Overclocking a Athlon 64 X2 4200+

[21stHermit]

I have a Athlon 64 X2 4200+ Manchester and a MSI K8T Neo2-F MB. Currently using the retail HSF from the old single core A64.

I'm using the 2 YO Isolate & Consolidate thread as an OCing guide. Couldn't find it with search to link in, gone?

I have CoreTemp 0.95 running and the idle temps are 39C and 44C for cores #0 & #1. That difference surprises me. In step 2 of the guide, it has me run Prime95. In doing so, within 15sec I get 49C and 55C on the two cores.

My recollection is to not have a A64 above 50C so I terminated Prime95. Not sure what next, the temp jump was so fast that it would likely gone ballistic had I not ended Prime.

Could someone will memory of their A64 OCing days please give me some hints.

Thanks

 

[John]

IMO you're fine if the full Prime/Orthos load temps are under 70C when monitoring w/ CoreTemp.

 

[Rhoxed]

max temps in the A64 x2's are from 60C to 70C - my 3800x2 powers off at 61C running it at 2.7 full load 24/7 with F@H SMP getting 54C and that is just fine.

 

[21stHermit]

Originally posted by: John
IMO you're fine if the full Prime/Orthos load temps are under 70C when monitoring w/ CoreTemp.

Originally posted by: Rhoxed
max temps in the A64 x2's are from 60C to 70C - my 3800x2 powers off at 61C running it at 2.7 full load 24/7 with F@H SMP getting 54C and that is just fine.

Cool!!! I'll continue on with the OC.

Thanks

 

[21stHermit]

Turns out this whole project is a bust. At stock speed the X2 4200+ is only 5% faster than my single core 3000+. This is based on my text to binary complier task.

Because I have a VIA chipset MB, the PCI bus will not lock, hence PCI devices like hard drives become unstable at 225MHz FSB, stock is 200 MHz. I was only able to get semi-stable boots to 210MHz. Predictably, I got a 5% gain at 210 . . . BFD!!!!

So this X2 is pointless for my task. Turns out Newegg's 7-day RMA includes shipping time so I can't return it. Ebay, perhaps.

Hermit

 

[mruffin75]

Originally posted by: 21stHermit
Turns out this whole project is a bust. At stock speed the X2 4200+ is only 5% faster than my single core 3000+. This is based on my text to binary complier task.

Because I have a VIA chipset MB, the PCI bus will not lock, hence PCI devices like hard drives become unstable at 225MHz FSB, stock is 200 MHz. I was only able to get semi-stable boots to 210MHz. Predictably, I got a 5% gain at 210 . . . BFD!!!!

So this X2 is pointless for my task. Turns out Newegg's 7-day RMA includes shipping time so I can't return it. Ebay, perhaps.

Hermit

Would running 2 instances of your compiler work? Afterall you have two cores now? Wouldn't that make it faster?

 

[21stHermit]

Originally posted by: mruffin75
Would running 2 instances of your compiler work? Afterall you have two cores now? Wouldn't that make it faster?

Interesting Suggestion.

Turns out my complier is multi-threaded and runs in MS-DOS. Never thought I'd combine those two in the same sentence. I was using it instead of Prime95 for stress testing. So I watched it in Task Mgr. Both cores were being used, but only at ~25-40%. Don't know why. Also, only using ~150MB of my 2000MB of RAM.

Another program I would also use but only 10% of the time, the developer suggested your very solution, run two instances.

Not sure its worth the hassle to remove the X2 and sell it, probably only get $50.

However the 5C delta at idle grew to 8C under load, so some tweaking required.

Thanks
Hermit

 

[bryanW1995]

reseat the cpu, reapply thermal paste. You're right, you may as well get the most you can out of that cpu, it's not worth much on the open market

 

[myocardia]

Hermit, that compiler isn't multi-threaded. With dual-cores, when running a single-threaded app, the Windows Task manager will show ~50% usage on each core, even though one core will be @ 100%, and the other will be idle. So, you could easily run two instances of it, and be able to do twice the work per hour/day/week.

 

[GuitarDaddy]

Originally posted by: myocardia
Hermit, that compiler isn't multi-threaded. With dual-cores, when running a single-threaded app, the Windows Task manager will show ~50% usage on each core, even though one core will be @ 100%, and the other will be idle. So, you could easily run two instances of it, and be able to do twice the work per hour/day/week.

:thumbsup: Windows "simulates" load balancing of single thread apps on dual core cpus, and the tell tale sign of a single threaded app is the 50/50 core usage. If the app is multi-threaded you will get near 100% usage on both cores. Make a copy of the compiler in a different folder and run two instances

 

[mruffin75]

Originally posted by: GuitarDaddy

Originally posted by: myocardia
Hermit, that compiler isn't multi-threaded. With dual-cores, when running a single-threaded app, the Windows Task manager will show ~50% usage on each core, even though one core will be @ 100%, and the other will be idle. So, you could easily run two instances of it, and be able to do twice the work per hour/day/week.

:thumbsup: Windows always load balances single thread apps on dual core cpus, and the tell tale sign of a single threaded app is the 50/50 core usage. If the app is multi-threaded you will get near 100% usage on both cores. Make a copy of the compiler in a different folder and run two instances

Load balances single thread applications???? I highly doubt that..

From what I've seen on my X2, Windows will only report 100% CPU usage if *both* cores are at 100%... if *one* core is at 100% and the other is doing virtually nothing, it will then report the usage at 50%.

I've seen many instances where Firefox will max out (I'm sorry.. I tend to keep too many tabs open! ).. and my CPU usage will be 50%, but if you look at the core usage, one will be maxed out (Firefox) and the other will be doing virtually nothing..

If Windows could actually "Load balance" a single threaded app to run on more than one core, what would be the point of programming for multi-threaded apps?????

 

[GuitarDaddy]

Yeah I didn't state that very well, I will edit my post. I should have said something like "simulated" or "virtual" load balancing. In other words Windows leads you to beleive it is balancing the load across both cores, when in fact one core is doing all the work.

 

[21stHermit]

Good New - Bad News!!!

As has been suggested, the program is not multi-threaded. I was able to run two seperate DOS windows, each running from a seperate folder and indeed the CPU usage went up to 85-100%. However, I kept getting access errors, when that occured, the process would stop and I'd be missing a compiled file and I'd have to click OK to restart. That's not good.

For whatever reason most of the errors occured in the first DOS Window, but also a few in the second.

In the spirit of testing, I attempted the same two Window test in my Core Duo laptop. Within the first minute I had a couple of access errors and I terminated the test.

My conclusion, for this DOS complier, running two windows is NOT practical.

What say you?

Hermit

 

[VirtualLarry]

Originally posted by: GuitarDaddy

Originally posted by: myocardia
Hermit, that compiler isn't multi-threaded. With dual-cores, when running a single-threaded app, the Windows Task manager will show ~50% usage on each core, even though one core will be @ 100%, and the other will be idle. So, you could easily run two instances of it, and be able to do twice the work per hour/day/week.

:thumbsup: Windows "simulates" load balancing of single thread apps on dual core cpus, and the tell tale sign of a single threaded app is the 50/50 core usage. If the app is multi-threaded you will get near 100% usage on both cores. Make a copy of the compiler in a different folder and run two instances

That's screwed up how XP ping-pongs threads across available CPUs. It really shouldn't do that. Try installing the MS multicore patch for XP (version 4 or newer), that hopefully should fix that problem.

On a properly functioning OS, one CPU core should be pegged at 100%, and the other should be idle, when faced with a single-threaded heavy task.

 

[VirtualLarry]

Originally posted by: 21stHermit
Good New - Bad News!!!

As has been suggested, the program is not multi-threaded. I was able to run two seperate DOS windows, each running from a seperate folder and indeed the CPU usage went up to 85-100%. However, I kept getting access errors, when that occured, the process would stop and I'd be missing a compiled file and I'd have to click OK to restart. That's not good.

For whatever reason most of the errors occured in the first DOS Window, but also a few in the second.

In the spirit of testing, I attempted the same two Window test in my Core Duo laptop. Within the first minute I had a couple of access errors and I terminated the test.

My conclusion, for this DOS complier, running two windows is NOT practical.

What say you?

Hermit

It sounds like the two instances are sharing files. Perhaps temporary files. Try setting the TEMP or TMP environment variables to two different directories, one for each DOS window.


I would suggest, if this is a "work" (paying) project, that you bite the bullet and put together a C2D rig, it should significantly speed up the project, even if you are only using a single core of it. Heck, you could buy a Celeron 420 or 440, and overclock it, and you would have a fast, single-core, C2D.

(I built a rig for my mom using a Celeron 440 2.0Ghz, and it's quite fast and fine for desktop apps. It even calculates on SeventeenOrBust nearly as fast as one of my cores from my desktop C2D rig, even though it's at 2.0Ghz and my rig is at 2.8Ghz. C2D Celerons do NOT suck.)

 

[21stHermit]

Originally posted by: VirtualLarry
It sounds like the two instances are sharing files. Perhaps temporary files. Try setting the TEMP or TMP environment variables to two different directories, one for each DOS window.

I have a bat file which launches the exe with arguments. Would I write a line in the bat file?
If so, how might I word it?

Found this:

SET TEMP=C:\WINDOWS\TEMP
This command establishes an environmental variable called TEMP. Environmental variables are stored in an area of RAM that all programs have access to. If a program wants to see just where the user wants to send all "temporary" files, it could check the environmental RAM area. It would then know (in this case) that the WINDOWS\TEMP directory was the place to go. This particular environmental variable helps reduce disk clutter by establishing one directory for all temporary files - as long as a program is willing to check and store them there. FYI - the PATH goes up in the environmental area. As such you may set the PATH with SET PATH= or with PATH=

So it would seem I could set a temp1 and temp2 in the bat file prior to launching the exe ?

Wrote this line into each batch file:

set temp=c:\windows\temp1

Just errored, damm.


Thanks
Hermit

 

[21stHermit]

Originally posted by: VirtualLarry
I would suggest, if this is a "work" (paying) project, that you bite the bullet and put together a C2D rig, it should significantly speed up the project, even if you are only using a single core of it.

You are correct, except that I "work" for myself and to date this project has netted a few $K, hardly enough to live on. I do have my eye on a 45nm E4700 Q1-08, should really zip things along.

Besides, I'm learning, priceless. Oh, BTW, cheap too.

Do appreciate the help. :beer:
Hermit

 

[VirtualLarry]

A 45nm E4700? Is this confirmed? I would love to get an 800Mhz FSB C2D that is 45nm, OCing should be incredible, without worrying about RAM speeds.

Edit: According to this, E4700 will still be 65nm.
E8200 will be a cheap 45nm, but with 1333FSB.

 

[21stHermit]

Originally posted by: VirtualLarry
A 45nm E4700? Is this confirmed? I would love to get an 800Mhz FSB C2D that is 45nm, OCing should be incredible, without worrying about RAM speeds.

Edit: According to this, E4700 will still be 65nm.

I didn't see anything in "this"

The entry-level Wolfdale 45nm part will be the E4700. That processor will be equipped with a 2MB L2 cache (presumably unified), and run at 2.6GHz on an 800MHz FSB with a 1K price of $133. Intel is reportedly releasing a second 45nm part with a 1066MHz FSB and 3MB of L2 cache, but neither clockspeed nor model number has been disclosed.

I'm not certain, I've seen it both ways, we shall see. The follow on E5xxx almost certainly will be 45nm. But its 2Q-08

Any thoughts on Set Temp ?

Hermit

 

[mruffin75]

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
It sounds like the two instances are sharing files. Perhaps temporary files. Try setting the TEMP or TMP environment variables to two different directories, one for each DOS window.

I have a bat file which launches the exe with arguments. Would I write a line in the bat file?
If so, how might I word it?

Found this:

SET TEMP=C:\WINDOWS\TEMP
This command establishes an environmental variable called TEMP. Environmental variables are stored in an area of RAM that all programs have access to. If a program wants to see just where the user wants to send all "temporary" files, it could check the environmental RAM area. It would then know (in this case) that the WINDOWS\TEMP directory was the place to go. This particular environmental variable helps reduce disk clutter by establishing one directory for all temporary files - as long as a program is willing to check and store them there. FYI - the PATH goes up in the environmental area. As such you may set the PATH with SET PATH= or with PATH=

So it would seem I could set a temp1 and temp2 in the bat file prior to launching the exe ?

Wrote this line into each batch file:

set temp=c:\windows\temp1

Just errored, damm.


Thanks
Hermit

Does the temp1 folder exist? If not, then it will error. You didn't say if you made the folder first..

It's a bit hard to troubleshoot an application that we don't know Also check to see if there are any other folders that it uses in the environment variables.

Also I'd hold off on buying any C2D upgrade for this. You'll need a new mainboard/CPU/RAM for starters and the amount of speed increase you get will probably be not worth the money at all...

You definately need to find out how to run two instances of this application to take advantage of the dual-core CPU you have now..upgrading your hardware won't help much..

 

[21stHermit]

That's a world class response.

Originally posted by: mruffin75
Does the temp1 folder exist? If not, then it will error. You didn't say if you made the folder first..

NO. I thought that the command would create the two folders Temp1 and Temp2. In re-reading the explanation of Set Temp, it appears only ONE Set Temp is allowed, so since I started 2 second, Set Temp would be at Temp2. Which may not exist.

I'll run another test after creating a temp folder.

It's a bit hard to troubleshoot an application that we don't know Also check to see if there are any other folders that it uses in the environment variables.

I'm going to email the code author, explaining the issue.

Also I'd hold off on buying any C2D upgrade for this. You'll need a new mainboard/CPU/RAM for starters and the amount of speed increase you get will probably be not worth the money at all...

You definately need to find out how to run two instances of this application to take advantage of the dual-core CPU you have now..upgrading your hardware won't help much..

Precisely my plan. I want to fully understand these software issues first, then hardware.

Many thanks,
Hermit