How much of a disadvantage am I at with a 2500K? (vs 2600K)

Come join the... uh... er... fold.

Moderators: just brew it!, farmpuma

How much of a disadvantage am I at with a 2500K? (vs 2600K)

Postposted on Thu Feb 09, 2012 4:43 am

I'm thinking about joining the folding effort with the new system I built, to test out its stability and stuff.

I read about this flag that is to be enabled on 8- and higher thread machines which gives extra point bonuses for completing work faster... My understanding is that even with an overclocked 2500K I've only got 4 threads so I won't be able to get these bonus points. Would I have half the ppd compared to if I got a 2600K?

Edit: I did some more reading... bigadv is probably out of my reach since I actually want to use the computer. But I'm supposed to be able to pull somewheres in the neighborhood of 20K ppd with the 2500K.

It seems like the default client only runs on one core, I'm getting 25% CPU utilization. I guess I need to set up the SMP client? Should I uninstall the default client? I am supposed to get more points if I run the SMP client, than if I were to try to run 4 instances of the regular client, yes? I did some back-of-envelope calculations and extrapolating the performance of my single threaded computation I get about 8000 ppd if I were to run 4 instances.

My GTX 260 is not a Fermi, does that mean I can't use GPU3? Looks like I can only expect my 260 to pump out 6-7000 ppd. I might not even bother if that's the case: it's not nearly as much as the CPU and consumes a similar amount of power. GPU's are supposed to have more compute power, even this one that's a few gens old. I guess the complex nature of protein folding simulation makes it difficult to harness all of it.

From the looks of it there's a much better chance of better folding results from the next generation of nvidia GPU's, which should be enough to push me to choose one of them over AMD 7900 series. Unless AMD cuts prices. A lot.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 6:20 am

If you want to test your system stability, why do you care about the points?

Why does anybody care about the points in the first place, the motivation to do scientific computing should be "to help science" and not E-Peen. ;)

That being said, I read that HT actually slows you down if both threads need to use the same resources of the CPU, so I don't really know if it'll be any good for scientific computing.
The backbone of modern industrial society is, and for the foreseeable future will be, the use of electrical Power.
Ifalna
Gerbil
 
Posts: 51
Joined: Sat Jan 28, 2012 11:14 am
Location: Celestis

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 6:50 am

Well, I like to think of points as a measure of how much is contributed and even though it makes no difference in terms of stability per se, there is this curious psychological aspect of it that extends a bit beyond even the e-peen factor. Even if we can accept that the points don't matter, they're *points* dammit. I need to tweak it so that I can get the highest rate of production based on the most normalized metric for it. Maybe it is a particularly non-optimal way to "benchmark" a system, but it does also happen to contribute for a very good cause...

Since points are intended to represent units of work, by corollary if I can tweak more points out I am increasing production, or there is something else that is given point value that I am improving.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 6:54 am

Just out of curiosity: Does scientific computing even stress the system enough for stability analysis?

I have Boinc running almost always (except when I'm gaming of course) but the temps of the CPU are nowhere near the temps I would get when testing with prime.
The backbone of modern industrial society is, and for the foreseeable future will be, the use of electrical Power.
Ifalna
Gerbil
 
Posts: 51
Joined: Sat Jan 28, 2012 11:14 am
Location: Celestis

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 7:26 am

Well. Prime testing is scientific computing. Whatever it is that the Lucas Lehmer FFT algorithms do sure puts a lot of stress on the CPU though.

I doubt folding is quite as heavy a load.

Part of why I want to test is I want to see if my motherboard's CPU throttling issue comes up if I try to fold. The motherboard likes to throttle my chip back down to 3.3Ghz, and it continues to hop back and forth, staying at 4.5Ghz about 80% of the time, once it's a few minutes into Prime95. The amount of throttling is correlated with the amount of cooling on a few mosfets near the socket on the board. Somehow it is able to sense the temperatures there, or perhaps it's some other side effect from the temperature. Either way it continues to be stable which is quite good.

Since Prime95 gets it to throttle, and when it runs at 3.3Ghz the voltage is also automatically lowered, I find that Prime95 actually puts *less* of a load on the chip. It's just less consistent than a system which does not throttle. Would the machine eventually reach an unstable state if it could be somehow forced to prevent throttling? Nobody knows.

I'm trying to come up with a proper cooling solution but i'm fairly certain it requires custom heatsink fabrication. A fan and some tiny heatsinks aren't cutting it.
Last edited by APWNH on Thu Feb 09, 2012 7:33 am, edited 1 time in total.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 7:32 am

As far as I understand it, this jumping between frequencies is normal sandy-bridge behavior. (Personally I'm still stuck on Core 2 Duo so I can't serve you with first hand experience. :/ )
Or am I wrong here? You can't OC the base clock, just the turbo, thus your chip will always jump.
The backbone of modern industrial society is, and for the foreseeable future will be, the use of electrical Power.
Ifalna
Gerbil
 
Posts: 51
Joined: Sat Jan 28, 2012 11:14 am
Location: Celestis

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 7:37 am

Yes, I am using turbo with a setting in my BIOS which lets me choose the max multiplier that turbo reaches. So it only runs the processor at 45x when necessary. I don't even think it distinguishes between single threaded loads and multithreaded loads, though I'm not sure Sandy Bridge is meant to set different multis for those conditions, like with Nehalem. But it should never drop down to 33x in the middle of a prime95 run.

1) No cooling on mosfets, 40x overclock: throttling to 33x begins within a minute
2) Heatsinks attached to mosfets and fan mounted near it, 45x overclock: throttling to 33x begins after 5 minutes

This behavior is quite consistent too.

Throttling never occurs under lighter loads such as single threaded applications and gaming.

But let us not stray from the subject. I'll get around to figuring out the SMP client soon enough and post some ppd numbers for my system.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 8:56 am

APWNH wrote:But let us not stray from the subject.

You're going to have to remind me what the subject is at this point. Is it folding, upgrading to a 2600K, upgrading to a Radeon or to a Geforce, or CPU overclocking / throttling?
flip-mode
Gerbil Khan
Silver subscriber
 
 
Posts: 9101
Joined: Thu May 08, 2003 12:42 pm
Location: Cincinnati, OH

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 9:06 am

Just folding.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 9:08 am

Ah, folding. OK. I'll be of no help here then. Good luck to you.
flip-mode
Gerbil Khan
Silver subscriber
 
 
Posts: 9101
Joined: Thu May 08, 2003 12:42 pm
Location: Cincinnati, OH

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 9:53 am

You already have a $200 2500K, it sounds EXTREMELY pointless to pay $330 for a 2600K when it gives you very little increase in performance. I think you're getting a little too hung up in the "ppd" and its clouding your judgement. Remember, like Ifalna said, you're voluntarily contributing to science, that should give you enough warm fuzzy feeling, what does points matter? (there will always be someone with more points than you)

As far as the "throttling" you speak of, that max multiplier is not made to be run for extended periods of time. It's controlled like an internal timer, once the buzzer goes off, the chip returns to the "stock" frequency. Once it's sat at stock frequency for a while and cooled back down, the turbo boost will kick back in. If you disable "C-States" and/or "Speed Step" in your motherboard BIOS, it should allow the processor to constantly run at max turbo boost multiplier for an unlimited period of time. This is what is done in most articles benchmarking the overclocking performance of K-series CPU's so that they can get a long enough sample time. I wouldn't do any of this if you're going to be leaving your processor folding for long periods of time. However, if you do, be aware that you may have to increase the CPU voltage a bit to achieve stability. Also, don't be suprised if your CPU lives a short life.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1714
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 10:16 am

I'm pretty sure a lot of people have tried switching off the C-states and whatnot on this motherboard and it still throttles the same way. The only way to improve it seems to be to draw more heat from the mosfets, and my observations tend to agree with that... when I had it at 4.0 without any heatsinks I measured 94 degrees C with an infrared thermometer, so it was likely even hotter than that (the thermometer measures in a cone, and my case is extremely cramped).

And I was never thinking about upgrading to a 2600K, I made my decision to get a 2500K because I knew the 4 extra HT threads are not worth $100 and they still aren't. I found on a thread on some forum of some guy who had a 2500K at 5.3Ghz @ 1.5V, running Linux dedicated to folding, emulating HyperThreading with software and getting 80,000 ppd. Well, I'm not gonna do that, and I couldn't even get that much ppd with an overclocked 2600K because I'm not going to dedicate the machine to folding!

But it will be interesting to see if this thing throttles with 100% folding load.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 10:18 am

VRM cooling is essential to avoid throttling the CPU under heavy, continuous, loads.
Z68XP-UD4 | 2700K @ 4.7 GHz | 16 GB | GTX 780 | PCP&C Silencer 950 | XSPC RX360 | Heatkiller R3 / X3 | D5 + RP-452X2 | HAF 932 | 480 GB Extreme Pro
Waco
Gerbil Elite
 
Posts: 794
Joined: Tue Jan 20, 2009 4:14 pm

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 10:22 am

TheWacoKid wrote:VRM cooling is essential to avoid throttling the CPU under heavy, continuous, loads.

Yeah, it makes sense. The manufacturer was cutting costs. Figured they'd skip the heatsink for it.

I'm curious though. How does it *know* when to throttle? mosfet chips don't have integrated temp sensors. This was something that nobody gave an answer for on the thread at HardOCP.
Silverstone SG06: 2500K@4.5Ghz on Corsair H80 Push+Pull Scythe Kama Flow 2, Gigabyte Windforce GTX 670, ASRock Z68M-ITX/HT, 16GB Patriot DDR3-1600 @1T, OCZ Vertex 3 120GB
APWNH
Grand Gerbil Poohbah
 
Posts: 3161
Joined: Thu Oct 31, 2002 9:10 pm
Location: Burlington MA

Re: How much of a disadvantage am I at with a 2500K? (vs 260

Postposted on Thu Feb 09, 2012 4:30 pm

How does it *know* when to throttle? mosfet chips don't have integrated temp sensors.


You sure about that?
Besides: A temp sensor somewhere in the cluster of voltage regulators would be more than enough.
Or the board could monitor the output of these regulators, if it spikes or reaches a certain value based on temps the throttling begins.
The backbone of modern industrial society is, and for the foreseeable future will be, the use of electrical Power.
Ifalna
Gerbil
 
Posts: 51
Joined: Sat Jan 28, 2012 11:14 am
Location: Celestis


Return to TR Distributed Computing Effort

Who is online

Users browsing this forum: No registered users and 0 guests