey. Thanks to all who have advised, urged, nudged, and downright dragged, me in the right direction.

Phil

tbret
tbret
Joined: 12 Mar 05
Posts: 2115
Credit: 4812613666
RAC: 79502

RE: Tbret wins our

Quote:

Tbret wins our friendly debate over cooling issues by default. Not that he cares about winning, he was just trying to keep me from making some costly mistakes. Big smile there tbret, lol. Free Scooby snacks for Christmas dude.

Wait... what? I don't care about winning? Since when?!? DO YOU THINK I JUST...

Oh, wait, that's right.

I've got a boss, many government and quasi-governmental regulators breathing down my neck, three kids, and I've been married for over thirty years. I've had to give that up.

But yeah, costly mistakes can be errors,...and also, pricey.

But winning by default of the guy for not carrying a tape measure with him? I can't celebrate that. (Mostly because I'm too old to celebrate much, but not because I'm above enjoying it when my kind of misfortune befalls someone else. Since you made that mistake, now I guess it's my turn. Again.)

Don't be surprised if you have to put a box fan on all of that open gear. And don't be surprised if you hear a banshee shriek, a wail, a grinding sound, a pszzzzzzzft, and a really unearthly moan, and a crash if you leave the door open and let the cat in. If you don't already have an air-compressor, put that on your list to ask from Santa.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.