*****Official***** DFI NF4 ultra-D first look

[Tony]

To help you guys DFI again have shipped me two absolute final revision NF4 boards to test.

I have been qualifying OCZ memory on the Ultra D now for 14 hrs and can pretty much now comment on how the boards overclock ram.I have tested with the following modules listed below and will post results here in this thread. I also have some huge news that is under NDA and I will post this also as DFI release me to do so...please don't ask...i will post as soon as they say go

Tested modules.

3200Plat rev2 =TCCD
3500Plat = BH5
3500EB = Micron EOL
3200VX
4200EL = Hynix BTD 43
4400gold = Hynix CTD5
4000gold rev2 = Hynix DTD5

As you can see I have been pretty busy, none of the above are cherry modules, infact some are actually fails which needed a little extra juice.

Now im using an IDE HDD and have sound, USB, ports etc all disabled. The PCIe bus is locked and can be independantly manipulated.
You have already seen the board has the ability to run Vdimm off the 5vline. What you don't know is that i found heat became an issue with the dimms on load at high voltages.The mosfets have heatsinks and you are going to cool them for stability. I have chosen to tweak the 3.3V line up to 3.6V on my powerstream and this allows 3.5Vdimm or so and does seem more stable for me.

Bios is pretty much the same as the 250UT with the addition of 8 Dram drive strength values and increased vdimm and vcore.So if you are used to tweaking the NF3 your going to slot straight into tweaking the NF4 . Chipset voltage does help when overclocking and async ram clocking does work quite well.

Both the ultra D and the SLI board share the exact same PCB. All that is missing on the ultra D is a SI raid controller Chip and 4 sata port and obviously they have a different chipset. regarding SLI etc i will add comment later in the thread. For now im just using an ATI X700 pro and a 3800+ cpu air cooled. No fancy phase change so results are pretty much what you can all expect as a base line.

[Tony]

2x256BH5 3500Platinum.

Now these modules did 240fsb 3.3v on the NF2 UltraB revB which i didn't think to bad....they do a LOT better on the NF4

Everest is a pretty slick bench as it also reports memory latency. Im running cas1.5 here at 264fsb 1:1 for the lowest latency bench i have seen with my dimms. Also the jump from 240 on the NF2 to 264fsb is pretty astounding on this NF4....seems the board loves BH5 in a BIG way.



Moving onto sandra i did buffered and unbuffered. remember the CPU speed is around 2.6+GHZ, a higher cpu speed would create much bigger benches. The unbuffered bench was not as good as i had hoped as i have seem 4k with a modded Epox, still it gives you a rough idea of what you will see.




I added the cas1.5 into the CPU-Z shot to clarify what timings i was running and the cpu used. Pi was 32Secs which is a little higher than i have seen with a Dothan but still well on the money.



Overall with BH5 dimms you will see some nice clocks and i would say you should see a little higher than you have seen before...this is due to board design and the advances in the memory controller.

[Tony]

VX

Well my VX do 260 on nearly everyboard i have and they do 260 on this DFI and no more. Where i saw some nice gains with the BH5 i just matched previous results with the VX.Still we all know they do 2-2-2 at some xtreme speeds and others do have better modules than i do.

The benches.



Timings were tweaked out and voltage was 3.4vdimm.

Sandra.


Numbers real similar to the BH5 although we are at 260fsb here and not 264.

Everest again shows some sick memory latency which is nice.


And a CPU-Z shot to finish.


EB next.

[Tony]

EB

Well 282fsb or so was the max i could do here. I have hit higher on a modded epox so i think i will be asking for a few bios tweaks to try and get performance at least to match it.I have a feeling to many TCCD tweaks are in bios which may harm performance with other dimms, Hynix for sure is not as good on this board.

Anyway some benches, 9X multi for these.

sandra


better than the BH5 probably due to the higher fsb

Everest.

again this is better but latency is down...BH5 is thre king on this one.

Settings used.





I will add my TCCD results a little later.

[cantankerous]

Excellent job matey! Keep up the great work! I can't wait to see the results for the VX. That is pretty impressive for the BH-5. Can you see the same being for BH-6?

I can't wait to hear what the super secret thing is you have to tell us. I sure hope it is runing ATI cards in SLI!

[Tony]

Excellent job matey! Keep up the great work! I can't wait to see the results for the VX. That is pretty impressive for the BH-5. Can you see the same being for BH-6?

Good BH6 is just as good as BH5 as it is actually the same IC. I would say this board clocks my BH5 10 to 15fsb higher than i have seen before on ANY board.

[cantankerous]

Awesome. Well I have OCZ Dual Channel Platinum 2 x 512 BH-6. They do pooh on my NF2 so am hoping for similar. Reading back now I noticed you used 2 x 256mb chips. I guess the VX will be better as those are 2 X 512 sticks. Anyhow we all look forward to your results. Thanks.

BTW when is the NDA up so you can spill?

[Tony]

Probably next week now....all depends on DFI

[GinTonic]

As it was said before, I'm also anxious to see the results of OCZ VX, but those BH5 can be XTREME!!

Good work Bigtoe!

[spinboy]

Wow, seriously that was some mad specs there bigtoe!

[eva2000]

yay you put my mind at ease to know the board likes BH-5/6

Gonna be able to test OCZ PC3700 Platinum and OCZ PC4200 Platinum on this board too ?

[Cranox]

Nice to know the board loves bh-5 , i got 2 x 512mb here

[Tony]

yay you put my mind at ease to know the board likes BH-5/6

Gonna be able to test OCZ PC3700 Platinum and OCZ PC4200 Platinum on this board too ?

I have an unmarked pair of TCCD that meet 3200plat rev2 specs, thats why i quote them as 3200plat rev2.

The board actually seems tuned to BH5 and TCCD.

[Bennah]

Nice work on the ram testing.

Now we got to wait for the NDA to end to see the full extent of what the DFI NF4 boards can do

[choseone214]

for the VX if oyu haven't try

TRFC OF 15
TWR 3
TRTW 3
READ PREAMBLE 5.5

SHOULDN'T CASUE A HIT OF PERF MUCH AT ALL AND SHOULD LET YOU GO HIGHER ON THE FSB

[mrlobber]

You have already seen the board has the ability to run Vdimm off the 5vline. What you don't know is that i found heat became an issue with the dimms on load at high voltages.The mosfets have heatsinks and you are going to cool them for stability. I have chosen to tweak the 3.3V line up to 3.6V on my powerstream and this allows 3.5Vdimm or so and does seem more stable for me.

So for, say, 3.6Vdimm you still need the 3.3v line of the PSU to be tweaked higher than that?

[synergy321]

i'm going to get pdp 3200 tccd 512x1, any ideas on how that would do, or should i look into some ocz3200plat. i would really like tccd, and price as low as possible, but if the ocz is worth it on here, then i can do it. thanks! and keep the tests going and hurry up and spill the secret

[jikdoc]

for the VX if oyu haven't try

TRFC OF 15
TWR 3
TRTW 3
READ PREAMBLE 5.5

SHOULDN'T CASUE A HIT OF PERF MUCH AT ALL AND SHOULD LET YOU GO HIGHER ON THE FSB

do you give michael jordan basketball tips as well? bigtoe works for ocz and dfi sent him 2 of the 4 retail boards in the wild to qualify memory for the board -> i'm sure he knows what settings to use!

[Bennah]

So for, say, 3.6Vdimm you still need the 3.3v line of the PSU to be tweaked higher than that? "Yes" or "no" would suffice.

What's the point then on the ability to switch the Vdimm voltage to be powered from either 3.3 or 5v line, by user selection, as Oskar commented before a week or two ? Or was that only a feature of the beta boards and is now removed?

No. With the 5v setting it will draw the vdimm from the 5v line. This puts more heat on the vdimm mosfets and around that area, so it needs to be cooled good.

With the 3.3v setting enabled, it would draw the vdimm from the 3.3v line. Max vdimm would be 3.3v but this is 'better' for users 24/7 not running high fsb's and not having to add extra cooling to the vdimm mosfet area.

With OPB using only 2.9v for his GsKill TCCD why use the 5v feature when you can use the 3.3v feature and not put as much 'strain' on the vdimm mosfets.

[mrlobber]

No. With the 5v setting it will draw the vdimm from the 5v line. This puts more heat on the vdimm mosfets and around that area, so it needs to be cooled good.

With the 3.3v setting enabled, it would draw the vdimm from the 3.3v line. Max vdimm would be 3.3v but this is 'better' for users 24/7 not running high fsb's and not having to add extra cooling to the vdimm mosfet area.

With OPB using only 2.9v for his GsKill TCCD why use the 5v feature when you can use the 3.3v feature and not put as much 'strain' on the vdimm mosfets.

I know all the points. But notice what bigtoe said (what I quoted), he has tweaked the pot of the 3.3v line of his Powerstream to 3.6v, to allow 3.5Vdimm. At least that's how I understood the quote. That would indicate, he didn't use the 5vdimm option even though he tested BH-5 and VX, which are the voltage hungry chips. Why? Did the board become too hot even if actively cooled? Was he too busy to mess around with the respective jumpers? Many speculations might be derived... So let's wait what he has to say on that himself

[Tony]

I had stability issues using the 5V setting thats why i went with the 3.3V line tweak to give me upto 3.5V.

[mrlobber]

I had stability issues using the 5V setting thats why i went with the 3.3V line tweak to give me upto 3.5V.

Was that one of the "minor issues" you were talking about in another thread?

[bachus_anonym]

I know all the points. But notice what bigtoe said (what I quoted), he has tweaked the pot of the 3.3v line of his Powerstream to 3.6v, to allow 3.5Vdimm. At least that's how I understood the quote. That would indicate, he didn't use the 5vdimm option even though he tested BH-5 and VX, which are the voltage hungry chips. Why? Did the board become too hot even if actively cooled? Was he too busy to mess around with the respective jumpers? Many speculations might be derived... So let's wait what he has to say on that himself

it's simple as that:
switching to 5V generates a lot of heat on the vdimm mosfets and may cause instabilities if not cooled properly. but if one has no PSU with adjustable rails (be it either modded or factory pots) one needs to use that feature to get more than 3.1-3.2v
on the other hand, if one has the right PSU that gives him enough juice for feeding memory, then just stays with the 3.3v option by cranking up 3.3v rail on the PSU.

[kiwi]

nice
I'd like to see platinum rev 2 sticks

[Tony]

Regarding the minor issue , yes the 5V line for me was unstable, that doesn't mean it will be unstable for you guys.
I also just tested ultra high HTT, i won't be breaking any records with my EB and 400+HTT on this one but i haven't given up yet.

If you guys want screenies of the bios just let me know..i will do my best to post them tomorrow.I will also enable the audio and test with the daughter card, its high def and deserves a looking into