Core Duo vs Athlon 64 vs Celeron D Benchmarks

[21stHermit]

I have three available PC's for my map making projects which will require 100's - 1000's of hours of CPU time. So I've been running specific benchmark tests to determine which CPU is best and I've gotten surprising results.

The PC's:
1. Dell E1505 w/ Core Duo T2300 1.66GHz, 1.5GB RAM
2. Celeron D 360 Cedar Mill, 3.46GHz, 2GB DDR2
3. Athlon 64 3000+ Winchester, 1.8GHz, 2GB DDR

Two Benchmark Test:
1. Compile text files to binary machine code, Text files 5-20MB, Binary files 100K-2MB
2. Convert elevation data, elevation points on a regular array, to contours. Conceptually identical to Isobars, converting weather pressure data to lines of constant pressure. Very large data and contours files so memory is likely important.

The Test Results:

Test 1 15 finished files totaling 10.4MB
E1505....: 87min
Athlon64: 47min
Cele360.: 13min

Test 2 8 finished files totaling 68.5MB
Cele 360: 4hr 7min
Athlon64: 2hr 4min

Test 1 is by far the more important inasmuch as their are 10X as many files to produce.

If you use this Pentium 4 641 Cedar Mill to Athlon 64 3000+ Winchester comparison, the Cedar Mill wipes the floor with a Winchester. I could not find a single benchmark on Tom's where the A64 bettered the P4. Not so in my test.

Edit:
It would appear the cacheless wonder, Celeron D 360, is a real horse at code compiling (Test 1). My take is that code compiling is a linear task where GHz rules, unlike Contour Creation (Test 2) which has more "balls in the air" hence the need for more cache.

I'm blown away!!! :Q

Thanks
Hermit

 

[aigomorla]

Originally posted by: 21stHermit
I have three available PC's for my map making projects which will require 100's - 1000's of hours of CPU time. So I've been running specific benchmark tests to determine which CPU is best and I've gotten surprising results.

The PC's:
1. Dell E1505 w/ Core Duo T2300 1.66GHz, 1.5GB RAM
2. Celeron D 360 Cedar Mill, 3.46GHz, 2GB DDR2
3. Athlon 64 3000+ Winchester, 1.8GHz, 2GB DDR

Two Benchmark Test:
1. Compile text files to binary machine code, Text files 5-20MB, Binary files 100K-2MB
2. Convert elevation data, elevation points on a regular array, to contours. Conceptually identical to Isobars, converting weather pressure data to lines of constant pressure. Very large data and contours files so memory is likely important.

The Test Results:
Test 1, E1505 & A64, 15 finished files totaling 10.4MB
E1505: 87min
A64...: 47min

Test 2, Cele 360 & A64, 8 finished files totaling 68.5MB
Cele 360: 4hr 7min
Athlon64: 2hr 4min

Test 1 is by far the more important inasmuch as their are 10X as many files to produce.

If you use this Pentium 4 641 Cedar Mill to Athlon 64 3000+ Winchester comparison, the Cedar Mill wipes the floor with a Winchester. I could not find a single benchmark on Tom's where the A64 bettered the P4. Not so in my test.

Does anyone have an explanation?

Thanks
Hermit

first off i find it kinda funny how your benching 2 desktops against 1 laptop being the C2D.

Also i would check your speed on the C2D. Speedstep might be preventing it from going at full blast. There is absolutely no way in hell a C2D would lose to a A64 @ near same clock.

 

[21stHermit]

Originally posted by: aigomorla
first off i find it kinda funny how your benching 2 desktops against 1 laptop being the C2D.

Also i would check your speed on the C2D. Speedstep might be preventing it from going at full blast. There is absolutely no way in hell a C2D would lose to a A64 @ near same clock.

Those are the PC's I own, I'm not a magazine/website with access to loaners for test. The laptop in NOT a C2D, it is as I stated a Core Duo, their is a real, albeit small difference.

In Task Mgr, the Core Duo showed 50% CPU useage or 100% for one core and 0% for the other. Neither of the test applications is multi-threaded.

The PC's and the benchmarks are my real world, I'm hoping someone here can help me understand the results better.

Thanks for your thoughts.
Hermit

 

[Aluvus]

The Pentium 4 you are using as your basis for comparison has 4 times as much L2 cache and a higher Front Side Bus speed (and a bit lower clock speed) than the Celeron D that you have.

 

[21stHermit]

Originally posted by: Aluvus
The Pentium 4 you are using as your basis for comparison has 4 times as much L2 cache and a higher Front Side Bus speed (and a bit lower clock speed) than the Celeron D that you have.

You find me a Celeron D 360 Cedar Mill vs Winchester comparison and I'll use it. In an imperfect world ones use the closest comparison one can find.

What I REALLY care about are my benchmark test, why do my Core Duo and Celeron D 360 perform so poorly relative to an ancient Winchester? Sorry that wasn't clear.


Thanks for your thoughts,
Hermit

 

[Dman877]

Because the winchester is faster

 

[InflatableBuddha]

From Wikipedia:

Core Duo

Limited Floating Point Unit (multiply/divide) throughput for non-parallel computations or single-threaded processes; this is due to the smaller number of floating-point units in each CPU core compared to some previous designs

Other than that, maybe the smaller amount of RAM is hurting the Core Duo system. Are there differences in hard drive speed or cache too? Maybe the laptop's hard drive is only 5400 RPM or has less cache, or both.

As for the Celeron D, I'm not at all surprised it got smoked by the Winchester. Aluvus explained why.

 

[21stHermit]

Originally posted by: InflatableBuddha
From Wikipedia:

Core Duo

Limited Floating Point Unit (multiply/divide) throughput for non-parallel computations or single-threaded processes; this is due to the smaller number of floating-point units in each CPU core compared to some previous designs

Other than that, maybe the smaller amount of RAM is hurting the Core Duo system. Are there differences in hard drive speed or cache too? Maybe the laptop's hard drive is only 5400 RPM or has less cache, or both.

As for the Celeron D, I'm not at all surprised it got smoked by the Winchester. Aluvus explained why.

The meat and potatoes reply I was seeking!!! :beer: :beer:

I have another 1GB SO-DIMM in a UPS box today, will install and re-run bench. Yes 5400rpm and not looking to change the HDD. However, Test 1 has very small files so should all reside in RAM, me thinks?

You say you're not surprised, 2:1 is pretty smoked. :shocked:

Thanks
Hermit

 

[InflatableBuddha]

Originally posted by: 21stHermit

Originally posted by: InflatableBuddha
From Wikipedia:

Core Duo

Limited Floating Point Unit (multiply/divide) throughput for non-parallel computations or single-threaded processes; this is due to the smaller number of floating-point units in each CPU core compared to some previous designs

Other than that, maybe the smaller amount of RAM is hurting the Core Duo system. Are there differences in hard drive speed or cache too? Maybe the laptop's hard drive is only 5400 RPM or has less cache, or both.

As for the Celeron D, I'm not at all surprised it got smoked by the Winchester. Aluvus explained why.

The meat and potatoes reply I was seeking!!! :beer: :beer:

I have another 1GB SO-DIMM in a UPS box today, will install and re-run bench. Yes 5400rpm and not looking to change the HDD. However, Test 1 has very small files so should all reside in RAM, me thinks?

You say you're not surprised, 2:1 is pretty smoked. :shocked:

Thanks
Hermit

You're welcome. Good luck with the next round of benching.

Oh, another thought...how was the RAM in the laptop configured? 1.5 GB would mean 3x512MB, correct? If so, that would be slower than 2 sticks running in dual channel. Configure the laptop with 2x1GB and hopefully that will speed things up a lot.

 

[formulav8]

That Core based lappy chip IIRC is based on the original Core not Core2 and is much slower. The original core cpu, when Anand benched them along time ago put them slower than the Turion to slightly faster in some areas.

So if that is a Core1 based cpu, then don't expect near Core2 performance

As for the Celeron, they are absolute dogs and are a waste, especially the northwood netburst based junk. Even the newer ones are dogs compared to whats out there and unless that is absolutely all you can afford, and will be going to do a upgrade to a better cpu soon, then consider such a cpu. Otherwise, save your money until you can afford the cpu.

Anyways, this all only my experience and opinion. Take it with a grain of salt


Jason

 

[21stHermit]

Originally posted by: InflatableBuddha
Oh, another thought...how was the RAM in the laptop configured? 1.5 GB would mean 3x512MB, correct? If so, that would be slower than 2 sticks running in dual channel. Configure the laptop with 2x1GB and hopefully that will speed things up a lot.

I was wondering if anyone would notice the laptop RAM.

No, it's only two SODIMMs, so a 1Gb and a 512MB, which I assume is running single channel. Now that DDR2 is below $25/GB, it was a no brainer.

Anyone looking for a 512MB SODIMM? Cheap!!!

Thanks
Hermit

 

[21stHermit]

Originally posted by: Dman877
Because the winchester is faster

That may very well be. Certainly my first conclusion.

 

[Markfw]

First, forget a P4 netburst based Celeron.

Now as to the other two, the first thing (already noted above) is the disk. You HD in the laptop is 4200 or at best I think, 5400, so you are SEVERELY handicapped with the otherone certainly at 7200 rpm. That alone can totally screw up any benchmark, so why even try ?

If you try a superpi 1m, that will probably leave out the HD, try that and report back.

 

[21stHermit]

Originally posted by: Markfw900
Now as to the other two, the first thing (already noted above) is the disk. You HD in the laptop is 4200 or at best I think, 5400, so you are SEVERELY handicapped with the otherone certainly at 7200 rpm. That alone can totally screw up any benchmark, so why even try ?

If you try a superpi 1m, that will probably leave out the HD, try that and report back.

Absolutely no interest in SuperPi, as I explained in my original post, those are exactly the programs I have to run to build my maps. SuperPi is mere entertainment.

First, forget a P4 netburst based Celeron.

Turns out it is as you said, a 65nm shrink of the old P4 architecture. I incorrectly thought it was Core based. However it is a computer I own and can run a bench on.

Thanks for your thoughts
Hermit

 

[21stHermit]

New benchmark in the OP. Amazing result. Celeron D 360 has found its calling!!!