• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

best out out of Asus A8V Deluxe......(939)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Overlag

Registered
Joined
Jul 1, 2002
Location
Essex, UK
after a slightly hostile/heated "discussion" in someone elses thread i decided to make my own so not to spoil the other thread anymore........

anyway after "upgrading" from an A64 3200/2000/1mb skt 754 to a A64 3200/2000/512kb skt 939 i found quiet a large performance drop, which at first (and i still do slightly put down to loosing 512kb cache). looking around reviews people posted or others ive seen i noticed that 1mb 754 vs 512kb 939 is around 50/50 wins for both IE 3500 vs 3400 yet right now most the things ive tested have clearly show a 90-100% win streak for my old 1mb system. Something clearly must be wrong......

quote from the "other post"
BeerHunter said:
What help you want? :)

I could suggest things like are mem timings the same? same board? Same OS and same drivers? But I'm sure you thought of these things.



i cant run it in the same board can i, different pins ;)

look back anyway, i edited after you quoted.

in truth i should be expecting close to the exact same performance between the two right? so theres got to be something i missed.

Same parts: Ati 9700pro, Dlink 538xt, SB audigy, maxtor 160gig sata hdd, 431watt psu.

2 systems:

Skt 754:
A64 3200/2000/1mb
Asus K8V deluxe
512mb PC3200 2-2-2-6
Overclocked to 2100mhz, ram 210mhz, HTT bus@ 840, PCI/AGP 35/70 (AH thats one thing thats changed, AGP overclocking!? much difference?).

Skt 939
A64 3200/2000/512kb
Asus A8V rev2 deluxe
1gig PC3200 2-2-2-6 (err now back to 512 since the crucial died! :()
Right now its only stable at 2200ish, Ram 220, 2-3-3-6 (another difference), HTT is 1100mhz or 880, PCI/AGP locked at 33/66 (i think)

sargelarge said:
@Overlag

Ok certainly i did'nt meant to be hostile to you. And look, i can't run those tests WITH MY OWN EYES (just using your emphasis on this) because i lack the money in buying parts just for testing ;)

Also, your very private Testbed and your comparsion of the two sockets is not what i call clean. I could even post another six benchmarks of independent web sites prooving my point, even others do post other results, yet you only trust your eyes without doubting that you are doing something wrong.

Anyway, i see your results and indeed, your s939 rig is performing weak compared to the 754, but did you notice that the 939 is actually faster than your 754 by a small margin? Second, did you notice that you are largely peaking out into the same range because you are not hold back by cpu performance but by your 9700 gpu?
You can even comapare to an s754 2800+ at stock with identical configuration and i bet it won't be behind by a large score.

What you can do, if really something is wrong: Check htt multiplier on your s939, do you run 1t rate? Do you ûse correct ram timings or some horrible settings? Well i doubt it, i bet everything is correct with your 939 rig.

What you can do to test further:
Use SuperPi or switch to 640x480res in 3dmark01 so that you are not gpu bound, or even use 3dmark05 as it has a dedicated cpu performance bench and report back with your results (if you want to).

started well, then got hostile near the end ;)

i will look into it, just a bit pissed this ram died :bang head

so from that post i see the main differences are, the timings

210 2-2-2-6 SC vs 220 2-3-3-6 DC (although 210 2-2-2-6 does no favors)
and AGP/PCI lock?

btw i can also run at 8.5x260 2.5-4-4-8 with my single 512mb crucial, but thats single channel and will hurt (as i said the other crucial chip i had died today) so instead i got 2x256 corsair, which can run slightly tighter timings at under 230, but stops working after about 247

2200mhz seems the top for 1.47volts, and 2400 is "benchable" @ 1.6volts but thats hardcore for 90nm so i wont be doing too much of that.....

/end
 
Last edited:
It's the tighter timmings and overclcoked video card in the 754 setup. I'm pretty sure.


Alright, to test, the first thing you need to do is run both setups at thier default speeds and bench.

Meaning 200Mhz FSB, 10X multi, 2-2-2-9 1T mem same mem in both boards is preferable, same video card. (you are overclocking the video in one), and same/similar bios settings..like disable system bios cachable..enable fast writes... disable video bios cacheable etc etc.

Next you may want to run same OS, latest mobo drivers, latest bios on both boards and same video drives.

Then the benches should be even.
 
BeerHunter said:
It's the tighter timmings and overclcoked video card in the 754 setup. I'm pretty sure.


Alright, to test, the first thing you need to do is run both setups at thier default speeds and bench.

Meaning 200Mhz FSB, 10X multi, 2-2-2-9 1T mem same mem in both boards is preferable, same video card. (you are overclocking the video in one), and same/similar bios settings..like disable system bios cachable..enable fast writes... disable video bios cacheable etc etc.

Next you may want to run same OS, latest mobo drivers, latest bios on both boards and same video drives.

Then the benches should be even.
the video card is overclocked in all the top tests, since i got an artic cooler now and thats what its fixed at (372,351)

thats another thing i gota check tho, i think fastwrites are disabled on this board atm. i cant really do any "new" tests on the old system because it would mean rebuilding it.

enabling 1t was another thing someone said, not sure how to do that on this board, im guessing it would be disable 2t? right now im on enable 2t

i think 2-2-2-9/2-2-2-11 are only good for nvidia chipsets too? i though 2-2-2-5 was better for VIA/A64?
 
ok i disabled T2 in the bios and ive jumpped from 5293/5296 to 6142/6079 in bandwidth terms! nice improvement from one little tweak

fastwrites are now ON in the bios, but still off in windows, just going to run ATIgart thingy.

HTT is now at 220 x5 1100 and seems stable.
 
Overlag said:
ok i disabled T2 in the bios and ive jumpped from 5293/5296 to 6142/6079 in bandwidth terms! nice improvement from one little tweak

fastwrites are now ON in the bios, but still off in windows, just going to run ATIgart thingy.

HTT is now at 220 x5 1100 and seems stable.

The difference in raw memory bandwidth between 1T and 2T is around 15% at 200 MHz level (rule of thumb). Your number showed that (15-16%). The increase in bandwidth would translate into better overall performance, though not as much as 15% and is application dependent.

Just to make the terminology consistent, please note that in your case
the 220 MHz is called HTT,
the 220 x 5 = 1100 MHz is the frequency of the HT (HyperTransport) bus, an external bus between the CPU and the chipset.

HTT is an internal frequency (aka FSB in some A64 bios) setting used to set the CPU clock frequency, the HyperTransport bus frequency and indirectly the memory bus frequency via the memory to FSB ratio in bios.

Overclocking setting for various bus frequencies (post 8)
 
Last edited:
testing 3dmark 2001 with those settings, and fastwrites now enabled got me within 100 points of the 2100mhz 1mb system....so im much happyer now. memory bandwidth loving tests are now miles ahead of my socket 754 system.....thanks ;)

anymore tweaks i should know about? :D
 
Back