More 1:1 vs 4:5 on DFI 975 X/G

[Yellowbeard]

Recently, N7 posted a thread on 1:1 vs 4:5 results

I finally got some things sorted with this DFI 975 X/G MOBO. If you have this MOBO, the recent 12-01 bios has several improvements.

I got it stable and then ran some of the benchmarks suggested by N7. My results parallelled some of N7s predictions and some by Yoxxy in the first thread. The other thread got diverted at the end so here is a new thread with some results and conclusions.

First, at the same CPU speed, 4:5 will show a very slight increase over 1:1. However, on this board, my max OC at 1:1 is 430mhz fsb and the max at 4:5 is 400mhz fsb. So, while 4:5 is slightly better at the same CPU speed, the increased CPU speed of the 430mhz/1:1 OC is better than 400mhz/5:4 OC. The extra memory performace gained at 4:5 is not enough to warrant giving up 430mhz at 1:1. The board is absolutely unstable at any FSB past 400mhz when using the 4:5 so, 1:1 is easily faster overall. This may not be the case with other MOBOs so, I'll try to get another board and test.

So, on this DFI 975 MOBO, IMO, synch is best. However, if the anticipated chipset and bios revisions allow a better handling of dividers for the 975 chipset, the advantage may fall back to asynch. Now, if we get dividers where we can upclock the CPUs well (this board won't do it well at all) , then asynch may be viable, ala 875 chipsets.

The 2nd 430mhz result in each part is with 4 x 1gb of PC8500 Dominator. Apparently, my MAX FSB is just as with 2 x 1gb as with 4gb. With Vista coming I added a 4gb result just for giggles. I don't have a 64 bit OS to play with yet but, the results still look really good.

I ran Aquamark, SuperPI 1M, 3DMark 01, and Sandra Memory/Buffered. I did not do the full suite N7 did due to time constraints. I will probably add 3DMark 03, 05, and 06 but, most likely they will be GPU limited for the most part.

System Spex:
DFI 975 X/G, Rev 1.0, bios 12-01
E6300
2 or 4 x 1gb Corsair Dominator PC8500 w/Dominator Fan
EVGA 7800GT
Corsair HX520 PSU
Corsair Nautilus 500 on CPU
BIG, LOUD, FAST 120mm Sunon fan on NB (stock sink) and sinks on clock gen.

This board could REALLY use better NB cooling.

Aquamark
400mhz FSB 1:1 112,544 (4,4,4,12)
400mhz FSB 4:5 112,928 (5,5,5,15)
430mhz FSB 1:1 114,557 (5,5,5,15)
430mhz FSB 1:1 114,998 (4 x 1gb Dominator PC8500 5,5,5,15)

3DMark 01
400mhz FSB 1:1 35,498
400mhz FSB 4:5 36,030
430mhz FSB 1:1 36,821
430mhz FSB 1:1 36,484

SuperPI 1M
400mhz FSB 1:1 20.250 sec
400mhz FSB 4:5 20.141 sec
430mhz FSB 1:1 19.031 sec
430mhz FSB 1:1 18.79 sec

Sandra Memory Bandwidth, Buffered

400mhz FSB 1:1 7373/5091
400mhz FSB 4:5 7612/5091
430mhz FSB 1:1 7864/5091
430mhz FSB 1:1 7999/5091

*Apparently there is an issue with this chipset and the latest Sandra reporting the Integer benchmark correctly as it never changed from 5091.

This is obviously by no means complete but, the results are consistant.

 

[n7]

Exactly as i expected; thanx very much for adding your results.

Some game benches & video encoding ones would be nice, but really, we already know what the results will be.

Intel does indeed love bandwidth, so when one has a choice of running their RAM at higher speeds with looser timings (while maintaining the same max CPU clock), do it!

(Excepting on nF 6xxi, where 1T is best.)

 

[Yellowbeard]

Your link for the Quake Demo gave me a "bad" version. When I run it basically tells me to buzz off and go buy the real game. Is there a known working version around? Also, is there an encoding benchmark in the WinRAR program or, do you encode a specific size/type of file? I'll run a few more as I also want to see what this setup does with 4gb (even with 32 bit OS).

 

[n7]

That Q4 bench needs the full game AFAIK.

WinRAR bench utilization pic:
http://img82.imageshack.us/img82/7038/winrawrqi8.jpg

I ran the multithreaded one.

I would say that bench has got to be one of the most memory sensitive benches out there.

 

[Duvie]

Yellowbeard I have this board coming. I just picked one up for 105 shipped. I will be running (2) X1900's on it with my E6600 which cannot do more then 3.4ghz anyways....I downloaded latest bios and will flash it when I get it.


I would like to run it at 8x425 1:1.......

What is the vdroop like?

Do you actively cool your NB?

 

[Yellowbeard]

Originally posted by: Duvie
Yellowbeard I have this board coming. I just picked one up for 105 shipped. I will be running (2) X1900's on it with my E6600 which cannot do more then 3.4ghz anyways....I downloaded latest bios and will flash it when I get it.


I would like to run it at 8x425 1:1.......

What is the vdroop like?

Do you actively cool your NB?

With the latest bios mine is doing 430x7 1:1. I don't think you will have a problem as mine is doing 430mhz with 4 x 1gb DIMMs just as easily as with 2 x 1gb so, FSB seems to be artificially limited in the bios.

TBH, I have not even paid any attention to Vdroop. However, I have not had any issues at all that seemed voltage related, which is what I would expect of DFI.

Currently, I am using the stock sink with a LOUD FAST Sunon 120mm fan over it. But, I would strongly suggest an aftermarket cooler such as the Thermalright HR-05 ("plain, not SLI) or the Swiftech or TT coolers. The NB goes well over 40c quickly with this stock cooler and no fan. Right now, under 2+ hours of Orthos with the 4gb of memory, stock sink with AS Ceramique, it's at 29c but, I have an ambient temp of 16c in my lab. The stock sink sux, say that 3 times fast.

If you order a cooler before you get the MOBO, beware that it has the 2 clip retention system, not holes and not 4 clips. So, your choices are limited somewhat. Personally, I am wary of a HEAVY NB cooler hanging over my video card suspended by only 2 clips. So, I'll prolly order a Thermalright as it is lighter than the TT or Swiftech models.

 

[Duvie]

I will remove it and reapply AS5 on it, and then mount one of my 50x50x15mm fans on it... my S3, and DS3's dont even use fans unless i go over 450 fsb.....

 

[Yellowbeard]

You should actually be fine. But, I always tend to build for the long haul so I tend to overcool if that is possible.

 

[Yellowbeard]

Duvie, just as a test I removed the 120mm fan. In under 1 minute the NB temp went from 29c to over 50c. I put an 80mm fan back on it after it passed 50c so I have no idea how far it would have gone. An 80mm fan keeps it at about 49-51c in 16c ambient. The 120mm hovercraft engine keeps it at 29c.

One tuning note, I had better luck with stability with the Vmch at 1.7v as opposed to the max, 1.75v recommended at DFI Street. Apparently with the stock cooler, that last .05v was not worth the extra heat.

 

[Duvie]

I can balance a bit since I will be using my E6600 I can keep it below 400fsb if needed....9x378 will still give me same results of 3.4ghz but I could use 4:5 to get 945ddr2 which my memory does just fine....

This board will have enough case heat with 2 x1900's....I have excellent case cooling now....I have a 120mm fan blowing on ram and northbridge...the vid cards will exhaust heat out back and an 80mm fan on the side will blow on top of the vid cards.

i will put the 50x50x15 on the heatsink and keep voltages lower.....


I may try a quick spin with my QX6700 to see how the boards superior power can handle it...

 

[Termie]

Hey guys - good to see people using this board over here. I don't push my board, but the new bios (12/01) did let me load up a much higher FSB than I could do before. 366 easy, no voltage mods. I will say that during the summer at an ambient temp around 23 C, my NB went to about 55 degrees at idle without cooling, 68 at load. I propped a 120mm fan in my case, and I now hold at 38-39 C idle, 41 C load, and my ambient is about 20 C (it's winter). You want some active cooling for sure.

Also, I have big problems loading anything other than 1:1 - and because I have to go in and clear the CMOS, I don't bother. Like Yellowbeard said, the higher FSB pays off more than tweaking the memory divider.

 

[The-Noid]

Try 4-3-2-1 @ 1:1 or 3-3-3-8 or 3-4-3-9 that is where this board and all 975x fly. They love Cas 3.

My theory on 975x was always that CAS 3 would beat CAS 4 ddr-1000.

CAS 4 800 and CAS 5 1000 are completely different ball games.

 

[n7]

Originally posted by: Yoxxy
Try 4-3-2-1 @ 1:1 or 3-3-3-8 or 3-4-3-9 that is where this board and all 975x fly. They love Cas 3.

My theory on 975x was always that CAS 3 would beat CAS 4 ddr-1000.

CAS 4 800 and CAS 5 1000 are completely different ball games.

I fail to see how that makes sense.

Based on simple fractions, the difference between CL 3 800 & CL 4 1000 vs. CL 4 800 & CL 5 1000 should indeed be less, but 4:5 1000 should still win by a slight margin, unless there are some major changes in the subtimings.

 

[The-Noid]

Theory and Practice are not always the same unfortunately.