2.4B vs 2.4C

[ThugsRook]

2.4B vs 2.4C



Test System
P4 2.4B SL6RZ-3252
P4 2.4C SL6WF-L320
Abit IC7 (i875)
Kingston HyperX 3000 2x512mb c2622
BFG GF FX5900 @ 450x950
Swiftech MCX4000 | 92mm M1A | AS3
SBLive5.1 | WD800jb | Enermax 550w



>>>

[ThugsRook]

@ 3000mhz ~ Test Specs
2.4B = 167fsb 334ddr 1:1 1.525v c2622
2.4B = 167fsb 418ddr 4:5 1.525v c2622
2.4C = 250fsb 400ddr 5:4 1.525v c2622



Benchmarks


http://www.bleedinedge.com/images/thugs/ico-q3a.gif Quake III Extreme 1280x1024
2.4B 1:1 = 192.0 fps
2.4B 4:5 = 192.3 fps
2.4C 5:4 = 192.7 fps



http://www.bleedinedge.com/images/thugs/ico-ut2k3.gif UT2003 Demo 1280x960 flyby / botmatch
2.4B 1:1 = 208.0 / 81.5 fps
2.4B 4:5 = 209.6 / 82.6 fps
2.4C 5:4 = 213.2 / 86.0 fps



http://www.bleedinedge.com/images/thugs/ico-x2.gif X2 Threat Demo
2.4B 1:1 = 81.0 fps
2.4B 4:5 = 81.4 fps
2.4C 5:4 = 83.6 fps



http://www.bleedinedge.com/images/thugs/ico-xisle.gif X-Isle Tech Demo 1280x1024
2.4B 1:1 = 125.2 fps
2.4B 4:5 = 126.9 fps
2.4C 5:4 = 128.4 fps



>>>

[ThugsRook]

@ 3200mhz ~ Test Specs
2.4B = 178fsb 356ddr 1:1 1.525v c2622
2.4C = 267fsb 356ddr 3:2 1.525v c2622



Benchmarks


http://www.bleedinedge.com/images/thugs/ico-q3a.gif Quake III Extreme 1280x1024
2.4B = 192.4 fps
2.4C = 192.6 fps



http://www.bleedinedge.com/images/thugs/ico-ut2k3.gif UT2003 Demo 1280x960 flyby / botmatch
2.4B = 214.2 / 86.8 fps
2.4C = 216.7 / 89.7 fps



http://www.bleedinedge.com/images/thugs/ico-x2.gif X2 Threat Demo
2.4B = 86.1 fps
2.4C = 86.8 fps



http://www.bleedinedge.com/images/thugs/ico-xisle.gif X-Isle Tech Demo 1280x1024
2.4B = 127.7 fps
2.4C = 129.3 fps



>>>

[ThugsRook]

@ 3400mhz ~ Test Specs
2.4B = 189fsb 378ddr 1:1 1.550v c2622
2.4C = 283fsb 378ddr 3:2 1.550v c2622



Benchmarks


http://www.bleedinedge.com/images/thugs/ico-q3a.gif Quake III Extreme 1280x1024
2.4B = 192.8 fps
2.4C = 193.0 fps



http://www.bleedinedge.com/images/thugs/ico-ut2k3.gif UT2003 Demo 1280x960 flyby / botmatch
2.4B = 218.8 / 92.3 fps
2.4C = 220.2 / 94.3 fps



http://www.bleedinedge.com/images/thugs/ico-x2.gif X2 Threat Demo
2.4B = 88.0 fps
2.4C = 89.1 fps



http://www.bleedinedge.com/images/thugs/ico-xisle.gif X-Isle Tech Demo 1280x1024
2.4B = 129.8 fps
2.4C = 130.3 fps



:wave: -thugsrook

[dbwillis]

So, very close they are huh?

[ThugsRook]

yes, extremely close concidering the huge fsb differences.

[jcarno]

:scratch:

[PiLsY]

Youre not running high enough memory speeds to see any advantage form the C cpu m8.

Dont forget all youre getting if you keep mem speed the same as your B chip is an increase based on L2 cache speed alone. To get any benefit form the cache speed you need to run higher frequency on your memory.

Get it running @ around DDR500 and itll beat the crap out of a B cpu @ 200mhz 4:5 or 187 3:4. You get increasing gains based on ram throughput to cache speed compared to the B cpus. To get the benefit of 250+ fsb you need to make sure your system is fully keyed to it.

PiLsY.

[ThugsRook]

Pilsy ~ thats an age old debate that im not willing to get into.

and im not spending $450+ to find out for myself ;)

[jcarno]

So does this mean you're staying with the "b" chip???? :D

[oldfart]

Yes, it is and has been proven many times over that the higher DDR speed with the slower cas timings that would be needed to do so will ~ cancel each other out. It is just more $$ for ~ the same performance (often lower performance).

Your results are ~ what I would expect. Not that much difference. People get all worked up over memory speed when in reality, its all about CPU speed. DDR speed and latency are a small part of the equation. Why people go out and spend more $$ on ram that their CPU and mobo combined for such a minor performance difference is beyond me.

How about a TMPGEnc bench? Be sure you enable the Hyperthreading option for the C chip.

[Jim]

Hey, I'd pay money to see Thugsrook do this "B" vs "C" comparison.....and here it is for free. :D ;)

[oldfart]

Thugs, What OS are you running for this test? Obviously XP is needed for any HT benefit.

[ThugsRook]

jcarno ~ yea im staying on the B chip.

oldfart ~ i dont use a HT capable OS (altho ive tried many, many times)

[ThugsRook]

hmmm, i can see the 3ghz results are little misleading due to me using a 4:5 ratio on the B chip.

im gonna add 3g 1:1 334ddr results for the B chip also ~ gimmie a few mins here....

[oldfart]

Originally posted by ThugsRook
jcarno ~ yea im staying on the B chip.

oldfart ~ i dont use a HT capable OS (altho ive tried many, many times)
It would be interesting to see if HT made much of a difference in these benches. It makes a very large diff in TMPGEnc since it is optimized for it.

[ThugsRook]

i think X2 supports HT, not sure tho.

[oldfart]

AQ3 has the option also, but I dont know if it is used in that standard benchmark.

[Bushe]

How to test the difference HT makes? i read a while ago that to use HT you need to be running the multiprocessor aware portions of XP, which are slighter slower and more stable that the single processor versions of the same. I'm not sure of the implications of this. DOes it mean

to test HT you'd need two install of XP, one with HT on and one with it off? Make them identical otherwise.

It makes for interesting permutations, like install with HT on, test with HT off, should be slower than install with HT off, run with HT off.

What happens when you install with HT off? will it recognise if you then switch HT on?

who has time for all this? I was getting ready for an XP install

[Flamer]

When i bought my 3.06 i put it in a system that was using a 2.53b and Windows did see the HT. I then reinstalled windows with HT enabled because that was the recomendation at the time but i saw no diffence.

[Bushe]

what would be the best benchmark to use?

3dmark2001 gives quite a lot of variation from run to run. What is the most consistent benchmark?

I have aquamark and 3dmark2003. and some others I can't remember the name of. haven't got the x2 benchmark, but as I'm interested in that game I think I'll download that. I'll just go look for it now

[Jim]

Originally posted by Bushe
I have aquamark and 3dmark2003. and some others I can't remember the name of. haven't got the x2 benchmark, but as I'm interested in that game I think I'll download that. I'll just go look for it now


X2 link here. (http://www.bleedinedge.com/forum/showthread.php3?s=&threadid=832) :wave:

[pointreyes]

Originally posted by samcolt
When i bought my 3.06 i put it in a system that was using a 2.53b and Windows did see the HT. I then reinstalled windows with HT enabled because that was the recomendation at the time but i saw no diffence.

I could never understand why people recommended that method either. The XP OS simply asks for you to reboot and viola you have a HT enabled system. The only way to not have HT enabled in XP is by disabling it in the bios.

What I wonder is if what impacts HT is multi-proc built software. Even though multi-proc and HT and completely different, there is one common area-more threads can be processed. However, most software does not support multi-threading processing. Does anyone know if this holds any weight?

[Mikki]

That's a good question pointreyes...have you gone thru the white papers yet? I may have a look...:)

[pointreyes]

Originally posted by Mikki
That's a good question pointreyes...have you gone thru the white papers yet? I may have a look...:)

No, I haven't looked but I do have another reason for thinking about this.

My dual Opteron has memory throughput close to that of a Canterwood. It also has the total processing power of 2.8Ghz. However, the 9800np card on this system for benching was slooooow; whereas it was very fast on the Canterwood system I had with a 2.8c. Reason: the benchmark software only used one proc instead of two so instead of all the 2.8Ghz processing power, it was only 1.4Ghz. Whereas the 2.8c is 2.8Ghz all on one proc. However, Oracle is very fast on duallies as well as the HT procs-but Oracle was horribly slow with a plan jane P4. Interestingly an XP1700 was faster than a P4 2.53 with Oracle and yet I thought both of these procs were multi-threading procs. See the confusion on this? :confused:

[Bushe]

thanks professor fate, it's coming down now!

pointreyes: do you mean the XP1700 is multi-threading? I didn't think it was. Also the P4 2.53 was would be a P4B? there weren't hyper threading were they?

[pointreyes]

Originally posted by Bushe
pointreyes: do you mean the XP1700 is multi-threading? I didn't think it was. Also the P4 2.53 was would be a P4B? there weren't hyper threading were they?

I'm well aware of that. :)
The AthlonXP is using QuantiSpeed™ Architecture and the P4 2.53 is using NetBurst® microarchitecture Hyper-pipelined technology without HT. However the QuantiSpeed™ Architecture just seemed to be faster until Intel incorporated the HT on their P4 procs.

[WeldZilla]

The C chip is vastly superior to the B chip!!!!! Especially when you consider HT technology as well as dual channel mem on the springdale boards. Run PCmark04 for compariso and stop underclocking the mem when running the C Chip. I moved from a P-2.66 running at 3350mhz to a 2.4C running at 3200mhz when you compare their CPU performance on 3Dmarko3, aquamark or PCmark04 it is an absolute slaughter!!! the C chip just destroys the B chip. Here is an example. Aquamark3 compare the CPU scores
http://img.photobucket.com/albums/v253/WeldZilla/Aquamark3compare.jpg
I later turned up my fsb on my C chip to a full 3200 mhz and tightened my mem timings to 2-2-2-5 and my cpu score jumped from 10,402 to 10,942 over 500 points just because of mem timings. to say the 2 are equal is just ludicrous. On 3d Mark03 with my B chip clocked to 3260 mhz my cpu score was 622. with my C chip clocked to 3200 my CPU score was 868.

[WeldZilla]

Memory speed and throughput has a huge effect on cpu score the processor can only process as fast as it's demands are met!!! and with HT enabled it can demand much more!!