DFI looks to have a winner here.

[G H Z]

Anyone try high vdimm yet, say over 3.5v via 3.3 modded and beta BIOS?

[Zeus]

Here is a nice HTT for me so far. Not a high overclock.
And I decided to take it out of the case

I didn't want to ruin my DFI box so I used my old MSI K8T Neo box lol

Edit! 400x5.0 I don't see why some people can't even hit 300 on HTT.

5x400

How many sticks of RAM? 1?
Which bios?
Settings?

[HKPolice]

CodeRed, I think you should provide a rough guide to include with A64Tweaker. Tell people which settings improve performance, and which are for stability, etc, just general stuff?

Thanks

[eva2000]

Some quick results no screenies yet still benching but

i'm at 10x 260HTT with 1:1 ram @260mhz 2-2-2-6 at 3.28v on this board now

1T 8-12-2-2-2-2 7x, 16x

Sandra 2004 Pro SP2b
Buffered = 4033/4036 MB/s
Unbuffered = 2428/2484 MB/s

SuperPi
1M = 33s
2M = 1min 17s
4M = 2min 51s
8M = 6min 30s

Everest Home
Read = 4037 MB/s
Write = 1721 MB/s

Hexus Pifast = 47.11s

Memtest86+ v1.26 = 1953 MB/s bandwidth

[CodeRed]

CodeRed, I think you should provide a rough guide to include with A64Tweaker. Tell people which settings improve performance, and which are for stability, etc, just general stuff?

Thanks

Yeah I know ...

I really need more types of memory to test on. My BH5 runs with the most aggressive timings so thats all I know. I havent a clue what the best settings are for each and every brand/type of memory ... people have to test.

[ColdFusion]

Is anyone using OCZ EL Gold PC3700 rev3 with the DFI board? What FSB have u managed to acheive, and what timings/voltage??

Cheers

[eva2000]

Is anyone using OCZ EL Gold PC3700 rev3 with the DFI board? What FSB have u managed to acheive, and what timings/voltage??

Cheers

yeah i have 2x 512MB OCZ PC3700 Gold rev3 here will get around to testing them.. but may take quite a while heh

[Bourch]

Eva, what freq is your cpu running at?
There is something that doesn't match when I compare your results and mine.

I also have Sandra SP2
My timmings are far from yours -> 2.5-4-3-3 1T 2CMD
2x512 ADATA PC4000 running at 500DDR

I got about 37xx Mb in the buffered test but If I do the unbuffered test I get about 30xx Mb!

The option to do the unbuffered test is the "Enable Buffering/block Prefetch benchmarks" one? What's wrong? I'm sure that mine should be lower than yours at the unbuffered test

My CPU is also far of your superPI time. Mine takes between 1:26 and 1:29 on 2M

[eva2000]

you're running unbuffered incorrectly you need to uncheck 9 boxes shown here



loose mem timings = slower superpi times

[andL64]

so u guys "r" searching for samsung TCCD chips?

XMS-PC3200 XL rev 1.1 (2-2-2-5)
Samsung 4ns TCCD

DDR PC3200 Platinium Edition rev2 (2-2-2-6)
Samsung 4ns TCCD

A-data Samsung TCCD PC4800 DDR600 ram

new A-Data DDR 600

u could also go for the 3500 and 3700 EB on oc-wear.de i think the ship every where ( in european room)

hope my dfi come this week to hit the shi/t out of my hardware :P

if u want good bh-5/6 and cheat search ebay for it because so many people didnt get it that they ve got bh-5 they bought their memory far ago... so its good for us !

seEn andL

PS: yeah do it codered its a must to stabilizise ur tweaker with more options ( very very great tool ! really! )

[SunTzu69]

Thanks for replying so quickly to my OCZ EB RAM questions...

I'm wondering why you all aim to hit a 1:1 ratio on your FSB/RAM with the A64? Some have suggested that this does not matter for the A64 because it is always using a divider with the on-board memory controller.

Is there really a performance advantage to running 1:1 on an A64 ? Or is it solely an achievement one strives for in the benchmarking world ?

Thanks !

[eva2000]

AMD64 performance all depends on how high your mem speed goes at tight timings, using high HTT with dividers on mem = lower performance unless with dividers mem is at higher speed than what mem would of been at 1:1

well that's what i have read

[Bourch]

The bios released some days ago says:
Fixes - Temp reading wrong(on some)

Are your temp readings fixed?
When posting, people could say if their mobo now reads the temps OK or not.
Just a way to see if there is too many ppl like me or not cause mine is not fixed

[andL64]

u ll get better performance on 1:1 than on others on the same mhz!

so ull be faster with 240x10 1:1 than 300x8 with divider its strange but it gains u little bit i tried it....

[RGone]

...eva2000!

how high your mem speed goes at tight timings, using high HTT with dividers on mem = lower performance unless with dividers mem is at higher speed than what mem would of been at 1:1

That almost word for word is the exact results that I see first hand. "Memory" is still the very most important consideration and always has been. There is plenty of slight of hand to make impressive HTT numbers appear but when you run a real calculation or application like SuperPi that is an application and not some bench> if you don't see a 'decrease' in the time to run SuperPi then somebody is kidding somebody.

My particular memory will run 240 straight up 1:1 and superPi is 36secs. In fact right out of the box. Been playing the slight of hand divider game and finally have an HTT of 335 and a so-called memory frequency of 260 both as shown in Cpu-Z but superPi has not come back down to 36seconds yet. Hehehehehe. Smoke and mirrors and still the "memory" and as fast and as tight as you can get it is the answer to big benches except where AGP performance is taken into account. This board and likely most boards that seem to slide AGP transfers to the HTT bus can really smoke.

As always if you have super memory = super benches that have been used normally on previous platforms. My guess is that as far as real applications go the A64 will run way ahead of the NF2 style platforms but to see that something besides 3dmark and memtest will have to be run and then the real horsepower of the A64 will show itself for those things that do data transfers using the HTT communications channels. Memory and it faster is still the biggest and baddest key to running fast. Has not changed a bit.

Sincerely, RGone...

[Ubermann]

My volt reading has gone totaly mad!
Tried to reflash bios but does not work.
What to do ?

*edit* Changed psu and still same weird volts *

[p4z1f1st]

My volt reading has gone totaly mad!
Tried to reflash bios but does not work.
What to do ?

*edit* Changed psu and still same weird volts *

measure with a digital multimeter

[eva2000]

smartguardion /ihdoctor app underreports voltages and isn't accurate so ignore

[Klisp]

My volt reading has gone totaly mad!
Tried to reflash bios but does not work.
What to do ?

*edit* Changed psu and still same weird volts *

Same reason why i brought a new powersupply.

[Ubermann]

smartguardion /ihdoctor app underreports voltages and isn't accurate so ignore

Ok but cpu temp that use to be 39c is now ~53 ?

[ugp]

5x400

How many sticks of RAM? 1?
Which bios?
Settings?

It was 1 Stick and I always run it in DIMM3

RAM Divider was set at 1/2 I didn't mess with it anymore after that last night though. CPU was alittle hotter, ran about 37 Idle but at 275x8.0 right now it is at 30 Idle which I like very much.

It was just a plan and simple clock at 400x5.0 which the Vcore at 1.8 and the Chipset was only at 1.8 as well.

[eva2000]

temp measurement in software is buggy = ignore

i get same idle temps in that app at 12x 200 = 2400mhz stock and at 11x 245mhz = 2695mhz at 1.75v !

[ugp]

Is your CPU really at 54? Mine reports mine acuurately. Mine idles at 30C and depending on if the Air is on it idles at 25-27 on air. Which isn't bad at all I think.

[masterofpuppets]

Is there any news on the 939 DFI LanParty board? I haven't been following this thread and I can't read all 39 pages as I am on dialup still! Stupid internet!

[ugp]

Quick question because I want to shoot for more out of my HTT...is there any real risks of burning the CPU out settings the HTT higher than 400. I want to try 450x5.0 But I am not sure on the dangers.