975X compare to X38 - who is better ? Simply test [56K friendly]

[DEVIL K-ce]

Hi

My and my team mate Matiz did some simply test who is better Old and good chipset named 975X or new baby from Intel named X38.


Aplication used :

Everest Ultimate Edition - memory read, memory write, memory latency. ( The same Everest version )

3D Mark 2006 - CPU test

Super PI mod 1.5XS - probe 4M ( without optimalization )

WinRar included test - the same version of progs. Stop after 1 minute of run test.


Test platforms :

The X38 system come with :

Gigabyte X38-DQ6 board F3 bios
Intel Core 2 Quad Q6600 @ two core enabled
2 x 512MB Corsair 5400UL
Galaxy GeForce 8500GT
BQ 500W


The 975X system come with :

Intel D975XBX2KR Bad Axe
Intel Core 2 Duo E6600
2 x 512MB Crucial Ballistix DDR2 800MHz
Asus GeForce 6600GT
Hiper Power 530W


All test done at 2400MHz CPU freq and mem divider 4:5 ( 667MHz 3-3-3-8 )

So let's see who is better

On left 975X chipset system, on right X38 system. - Click for full resolution image.


Everest Ultimate Edition memory read test

975X score : 6725MB/s ------- X38 score : 6263MB/s




Everest Ultimate Edition memory write test

975X score : 4847MB/s ------- X38 score : 4830MB/s




Everest Ultimate Edition memory latency test

975X score : 64.8 ns -------- X38 score : 75.5 ns




3D Mark 2006 CPU test

975X score : 1997 points ------- X38 score : 2116 points




Super PI mod 1.5XS 4M probe

975X score : 1:59.219s ------- X38 score : 1:55.937s




WinRar included performance test

975X score :1182KB/s ------- X38 score : 1052KB/s




The end


Regards

Martin & Mati

[Speederlander]

Outstanding. For all practical purposes, it's a wash between the two then.

[screwtech02]

Dude, thank's for saving me a BUNCH of $$$$$$.......

Looks like i'll just be getting some better memory, and some 2950's when the come out.....

[pumbertot]

well as is well known X38 shines best with DDR3. but I gues its possible the X38 is running at a looser strap or something else and when fine-tuned it would beat the 975X. also arent they supposed to overclock CPUs at lower voltages so again another benefit these tests wouldnt show.

[redphil]

Imho for "normal" users (ok, here we are at xt, so this species is a bit rare... ) I can not see much sense in upgrading from 975/965 to X38/P35, even if performance improvements would be a bit bigger.
So the xbx2 will stay with me for a while longer.

[Zeus]

Nice test, thanks for the info.
SuperPi 32M would be apprieciated, looks like the X38 is taking a leap there which might be good news to us SuperPi lovers.

[hersounds]

thanks 4 the comparation really interesting and 100% cool data.

x38 no much power so?

[Richard Dower]

So basically X38 is crap and just hype and extra money for no performance?

That sucks man!...Intel and the mobo makers should be ashamed!

[cuke2u]

Hi, so what we are comparing is a mature motherboard, with it's latest drivers and bios, with a young motherboard that has yet to be optimised with software that may not fully recognise the chipset. Perhaps we should test again in a few months time....
ChrisC

[toddm27]

So basically X38 is crap and just hype and extra money for no performance?

That sucks man!...Intel and the mobo makers should be ashamed!

I would say not, as the x38 is optimized for 3d and no 3d tests were run, only cpu and memory tests, I believe if 3d scores were compared is where we'd se the x38 shine esp with a crossfire solution

[Blauhung]

So basically X38 is crap and just hype and extra money for no performance?

That sucks man!...Intel and the mobo makers should be ashamed!

as was said before, X38 is also heavily optimized for DDR3. DDR2 is mostly legacy support and not much attention was paid there as far as MCH design.

It is not worth upgrading to the x38 if you have DDR2. The p35 takes it better because much more time was spent on DDR2 support there.

SO right now if you have a 65nm quad/single core and ddr2 your best choice is a p35. If you have DDR3 and want to upgrade to a 45nm chip soon, then you should look at the X38.

[nullface]

Very nice, hope someone does the same with X38 vs. P35/G33

as was said before, X38 is also heavily optimized for DDR3. DDR2 is mostly legacy support and not much attention was paid there as far as MCH design.

It is not worth upgrading to the x38 if you have DDR2. The p35 takes it better because much more time was spent on DDR2 support there.

SO right now if you have a 65nm quad/single core and ddr2 your best choice is a p35. If you have DDR3 and want to upgrade to a 45nm chip soon, then you should look at the X38.

What if you wont to keep DDR2 and go 45nm, still P35?

[Nosfer@tu]

RELAX everybody.

The results are VERY good, but they dont nessesaryly mean that X38 Sux

I think it RoX. The ease you have when overclocking is amazing

[Talonman]

Do we have a review showing how X38 does much better with DDR3 tha DDR2?

I would love to give it a read...

Post up the link!

I can't speak to weather it was worth it to upgrade from your existing 975X board, but if your just jumping in for the first time like me, X38 was a smart choice I think.

I don't believe X38 is crap. let us have a few BIOS updates for pete sake, before you tag the boards with an official lable.

[Stelios]

Problem with 975 is that it cant clock 470+ fsb easily .. and it can't run memory more than 1200 . Also o/c with quad core is limited.

For us benchers .... i believe X38 would be a far better choice.

[DEVIL K-ce]

Thanks for comment

For me 975X is not universal platform, but still very good. X38 hmm good with Quad Core CPU, also Dual Core, clocking mem good.

But upgrade from 975X to X38 today - I say no. Maybe if I complete my system today I chose X38. Nothing special for me with new chipset, hmmm only full 2 x PCI-E slots and low heat compare to 975X, not more good news.


btw : pumbertot say : with tunning X38 will be better, 975X also can be tunned


Martin

[hipro5]

You just can't compare two different systems with different windows instalations, CPU's (even 2 cores enable on a QUAD).....
You HAVE to use the SAME HD with the SAME windows and the SAME CPU/Rams......Then you can compare them by MAXING them out (ram/timmings/etc. - each platform) keeping the CPU's frequency the SAME for both of them.....Then you can copmare and say yes or no.....

For Example: X38 might play your rams at 600MHz......Why play them at only 333MHz?.....975 might play your rams at 400MHz.....Why play them at 333MHz......
You HAVE to get the MAX out of them so to compare them......Playing(for example) the rams at the same frequency for each platform, means nothing coz x38 might have higher internal latency than 975.....
There are SO MANY settings that MUST be done before comparing......
MAX them out......KEEP the SAME CPU frequency (for reference)......Use the SAME rig with the SAME HD......and redo your tests......This is the way that you will be sure.....

[BulldogPO]

Main reason for me to go from Asus P5W DH to Asus Maximus Formula was nuch better overclockability with quads, My old P5W DH did achieve whopping 366MHz FSB with Q6600

[ent]

For Example: X38 might play your rams at 600MHz......Why play them at only 333MHz?.....975 might play your rams at 400MHz.....Why play them at 333MHz......

I think it was just a fast comparision to show clock-for-clock benches ;-)

DEVILEK >> could you show some 32M? Its strange to me that X38 outperforms 975 in 4M, while 975 has better memory bandwith and memory latencies.

[v_parrello]

imho this comparison is not truthful, too much differences between two testbed:
- different windows installation
- very different hardware: CPU, RAM, video card, psu
with these hardware differencies the results can differ until 10%-20%, is too much to have "scientific" comparison.

And then at the end you have to consider other parameter in order to give a complete reviews:
- X38 is able to run RAM at very high frequencies, 975x works well only mukltiplier 1:1 and then RAM management is very limited.
- X38 is able to run with very high FSB, 975x works at max 450-470 FSB.

For the explanations I give before is not possible with your tests to answer your initial question "who is better?".

[DEVIL K-ce]

I think it was just a fast comparision to show clock-for-clock benches ;-)

DEVILEK >> could you show some 32M? Its strange to me that X38 outperforms 975 in 4M, while 975 has better memory bandwith and memory latencies.

Yes Ent it is fast comp to show ~clock for clock

Yes in weekend I try some 32M on only my system, so all hardware will be the same

btw. 4s better for X38 - this is maybe reason for my OS Pretty old gamer Windows.

Hipro - thanks for good comment and sugestion


Martin

[sky]

hm, it'd be interesting to see compares with maybe gfx cards involved. also, what i'd like to know is if the x38 boards would help when before your mainboard (i975x) limited your fsb-aspirations. i know both my p5w dh's crap around 430ish max, i think one only gets to 420ish...
an x38 that would do 500 would be sweet. especially if it's crossfire-able. but prices are what worries me, too.

[dinos22]

i know that the intention was good in this thread however these chips are not really represented well without being pushed to the absolute limit

afterall that is what separated men from boys in hardware world

you push the two to the max and i think you will see one very clear winner

[Jacky]

I don't understand the outcome, maybe it's just because of the difference in setups:
x38 faster in Superpi 4m/3dmark CPU - all CPU intensive tasks?

Does the 4m set fit into the 4mb cache, if not then pi used to be quite memory sensitive but it's faster even though any memory related numbers are worse? How's that possible..

I conclude: x38+DDR3 >> 975X, because DDR3 offers superior BW anyway (and thus makes up for the worse BW numbers with DDR2 on the x38), and the raw CPU performance increased with x38.
On the other hand: If my memory serves me right anandtech showed 1-5% increase when it comes to memory-BW on early x38? the EXACT opposite of your tests.
I should read the anandtech preview once again, it has been a long time..

[Thorburn]

You disabled two cores on a Q6600. Does this turn off one core on each die, or disable the second die? If it's one on each die then you are giving each CPU core 4MiB L2 cache, rather than sharing it, which could explain the difference in results.

You are also varying the memory, even setting to 3-3-3-8 won't produce identical results between two sets of modules, there are other timings which do make a difference.

Finally the Gigabyte board I believe has some BIOS issues which could be holding it back.

Even with that withstanding, until you test with identical components in each system it is wrong to draw conclusions and influence peoples buying decisions.

There are also many other benefits from moving from 975X to X38, for example the 6 SATA ports of ICH9R compared to the 4 of ICH7R.

Throw in the fact that 975X has fairly dire FSB headroom and even if stock performance isn't going to set the world alight it's still a far superior chipset.