Post Your AMD 64 Bandwidth Shots.

[Kunaak]

Just something for fun.
lets see how high you can go.

please list a few details.
what kinda ram, board, cas, voltages.
I just want to see how high they can go.

oh yeah, if you post a FX shot-please note that too

[koensa]

since this is kinda the tread
what mobo u would advice?and wich chipset?
i'm thinking the shuttle an50r or asus kt800
or wich ones are better? i just want 250fsb

=2500mhz?under a prommie seems possible right?

[shortcircuit]

4010MB ..... with an opener like that I think others will be embarassed to post their scores. I was psyched to hit 3700mb. I'll post mine tonight, on both NF3 and Via.

Koensa, the question you ask is asked every 2-3 days by someone here. Just search around, see what the people posting front-page scores on the ORB are using, and try the same... (hint hint NF3 cough cough Gigabyte).

[Evil_Spork]

it looks to me as if an athlon(vinilla) has more bandwidth that an a64 from that shot. or matbe im hilucinating. i got over 3500m/s at 230fsb on my nf2 board.. or is it the single channel thats killing the nf3 bandwidths?

[ctgilles]

Originally posted by koensa
since this is kinda the tread
what mobo u would advice?and wich chipset?
i'm thinking the shuttle an50r or asus kt800
or wich ones are better? i just want 250fsb

=2500mhz?under a prommie seems possible right?

250 1:1 You'll need good RAM and lotsa voltage and lotsa luck
I'd go for the VIA, but the nf³ seems to be easier... (well, for as far OC is easy it has locked AGP afaik), ah, heck, go for the VIA

[Evil_Spork]

holy!!! 387% efficiantcy!!!! omg!

yes i knwo im looking just at numbers here and not the whole picture. can you explane more what the efficiancy number there does and how it helps? (ive heard theirs somethign weird with sandas efficency number, not sure)

[shortcircuit]

Here's my best shot for this evening.

Gigabyte K8NNXP
A64 3200+ @ 9x300, 1.96v
256mb Corsair XMS3500 @ 3.2v
Prometia cooling

This is at 300fsb and 2700mhz with the /11 memory divider, giving 245mhz on the memory with 6-2-2 Cas2. Tried it with 5-2-2 also but got a lower score. I can't seem to get this board stable at 1:1.

This was done with my "dead" A64 chip. Seems that instead, the new abit kv8-max3 that I used to diagnose the A64 was DOA. I'm not sure if I should be happy or sad, but now I have 2 A64's... the old one clocks a little bit better.

[bowman1964]

Well if my stuff was up and running...i had a bunch of 4300+ shots on sandra...
that was running 280fsb 1:1 but unbenchable 3d

[Henry]

Here's mine :

Gigabyte K8N
Athlon64 3200+ @ 10x248 1:1
Winbond BH-5 @ 6-3-3-cl2.5

[mikem]

Koensa....
The Asus is good upto 250 with good memory (unless your name is macci) and is somewhat faster than the Nforce boards clock for clock...if your going beyond that get the Shuttle....

My 2 cents since I own both....

[shortcircuit]

I'm starting to figure out this gigabyte board... although I still cannot get 1:1 stable with cas2 much over 210mhz

This is with twinmos pc3700 and 3.25v. I still have more ram to test (one of the many benefits of having a folding@home farm). This stick surprised me, I was never able to get it much over 250mhz before.

[drunkenmaster]

shortcirucuit, is that volt modded up or does the knnxp version have higher voltage options, do the stock options actually work, vdimm and vcore? or stuck like on the k8npro?

And mikey, its not at all close to 347% efficient, the efficiency merely compares the estimated bandwidth of the cpu to the actual bandwidth sisoft reads. Its reading the cpu sestimated wrong, like in shortcirc's last shot, its reading it at 29Mhz. Diff version seem to read them all wrong, and diff boards too. You can't get higher than 100% efficiency in theory, in practice you could never beat, really, 96%? and most don't hit that, and won't. P4's efficiency is dire, but its all based on a silly comparison that doesn't show anything.

[shortcircuit]

The bios voltage options do work, but they are limited (vcore + 0.2v, vdimm +0.3v). I'm using the S940-F3 bios, and I've done both the vcore and vmem mods.

[shortcircuit]

Bump.

[mikem]

Damn......nice work!!

[JCviggen]

shortcircuit...

looks nice, but i wonder if the terrible latency wont make performance worse than 2-5-2-2 at 220-230 MHz... I know that your RAM cant do that though...looks like the same chips being used for PC4000-4200...sucks ass
It will bench terrible in everything except sandra

[shortcircuit]

Yeah at those timings everything but sandra sucks.

I think my best bet for benching 3dmark is 320x8.5 = 2725mhz with the /11 memory divider (248mhz) and my Corsair XMS3500. Too bad I only have a GF3 and a sucky 256mb 9600p to bench with for now. I'll start benching again with the GF3 soon so I can work all the bugs out of the system before my 9800XT returns.

Regarding memory in this gigabyte board, so far I have noticed:

256MB Corsair XMS3500 BH-5 runs great. 250mhz+ at cas2 with low timings.
256MB TwinMOS PC3200 BH-5 doesn't run good.... can't get cas2 at high speeds although it's done it in other boards. I think this board doesn't like single-sided BH-5 sticks (the corsair is dual-sided). I tried three sticks.
256MB TwinMOS PC3700 with 43B chips run great at lazy timings, bootable up to around 285 and can do some benchies around 290. I tried 2 sticks.
Corsair PC4000, both 256MB and 512MB (I assume with hynix chips) totally suck ass and will not even do DDR500 at SPD timings. But at least it has blinky LED's.

[mikem]

Here is my best with low latency...

[MrIcee]

Heres a little preview of my testing last night as I just recieved an A64 back from RMA yesterday afternoon. Trying to find the limits of the Shuttle AN50R at low multipliers. Utilizing the newest clockgen for FSB and multiplier manipulation, and volt modified the Vdimm to provide unlimited voltage, I was able to push the OCZ PC3500 Platinum Limited Edition as high as 290Mhz 1:1......this photo illustrates the Shuttle's capabilities with only the Vdimm mod at 6:5 ratio. The great thing is that the bandwidth produced is only at 1653Mhz

Randi

[Tony]

Good work MrIcee, nice to see our ram up there with the best.

[mikem]

Very nice MrIcee....I hope this time around the cpu/mb works out very nice for you.....does your new cpu look promising for high oc so far?

[MrIcee]

mikem

So far it is working very well indeed. The first thing I did was pull out my trusty old Promise Ultra133 TX2 Ide controller and put my WD JB on it. That alleviated the problems I was having before with shut downs and reboots at high FSB.

The A64 i just recieved is a week 34 CAAKC whereas my first 2 were week 41 CAAMC's. I do not know about clockibilty yet as I'm working on developing a stable 3DMark FSB speed. When in this mode I stay at one speed for a baseline, in this case 2000Mhz, the chips stock speed. I ran 3DMark 6.5x308Mhz FSB w/ the ram at 250Mhz 2-2-2-5 @ 3.5v successfully.

The dividers are unusual on these nForce3's, as it seems that they change depending on what multiplier you are at. With the memory set at 166Mhz....when at 5x the FSB is 331Mhz and memory at 275, CPU/6.

When at 6X....running at 334MhzFSB it is 250Mhz for the memory, CPU/8 according to CPU Z. Obviously at this ratio the bandwidth is not as great as my 5x screenshot but still darn good. What baffles me is that the memory goes to a lower ratio ie: 308/250 @ 6.5x vs. 334/250 @ 6x.

I have successfully benched 3DMark2001 at 334/250 as of today(screenshot coming soon for confirmation) and am incrementally trying for more. Running LDT at 2x for these runs too. 22,587 3dMarks at 2Ghz w/ stock Radeon 9800 Pro at moderate 460/760 clocks. The Asus Ti4800 will be here Monday:thumbsup:

The AN50R is an arse kicker once you get a good CPU and IDE controller card

Randi

[MrIcee]

Thanks Jason

Just needed the tools to get going with A64. This is the 3rd A64 in the last month...this one seems to take the 3.7v I shove up the OCZ's butt and it's memory controller to run max timings at 275Mhz + higher. The first 2 late vintage(Week 41 CAAMC's) only lasted days while I was working on bandwidth prior to 3DMark benching.

Randi

[MrIcee]

Jason

The Shuttle has sustained no injury thus far. It was my previous 2 A64's that stopped posting after attempting to run 1:1 at high mem voltage and 270-280Mhz FSB. It's speculation on my part that the high voltage took out the memory controllers on the A64's, but let's face it...I was at a low multiplier to keep total Mhz under 2000, so I could use stock Vcore. The only high voltage being run was to the memory. LDT voltage was at 1.4v as it is right now.

Randi

[charlie]

Great Work, Mr. Icee!!

Nice to see the Shuttle ripping along...
In fact my next benching rig will probably be AN50R + 3400+

C