3.4 <> 3.4
 

I just purchased a PIV 3.4 from work ... the exact same model and specs I have on my desk at work but it is running almost 20% faster than the one at work.

Both doing 18M DC...one at 0.032ms; one at 0.027ms. Same version of Prime95.

One difference is XP Home / fresh rebuild vs XP Professional / 3 year old build.

[quote=petrw1;164830]I just purchased a PIV 3.4 from work ... the exact same model and specs I have on my desk at work but it is running almost 20% faster than the one at work.

Both doing 18M DC...one at 0.032ms; one at 0.027ms. Same version of Prime95.

One difference is XP Home / fresh rebuild vs XP Professional / 3 year old build.[/quote]
Dust buildup on the slower one's CPU cooler, possibly?

While 20% is quite a lot, if the PC you have at work has been running for a while without a wipe it could have collected loads of crap on it, both physically on the CPU cooler, and softwarically (that's a word honest ... maybe).

Additionally, are you sure it's the exact same model? There were so very many models released, on different processes and architecture revisions. 3.4 GHz P4s were available in Northwood, Gallatin, Prescott, Prescott 2M and Ceder Mill families.

Since both computers come from the source and have probably been in use for the same time in the same conditions, dust build up would be more or less the same. The slow down is probably due to some other factors :

- An old installation, Windows gets slower and slower as time goes by. Registry values accumulate, files and directories get fragmented and the build in defragmenter is not really efficient.

- At work there are more maintenance tasks running : SMS for keeping the park up to date for instance.

- The antivirus software can have a huge impact as well. For instance where I work we use the antivirus program from a big OS vendor and it uses more than one CPU hour a day for "quick scans", updates... The previous antivirus program from a specialised company would typically use less than ten minutes a day for the same tasks.

- Software like Office Communicator are using a lot of resources, I saw a near doubling of Prime95 iteration times when it was active.

- The network is much larger and NetBIOS trying to keep up with all entities uses a lot of resources as well.

- ...

Jacob

I think throttling hits harder than 20% though. Probably a difference in either cpu cache or memory timing/speed, or a background process/service chewing cycles/cache.

Thanks everyone ... I think it must be a little of all the above comments.

Before I was allowed to take it home it was wiped and rebuilt.
It would have been dusted.
It has a fresh clean build.
It has XP-Home instead of Professional.
It has NO "Business Fluff" software.
It has not virus software (yet), no special network software, etc. etc.

I don't think it is overheating at work because the air coming out of the vents is cool.

I doubt the memory, cache, motherboard, etc are different since my company would order about 1,000 of a specific model/build from the supplier and configure them all the same to make support easy....unless the supplier itself does not use identical components. Note: we only buy: IBM, HP or Compaq (now HP) ... his particular one is HP Compaq.

As somebody posted above, PIV 3.4 came with different cores. Make sure one is not a Northwood and the other one a Prescott, for example. Prescotts were less efficient than Northwoods, and the difference in timings you posted is, AFAIR, compatible with one being a NWood and the other a Prescott. As the Prescott core was the evolution of the Northwood, it may have been the case that in a 1000 batch of computers some came with the Northwood flavour and others with Prescott, due to shortage of Northwoods in the market. But I think that the explanation may as well be on the software side: you may have processes running in one of the computers that are stealing cycles from P95. The fact that the one with the fresh install is the fastest is in line with that hypothesis.

try a cd linux and see which is faster then

[QUOTE=lycorn;164996] Make sure one is not a Northwood and the other one a Prescott, for example. [/QUOTE]

Is there an easy way to do this check? Sorry, I'm not a hardware guy.

Since you're running Windows, CPUZ will do nicely. It's a small, free, download. It doesn't need installing, just run the exe and it'll tell you all kinds of info on your CPU.

It can be downloaded from here:
[url]http://www.cpuid.com/cpuz.php[/url]

I maybe wrong, but I thought the Northwoods maxed out at 3.2GHz. The 3.4s are probably both Prescotts.

[QUOTE=lavalamp;164863]3.4 GHz P4s were available in Northwood, Gallatin, Prescott, Prescott 2M and Ceder Mill families.[/QUOTE][url]http://en.wikipedia.org/wiki/List_of_Intel_Pentium_4_microprocessors[/url]

[QUOTE=lycorn;164996]As somebody posted above, PIV 3.4 came with different cores. Make sure one is not a Northwood and the other one a Prescott, for example. Prescotts were less efficient than Northwoods, and the difference in timings you posted is, AFAIR, compatible with one being a NWood and the other a Prescott. [/QUOTE]

Here are the CPUID runs main screens from both ... both Prescott. But one has a 17.0 multiplier entry. Can someone explain this in Layman's terms...is this actually an indication of overclocking or just a faster bus? And would this alone explain the 20% better thruput or is it more likely still clutter on the work machine? Even in short bursts on the work machine (i.e I set the iterations low) the iteration time still does not do better than 0.032 compared to a consistent 0.027 for the Home version.

The fact that CPUZ can't get some information is interesting, that could indicate a driver issue on the work PC. Also interesting is the report of just one CPU thread on the work PC, so perhaps hyperthreading is disabled in the BIOS

If hyperthreading were disabled on the work PC, then other programs would have to steal cycles from Prime95, but on your PC, so long as an instruction doesn't use the same part of the core that P95 needs, it can share the cycle.

yeah something is not right with that second one. Which screen is from which computer? It seems HyperThread is enabled on the first but not the second, but it could be a side effect of whatever is causing cpu-z not to be able to see all the additional info.

edit: oh i see the pics are labeled. there is a probly a mainboard driver or cpu driver missing or messed up in the work machine, can you show screens of the mainboard tabs?

edit: also, cool air coming out of the vents doesn't mean it isn't overheating. actually that seems to me like the heat is actually getting [i]trapped inside[/i], same reason that a CPU heatsink [i]should[/i] feel hot, not cool, because that means it is taking the heat away from the CPU.

try running cpu-z while prime95 has been running for a while, and look at the cpu speed to see if it may be throttling at all.

[QUOTE=lavalamp;165226]The fact that CPUZ can't get some information is interesting, that could indicate a driver issue on the work PC. Also interesting is the report of just one CPU thread on the work PC, so perhaps hyperthreading is disabled in the BIOS

If hyperthreading were disabled on the work PC, then other programs would have to steal cycles from Prime95, but on your PC, so long as an instruction doesn't use the same part of the core that P95 needs, it can share the cycle.[/QUOTE]

Good point....when I ran this at work I did get an error about a missing driver but I ignored it and let it continue. I have limited access to my work PC; some software I cannot even install or run so I may not get past the driver error.

It definitely has multithreading though...prime95 tells me so.

Maybe I will open it up and check for excessive dust.

I'll see about printing other tabs.

Thanks

All screen shots for WORK PC....
 

Note a lot of blanks....still getting the driver error. Probably not enough Admin rights.

Home PC screen shots coming soon

HOME PC --- all screens
 

No driver issue....much more data available:

need all screens for WORK pc too...

[QUOTE=xorbe;165679]need all screens for WORK pc too...[/QUOTE]

Right above the Home PC post....but due to driver issues some values are blank.

On my PC CPUZ also blanks out values for some fields, I've found that it only does it if my core voltage is over 1.55 V (Intels reported max safe core voltage).

Though I can't see it being the same reason why your work PC has missing info in CPUZ, unless someone who didn't have a clue what they were doing got into the BIOS.

Since they do appear to be identical CPUs, and you can see that hyperthreading is enabled on both, and the same RAM is installed in both machines, it simply has to be software based, which probably means there's very little you can do, short of talking to a sympathetic admin.

Another Curious Oddity....
 

REMINDER: They are both Hyperthreading PIV 3.40 ...

The Work machine has slightly better iteration times with:
Test
Worker Windows...
Multithreading (CPUs to use): 1

The Home machine has slightly better iteration times with:
Multithreading (CPUs to use): 2