A64 OC Data

Status
Not open for further replies.
Darth_Fluffy said:
:eek: :eek: WHAT?! YOU CAN ADJUST IT TO 4.0vdimm in bios?! I want me one of these boards!
Click to expand...

Prebake oven to 450, proceed to insert ram stick on ur choice of oven holder thingy, prespray pam for extra fun! :p
 
Darth_Fluffy said:
:eek: :eek: WHAT?! YOU CAN ADJUST IT TO 4.0vdimm in bios?! I want me one of these boards!
Click to expand...

I thought for 4.0V on Vdimm on the DFI you have to adjust your 3.3V+ rail higher. Usually boards cannot get above 3.3V on the Vdimm unless you adjust the 3.3V+ pot on your psu, which is very dangerous.
 
I read the original thread on that, and they say that after 3.8vdimm, adjusting it doesn't do anything. That means that you can either get 3.8vdimm, or 3.7, which is pretty damn good.
 
With Gskill 1gble..
sandracpuid2920.jpg
 
Darth_Fluffy said:
Ouch @ the timings. :p
Click to expand...
heh, i'll assume you're just kidding.. i'll estimate that he gets just under 4000mb/s of memory bandwidth at that speed, with latency under 40ns
plus 2.9ghz = very very quick
 
Still, A64s aren't as dependant on pure bandwidth, as they are on low latency. Sure, increasing the frequency at no expense of timing is a wonderful developement, but if it costs you, then I'd rather not. Remember the astronomical gains when moving from single channel to dual on AMD64? What was it, 5%? While lower timings will have much more substantial gains. Yes, 2.9GHz uber fast, but if he spends $800 on a phase changer, why get a $160 CPU? Another thing I don't understand... oh wait, that's the 3700+ :eek:. Pretty damn good for a s754 ClawHammer. Still, the latency is killing the potential. I'd use BH-5 with that particular rig, not TCCD. The BH-5 will Oc better in 754's single channel as well..

heh, his ClawHammer owns our collective ClawHammers, I bet... and that's only 1.7vcore. I run mine at 1.74, and can't even break 2.5GHz :(
 
yeah, a64's do love low latency, but he has the ram running fast enough to bring down the absolute latency to a point where it's actually just as fast as ddr400 at 2-2-2-5 timings, if not a wee bit quicker. i could do some quick math do prove my point if you wish ;)

and yes.. his does own ours. curses to needed 1.75v to be prime stable at 2.4ghz :(
(especially when i need 1.45v at 2.25ghz and 1.55v at 2.3ghz.. the voltage/speed scaling on these cpu's make no sense to me sometimes)
 
(cf)Eclipse said:
yeah, a64's do love low latency, but he has the ram running fast enough to bring down the absolute latency to a point where it's actually just as fast as ddr400 at 2-2-2-5 timings, if not a wee bit quicker. i could do some quick math do prove my point if you wish ;)
Click to expand...
Yes, at that speed 3 clocks is about as low as 2 clocks at stock speed, but 2 clocks at that speed, would be much lower than 2 clocks at stock speed. Y'dig? :D :D

By the way, nForce 3 wants your tRas to be 10. Or burst length, I prefer the latter name for that latency.
 
heh, that would be quite true, but not much ram runs at that speed and timings without mad voltage + active cooling. and it seems like ddr2 is ust trying to bring the latencies up *cries*
 
Ok, we know for a fact that he had BH-5 that he ran at 280MHz 2-2-2-X. I don't see how 264MHz 2-2-2-x would pose a problem?
 
No way TCCD could do 2-2-2 at 260MHz, one stick, or two. My suggestion is to use your BH-5, which can do that. Besides, if you mean to just have 2 cas 3 trcd 3 trp X tras, then I'd rather have 3-2-2-X. Cas isn't quite as important, than, say trcd, or trp for that matter. In fact, the difference between 2-2-2-X, and 3-2-2-X is very superficial-- sometimes, not even marginable.
 
Darth_Fluffy said:
No way TCCD could do 2-2-2 at 260MHz, one stick, or two. My suggestion is to use your BH-5, which can do that. Besides, if you mean to just have 2 cas 3 trcd 3 trp X tras, then I'd rather have 3-2-2-X. Cas isn't quite as important, than, say trcd, or trp for that matter. In fact, the difference between 2-2-2-X, and 3-2-2-X is very superficial-- sometimes, not even marginable.
Click to expand...
Thanks. I will go to voltmod route f/ my 8kda3 with bh5.
 
2x512MB BH-5 on a S754 setup is not exactly a godsend for daily use...of course if you have some old nice 256MB BH-5 sticks hanging around :D
1) Very picky S754 single-channel memory controller, good chance it won't post even at stock. Weird how the best results I've seen seem to be from 256MB sticks of old TwinMOS BH-5.
2) 2x512MB fills up all 4 address banks, not good for oc'ing very far using 1T.
3) If you are lucky you might hit 250MHz 2-2-2-7 1T @ 3.3V but to run that 24/7 you will need a fan over those (should probably take off the heatspreaders too).
4) TCCD is the best choice right now available for purchase retail. 2.5-3-3- timings aren't bad at all. If you decide to go S939, 300MHz on the memory is within sight.

It's pretty much impossible to take 2x512MB to 280MHz on a S754 setup even with the voltage options of the DFI board.

AnandTech may find tRAS @ 10 to be the best but results differ. A lot of benchers I've seen find tRAS @ 7 to be the best setting.

A64 architecture already reduces latency drastically with the memory controller on-die and the elimination of the FSB bottleneck.
 
Ahh, DaveX, my arch-nemesis. Or, ally, we can never figure out which :D. You may remember me as M4d-K10wN from such films as ...err, never mind.

Anyway, he does indeed have those tough 256Mb sticks. And why would you steer away from 512Mb BH-5 sticks? Most work great with MSI K8N Neo Platinum, I've checked on MSI's forums, just abotu every BH-5 stick is in the "workin' fine" thread. Don't know about other boards, but the memory controller is on the die, as you say. 3.3vdimm is nothing to worry about, and Some 512 BH-5 sticks can do that at 3.2 as well... It's after 3.5vdimm that you should employ active cooling, IMO. Of course, by active cooling, I mean something dedicated to cooling the RAM alone, such as a special waterblock, or ramsinks + fans epoxied over. However, up to 3.5 you should be fine on good case airflow alone. He probably knows enough about overclocking than to ask for our adice on what's good or not, though.
 
Hm... here's my current OC on first attempt...

Temps are 30c idle 39c load right now. Running it on an ASUS A8V Rev 2.0 in a Silverstone TJ-05... Tested stable in prime95 for about 4 hours now. Going to try for higher clock.

i think i can get it higher but i want to get low timings. Why are my bandwidth #'s so much higher than you guys?! Does the low cas latency make THAT much of a difference?

3000OC.gif
 
So far so good. First time I have played with an AMD 64 system. This is my new work machine. I still need to work on the CAS latency issues. I find it hard to believe this XMS 3200 is doing 450mhz, but who knows. Not sure exactly how the memory is represented with CPU-Z. I am guessing it still defaults back to 400mhz for the memory. I just got it together and setup this afternoon so I am learning it's quirks. OC seems good so far. Running prime95 currently and will let it run overnight.

Memory is @ 2.7v
CPU is @ 1.6v
Board is an Asus K8N
Memory is the cheapest 512mb 3200 XMS, cas 2-3-3 rated
300wt In-Win powersupply.. or whatever is is that comes with an In-Win case
Leadtek GF4mx 64mb (hey it's a work computer ;) )

zeek-oc_11-9-04.JPG
 
Well, I suppose he's using a phase changer, but how stable will the mosfets be with that kind of current? I'm surprised a voltage spike hasn't killed his CPU yet.
 
Didn't you post this thing a while ago how this guy on Xtreme got a fx-53 Sledgehammer to run at 3.6GHz with his dual cascade setup?
 
Darth_Fluffy said:
Didn't you post this thing a while ago how this guy on Xtreme got a fx-53 Sledgehammer to run at 3.6GHz with his dual cascade setup?
Click to expand...

that was very long ago and it was only benchable @ 3.2GHz...this FX-55 @ 3.7GHz is fully benchable
 
Yes, and when I said it here in the "amazing" 3GHz fx-53 overclock, back when s939 was brand new, everyone started flaming me; and I asked you to give the linkage and you refused. You sly boots, you... :D I'm too lazy to hold a grudge, though.
 
Darth_Fluffy said:
Well, I suppose he's using a phase changer, but how stable will the mosfets be with that kind of current? I'm surprised a voltage spike hasn't killed his CPU yet.
Click to expand...

with good cooling it doesnt matter
according to those experience high vcore guys over @ xtreme.org
 
That's pretty sweet for only 1.65v, 'specially at 1:1 and with those timings. :cool:
 
Impulse said:
That's pretty sweet for only 1.65v, 'specially at 1:1 and with those timings. :cool:
Click to expand...

yeppers :).. I was running it @ 1.7 earlier but then decreased the vCore to check the stability.. and it did 9 1/2 hrs straight on Prime95 without any errors :D
 
dfx said:
with good cooling it doesnt matter
according to those experience high vcore guys over @ xtreme.org
Click to expand...
So, is that why those very same guys on Xtremesystems killed a bunch of A64s, phase changers, or no phase changers?
 
Darth_Fluffy said:
So, is that why those very same guys on Xtremesystems killed a bunch of A64s, phase changers, or no phase changers?
Click to expand...

phase changers
 
3500+ Newcastle @ 3.05ghz with 1.85 actual v-core
or
3500+ Newcastle @ 2.98ghz with 1.7 actual v-core
Both setting Prime 95 stable. Still haven't pushed how high bench setting I can get but preliminary around 3.15ghz... :p Gotta wait for my 2 sets of TCCD (one coming in today and one tomorrow :p) for pushing it harder.
Cooled by prometeia Mach II modified by baker 18 priming around -46c evap on 26c ambient ;P
As for mosfet, it doesn't get too hot considering the power I am pushing through mobo... around 58-60c... there are some air coming from fan situated above memory. Think neo2 just have used decent parts on it's PWM circuit at least to make it maybe more efficient... or the heat is getting absorbed through CPU :p
 
Darth_Fluffy said:
So, is that why those very same guys on Xtremesystems killed a bunch of A64s, phase changers, or no phase changers?
Click to expand...

Pffft. If you're not willing to sacrifice some hardware, don't overclock agressively.

Even with good cooling, he will be shortening the useful life of his processor significantly. But, honestly, the useful life of a processor at stock voltages is way beyond the amount of time anyone keeps them for. And, it COULD die suddenly, but honestly, if it's lasting this long, it's unlikely to die before he wants to replace it anyway. It's not like anyone running phase-change cooling is planning to keep their system unchanged for years on end.
 
etruscan said:
Pffft. If you're not willing to sacrifice some hardware, don't overclock agressively.

Even with good cooling, he will be shortening the useful life of his processor significantly. But, honestly, the useful life of a processor at stock voltages is way beyond the amount of time anyone keeps them for. And, it COULD die suddenly, but honestly, if it's lasting this long, it's unlikely to die before he wants to replace it anyway. It's not like anyone running phase-change cooling is planning to keep their system unchanged for years on end.
Click to expand...

Never burnt a single CPU yet... that is with some extreme method and through about 20-30 cpu :p Most of them didn't stay longer than 6 months though... With good cooling and decent sense of where to stop, it isn't too much risk
Can't say same about vmodding GPU though... Think I ate through 4 high end card in last 2 years. 9500pro->9700pro->9800pro->6800GT is on my dead list through volt modding.
 
jinu117 said:
Never burnt a single CPU yet... that is with some extreme method and through about 20-30 cpu :p Most of them didn't stay longer than 6 months though... With good cooling and decent sense of where to stop, it isn't too much risk
Can't say same about vmodding GPU though... Think I ate through 4 high end card in last 2 years. 9500pro->9700pro->9800pro->6800GT is on my dead list through volt modding.
Click to expand...

Of course, neither have I. But, when you OC the hell out of it, using some extreme/whatever methods, you're not going to break down crying and complaining about the unfairness of life if it suddenly stops responding, are you?

Yeah, I killed a 6800 GT. Overvolted it, and the memory went dead on me, stupid $400 doorstop. But my second one overvolts fine.
 
Status
Not open for further replies.
Back
Top