Abit IC7 oc'd with my old 2.4b and new pc3500

[Duvie]

Ok...here is the issue...I can run about 2fsb better then I could with the albatron at same actual vcore...so hooray for that!!!!

However I cannot get this Geil ultra pc3500 stuff rated at cas 2,3,3,7 @433mhz to do anything much better then 436mhz ddr...It will do that at cas2 but no matter what I give it in cas settings all the away to cas 3,4,4,8 it will not keep from freezing during bootup, after windows splash screen or just into windows....My guess is the 2.8v undervolting to 2.78 according to winbond may be the culprit.

I have tried cpu strap 533 and 666 with spd to 4:4 with all settings, voltage maxed to 2.8v and vagp set to 1.65v...nothing...I disabled a few settings that I could under game accelerator but the default for most of that stuff is auto.....

I am locked agp/pci and even tried disabling onboard sound though I don't think it has to do with onboard devices enabled. I can if I stay with 1:1 run 185fsb instead of 173fsb with onboard devices enabled and passes prime95 and tmpgenc test....


Is there a volt mod for the IC7 mobo to give me up to 3.0v like the max board???


Otherwise any other settings you would like to try??? All my voltage rails are tight and much tighter then my albatron where this sticks ran at 450mhz cas 2.5,3,3,7 with 2.83v actual...can't touch that now and I expected less out of the i845pe mobo since most boards other then dcddr mobos like multiple sticks and ocing....

The guy I bought this from said latest bios but I think it says like 1.4 or 14...isn't it up to 17 or 1.7 and if so were tere some fixes that maybe helped timing issues with the NB strap 533 and 666??


Otherwsie I run at 1:1 and te memory scores are 1400 better then before on sandra 2004 and my bench testshows improvements of generally 5% or less....

My 2.6c at best likely wont push to this problem i nfsb speed so Iam doing this just to learn and see what the ram is capable of prioir to the p4c coming....

 

[Thor86]

If you are running asynch with your cpu/ram, remember to disable CPC under your GAT settings, and also the setting right above it, can't remember now. Have fun, and good luck!

 

[Duvie]

Is it custom for this to happen???

1) 2.4b@3.24ghz w/ 180fsb 1:1 for 360mhz ddr cas 2,5,3,3 = 4514/4516 mem scores

2) 2.4b@3.1ghz w/ 172fsb 4:5 for 430mhz ddr cas 2,7,3,3 = 4325/4320

I would think the mem scores would be higher refgardless if the andwidth is more then the ocu can take.....Is what is in fact holding the number down the fact with a 4:5 ratio I am creating more bandwidth then the cpu can handle and therefore sandra only test what the cpu can deliver???


I have the CPC disabled but can't figure out how to disable the other ones...one of them has auto or the different accelerators....

However if above is correct then anything other then 1:1 doesn't seem like it actually results in any benefit.....

 

[jiffylube1024]

Originally posted by: Duvie
Is it custom for this to happen???

1) 2.4b@3.24ghz w/ 180fsb 1:1 for 360mhz ddr cas 2,5,3,3 = 4514/4516 mem scores

2) 2.4b@3.1ghz w/ 172fsb 4:5 for 430mhz ddr cas 2,7,3,3 = 4325/4320

I would think the mem scores would be higher refgardless if the andwidth is more then the ocu can take.....Is what is in fact holding the number down the fact with a 4:5 ratio I am creating more bandwidth then the cpu can handle and therefore sandra only test what the cpu can deliver???


I have the CPC disabled but can't figure out how to disable the other ones...one of them has auto or the different accelerators....

However if above is correct then anything other then 1:1 doesn't seem like it actually results in any benefit.....

Doesn't the IC7's 5:4 reduce the memory speed only? Ie at 172 FSB you would be running at 138 MHz memory? If not then you are just getting crappier bandwidth because 1:1 generally yields better memory benchmarks (except on 800FSB P4 C's, which like all the bandwidth they can get their hands on).

 

[Duvie]

That would be the case if I was using 5:4.....I am running the NB strap 533fsb and it gives you a spd, 1:1 and 4:5 ratio....The 4:5 ratio increases the memory speed....


I did some more research with the game accelerators....I can run turbo w/ emhanced at spd settings of 2,5,3,3....I cannot run F1 or street racer at these speeds...Haven't tried running manual and setting it to 2,6,3,3....cas 2.5 setting doesn't seem to change sandra much but I see it in tmpgenc and superpi.....


Some benches:

Tmpgenc file is a 20min clip from a divx file converted to mpeg2 DVD with constant quality of 85 and estimated motion search.
Winzip file is a 565mb project folder full of autocadd dwgs and project folders containing surface tins and other drawing setups.
Autocadd is rendering a file with shadows on and photo raytrace with highest AA.
Besweet is a simple wav to mp3 encoding with 192kbits on a 46mb wav file.

2.4@3.24ghz w/ 450mhz SCDDR Corsair 512mb cas 2.5,3,3,7

Sandra Mem = 3419/3410
Tmpgenc = 23min 59sec
Prime95 = 2048K = 65.817
SuperPi = 1M = 00min48sec
2M = 01min58sec
Winzip = 1min 03sec
AutoCad 2002 = 0min 43sec

2.4@3.24 w/ 360mhz DCDDR Geil 1gb (2x512) cas 2,5,3,3 spd, GAT auto

Sandra Mem = 4425/4390
Tmpgenc = 22min 45sec
Prime95 = 2048K = 62.478
SuperPi = 1M = 00min 47sec
2M = 01min 51sec
Winzip = 0min 58sec
AutoCad 2002 = NA

2.4@3.24 w/ 360mhz DCDDR Geil 1gb (2x512) cas 2,5,3,3 spd GAT Turbo w/ enhanced

Sandra Mem = 4487/4460
Tmpgenc = 22min 25sec
Prime95 = 2048K = 62.247
SuperPi = 1M = 00min 46sec
2M = 01min 51sec
Winzip = 0min 57sec
AutoCad 2002 = NA


.....................SCDDR vs DCDDR (Auto GAT) vs DCDDR (Turbi GAT)

Sandra MEM 0.0----------------29.4%-------------------31.2%/1.4%
Tmpgenc 0.0-----------------5.4%---------------------7.0%/1.5%
Prime95 0.0-----------------4.8%---------------------5.7%/0.4%
SuperPi (1M) 0.0-----------------2.1%---------------------4.3%/2.2%
SuperPi (2M) 0.0-----------------6.3%---------------------8.3%/1.8%
Winzip 0.0-----------------8.6%--------------------10.5%/1.8%

So DCDDR basically got me the 5-10% most talk about....


The question will be when I set my 2.6c at 3.24-.25ghz how will improve??? DCDDR was always and upgrade with p4c in mind.

 

[Duvie]

Oh and by the way here are my PCMark2002 scores

CPU = 8040
MEM = 9436
HDD = 992 (Ata100 and NTFS)

If that mem score is right ppl must be way over 10000 with DCDDR 1:1 ratios of 500mhz+ddr.

 

[orion7144]

Have you changed to the 17 bios? If you are running the 14 that is way old. The Abit flash utility that runs under Windows will go and check if you are up to date. To get the best out of your memory you only need to disable the last two settings under GAT and I leave all the others to Auto since I did not notice much of of difference.

Some people including myself (with the 16 and below BIOS) believe that the CAS ratings are backwards or atleast screwed up somehow.

 

[Technonut]

If that mem score is right ppl must be way over 10000 with DCDDR 1:1 ratios of 500mhz+ddr.

My PCMark2002 memory score comes in @ 12415 with my Kingston ValueRAM @ 464MHz / Street Racer.

It would not hurt to update your BIOS to V17... I think there is a Beta V18 floating around also.

I have not tried it, but
Here is the VDIMM mod for the IC7...

 

[Duvie]

Originally posted by: Technonut

If that mem score is right ppl must be way over 10000 with DCDDR 1:1 ratios of 500mhz+ddr.

My PCMark2002 memory score comes in @ 12415 with my Kingston ValueRAM @ 464MHz / Street Racer.

It would not hurt to update your BIOS to V17... I think there is a Beta V18 floating around also.

I have not tried it, but
Here is the VDIMM mod for the IC7...

That seems about right for score considering you are 100mhz faster in ram....What I am wondering now is since I would be going for a 4:5 asynch where the dual channel bandwidth is greater then my cpus theoretical bandwidth then it is all just about a number and not relavent to real world performance....

I will still flash the mobo to the latest cause I may gain stability at higher speeds I haven't thoroughly tested......