winchester overclock database

the problem is that i'll then have to ask the htt speed, cpu muli and ram divider settings in order to calculate the true speed.
it's definitly doable though.. and possibly even useful.

mamma, i even pm'ed kyle about it. he didn't think it needed sticky :(
though at this point, it wouldn't bother me if it didn't get stickied anymore, the venice oc database is the one that really needs to get up there.. word needs to get out BEFORE peopel put in their cpu's and forget to write down the opn ;)
 
Viper87227 said:
I would have to think that any memory dividor used would be more because the ram itself is not up to the high FSB. Someone who just dropped $300 on a winny and good OCing board doenst necesary have another $300 to drop on hi grade TCCD, so alot of times you see peopel overclocking with cheap ram and using a dividor. While it can be worth while info, waaay to much of a pain in the ass do it now. Perhaps the venice chary eclipse?
Click to expand...

While this is anecdotal I've seen guys report memory going pretty high on FSB with the CPU turned down. When they then try to ramp CPU and memory together they hit bottleknecks in the 2.4-2.5Ghz range. Applying a RAM divider then allows the CPU to clock to higher speeds despite an, on the face of it, inferior memory controller.

From what I've been seeing with my CPU this may be because going above a particular memory bandwidth value (can't remember the value but will track it on my next o/c run) seems to lead to problems - which may or may not be cured by higher voltages.

Following my theory, applying the RAM divider keeps the memory bandwidth low enough for the memory controller to handle the throughput. That said, I'm not saying people should go out and buy expensive RAM or FXs. Rather, I'm just pointing out that the quality of the memory controller seems, to me, too important an area to be overlooked in gauging relative performance. I do acknowledge the extra adminsitrative overhead that would impose though :D

That said, I may well be getting the wrong end of the stick in how I see all of this working. I am only the fount of some knowledge and not the fount of all knowledge ;) The memory controller may not be such an important factor after all.
 
laserbeam said:
From what I've been seeing with my CPU this may be because going above a particular memory bandwidth value (can't remember the value but will track it on my next o/c run) seems to lead to problems - which may or may not be cured by higher voltages.
Click to expand...
that's a good idea. i have a relatively low limit... but i can't seem to place it. it's somewhere around 235-245mhz. i never get errors in memtest, but windows will randomly reboot on occasion :(
even so.. if i get enough spare time, i'll see if i can track down that limit
 
(cf)Eclipse said:
that's a good idea. i have a relatively low limit... but i can't seem to place it. it's somewhere around 235-245mhz. i never get errors in memtest, but windows will randomly reboot on occasion :(
even so.. if i get enough spare time, i'll see if i can track down that limit
Click to expand...

You're seeing roughly the same behaviour as me
 
yeah, it's definitly the memory controller. try using one stick of ram. i get get to 250mhz at 2.5-2-2 that way. haven't tried for more though..
 
Hello people! well i just got my A64 socket 754 3200+ today in the mail the stepping is
CBAUC 0429VPMW does anyone know about this stepping and how it may clock still waiting on heatsink/fan will be running it on a msi neo fsr board.

Thanks
Jerry a.k.a DieHard_OCer
 
sorry mate. that's not a winchester, from the stepping, i'm guessing you got a 2.2ghz newcastle.

edit: though i would try to be more helpful, i don't know much about newcastle steppings. that's a really early one.. so i wouldn't expect anything awesome, but you never know :D
 
(cf)Eclipse said:
yeah, it's definitly the memory controller. try using one stick of ram. i get get to 250mhz at 2.5-2-2 that way. haven't tried for more though..
Click to expand...

Something else that backs the idea of the memory controller screwing with people's overclocks is the fact that I've run 3 different sets of dual channel in my rig. With going into the details I've seen that the faster the memory (by recorded bandwidth at a given speed) the lower my stable prime is at a 10xCPU multi. So, 'cos I'm running VX which is the fastest memory I've used up to now my stable prime at stock volts has actually dropped from about 2350Mhz to 2290Mhz. I bet if I was running cheap RAM I'd be prime stable at 2400! I'm not complaining but it is ironic don't you think?

It does appear that a very significant factor in the rating, and overclockability/ultimate all round performance, of these chips is the quality of the memory controller. The ability of venice to handle 4 ram sticks at DDR400 tells me that the memory controller is significantly more robust and scalable. I think it's this that's pointing towards people thinking venice overclocks like crazy. What I think they're actually seeing is, loosely speaking, what winchesters could do if not held back by the memory controller.

In short I'm thinking that the winchester memory controller is a prototype whereas the venice m/c is the finished product.
 
yeah. i might make another thread to find max cpu oc at a constant voltage and varying ram dividers now... fortunatly, this applies to everyone who has ram that can do somewhat respecable speeds.


also, updated picture for NHO. sorry i forgot about you mate ;)
 
(cf)Eclipse said:
sorry mate. that's not a winchester, from the stepping, i'm guessing you got a 2.2ghz newcastle.

edit: though i would try to be more helpful, i don't know much about newcastle steppings. that's a really early one.. so i wouldn't expect anything awesome, but you never know :D
Click to expand...

Ehg, its not as early as you think. Its a Socket 754 NewCastle, which have been around quite a bit longer than winnys. By all means, you could actually say thats a pretty late chip for the s754 run.
 
relatively speaking it's not late, no.. but considering that he just purchased it, it is. i would have epxected at least a week 40

also, my clawhammer that i bought in august is a week 20.. i just found that kinda amusing
(aside from the fact that it's 0420 ;))
 
(cf)Eclipse said:
that's a good idea. i have a relatively low limit... but i can't seem to place it. it's somewhere around 235-245mhz. i never get errors in memtest, but windows will randomly reboot on occasion :(
even so.. if i get enough spare time, i'll see if i can track down that limit
Click to expand...

My mem controller is kinda wierd. My Geil ram can run at 270Mhz...but get this, I can't run it at 270htt 1:1...I have to run it at 300htt with a 180 divider to get 270Mhz to run stable. screwy. Tried voltages and everything.
 
interesting. the opposite happens to me. my board hates taking the htt over the memory controller's limit, even if i run with dividers.
i can actually memtest at 250mhz 1:1 with lots of voltage, but if i try 250mhz 5:6, it fails miserably :confused:
 
(cf)Eclipse said:
interesting. the opposite happens to me. my board hates taking the htt over the memory controller's limit, even if i run with dividers.
i can actually memtest at 250mhz 1:1 with lots of voltage, but if i try 250mhz 5:6, it fails miserably :confused:
Click to expand...

What memory?
 
(cf)Eclipse said:
crucial ballistix and a dfi nf3-ut
Click to expand...

Well dont blame yoru CPU then. Ballistix is really picky ram according to what Ive heard. I have no idea why Crucial didnt go with TCCD.
 
not really. i can do 2.5-2-2 up to the limit of my memory controller. going to 2.5-3-3 or even 3-4-4 doesn't help any. also, i can do 2.5-2-2 at any speed between 200 and 245mhz with 2.7v. it's just getting over that that's a pain.
though i can do it with one stick in, no problem. and yes, i've played around with the dimm slots. tried every combination, two work well, one is good for getting to 225mhz at 2-2-2, the other is better for 2.5-2-2.. interestingly enough
 
(cf)Eclipse said:
not really. i can do 2.5-2-2 up to the limit of my memory controller. going to 2.5-3-3 or even 3-4-4 doesn't help any. also, i can do 2.5-2-2 at any speed between 200 and 245mhz with 2.7v. it's just getting over that that's a pain.
though i can do it with one stick in, no problem. and yes, i've played around with the dimm slots. tried every combination, two work well, one is good for getting to 225mhz at 2-2-2, the other is better for 2.5-2-2.. interestingly enough
Click to expand...

Far be it from me to question you....I dont need a lashing ;)

I find it odd tho that your memory controller is that crappy, most A64s I dont see the mem controller holding the ram OC back, its usually due to crappy ram.
 
hehe, nah. i don't get into arguments until people dissagree without reason. then i swoop down like a feather attacking water! bwahaha.

and i have a week 20 clawhammer. that might explain a few things. there's just some very.. messed up things with the memory controller. with my bh-5, i couldn't get past 209mhz at 2-2-2 at 2.85v, when i knew the same sticks could do 220mhz at 2-2-2 on my asus a7n8x with 2.8v :confused:

sounds like i need some advice. i'm thinking a mobile claw 3400+ sometime.. maybe for my b-day in may. hopefully newegg won't be out of stock. though that's if newark is out and is a wash. if it isn't, i'm getting a 1mb, 90nm DTR :D
 
I finally got my CPU today. Rememver, stepping is ADA3000DIK4BI CBBFD 0432MPM.

I was pleasently surprised when I clocked up to 2.35ghz at 1.36v. After that, it was BSODing galore. Raised the voltage to 1.55v, which took me to 2.5 stable. 1.6v took me to 2.6 stable. Right now, im looking at 1.7v, 2.7GHz stable, at at this voltage I have suicide shots at 2.85ghz. I tried to get 2.8 stable with 1.8v, but it would have nothing to do with it. Cooling is on water, just as last time.
 
hmm, a CBBFD :p
let me know which combination you want to use for the database, just as long as it's a stable one :D
 
(cf)Eclipse said:
hmm, a CBBFD :p
let me know which combination you want to use for the database, just as long as it's a stable one :D
Click to expand...

Combination or what? You got me all confused.
 
Viper87227 said:
I finally got my CPU today. Rememver, stepping is ADA3000DIK4BI CBBFD 0432MPM.

I was pleasently surprised when I clocked up to 2.35ghz at 1.36v. After that, it was BSODing galore. Raised the voltage to 1.55v, which took me to 2.5 stable. 1.6v took me to 2.6 stable. Right now, im looking at 1.7v, 2.7GHz stable, at at this voltage I have suicide shots at 2.85ghz. I tried to get 2.8 stable with 1.8v, but it would have nothing to do with it. Cooling is on water, just as last time.
Click to expand...

Can you give me an idea of what temps (idle and load at the various voltages) you've been getting during this process?

I ask because I'm starting to overclock mine after reseating and putting it through a few thermal cycles. I get to 2.30Ghz at stock volts before getting Prime95 errors. Inching voltage up from 1.36 to 1.40 has still produced errors but I'm concernd by the way my load temps. have risen from 38-39C to 40-41 with such a modest voltage increase. Idle temps are 25-26C.

Interestingly, I've seen no gain from "bedding in" my thermal interface. Eclipse is aware I've had temperature delta issues with this cpu before. If I've still got delta trouble - things are alot better but don't seem ideal - I might look at lapping my HS as I've observed a slight concavity to the base. Consequently, the best as5 application i've made has been to apply a blob centrally, 2/3rd towards the front of the core.
 
laserbeam said:
Can you give me an idea of what temps (idle and load at the various voltages) you've been getting during this process?

I ask because I'm starting to overclock mine after reseating and putting it through a few thermal cycles. I get to 2.30Ghz at stock volts before getting Prime95 errors. Inching voltage up from 1.36 to 1.40 has still produced errors but I'm concernd by the way my load temps. have risen from 38-39C to 40-41 with such a modest voltage increase. Idle temps are 25-26C.

Interestingly, I've seen no gain from "bedding in" my thermal interface. Eclipse is aware I've had temperature delta issues with this cpu before. If I've still got delta trouble - things are alot better but don't seem ideal - I might look at lapping my HS as I've observed a slight concavity to the base. Consequently, the best as5 application i've made has been to apply a blob centrally, 2/3rd towards the front of the core.
Click to expand...


To be honest, I cant tell you how much my temps were effected each voltage bump, but it wasnt much. I cant tell you my idle temps because I run F@H, but there somewhere around 22c. My currect temp, again running F@H, its 38C....at 300x9 and 1.7v. I do know that before I started overclocking I was at 35C load....so I only went up 3C total...but again this is on water, so that will make my temps fluxuate much less.
 
Viper87227 said:
...but again this is on water, so that will make my temps fluxuate much less.
Click to expand...

Right...explains alot. I'm on air but am not sure if I'm getting full value out of my XP-120/Panaflo combination. In the not too distant future I'll go to water, but not yet.
 
Viper87227 said:
Combination or what? You got me all confused.
Click to expand...
you listed about 3 or 4 different speeds at different voltages. which one do you want me to put up? :p
 
I've got 2500/1,33=1880!!!



My CPU is am 90nm Athlon 64 3500+, DIK4BI CBBID week 0503

Voltage: Default
HTT: 250x4
CPU: 10x
RAM: "180" MHz

AMD3500+@2500 MHz w/box cooler, Corsair 3200XL @ 227MHz
MSI K8N-SLI Platinum
 
Stability means hours of prime95, not minutes. You have primed for hours haven't you?
 
anyone had any luck burning in winchesters? I've d/l that cpu burn-in program and followed the instructions to "burn in" the cpu. I set my speed to 2500 @ 1.5v (just where it said, the point at which it "seems" stable in windows with minimal voltage but won't run Prime or SuperPi without errors).

So far, I've burned it in for a total of about 24hours. Max temp I ever got was 38c (idle is 27c) with the door to my room closed (pc is in a "closet" under the stairs). Haven't noticed much improvement.
 
(cf)Eclipse said:
burn-in is a myth ;)
Click to expand...


DAMNIT! :mad:

lol...what's the stock voltage on these winchester's anyway? I did notice that I can run mine at 1.26v at stock speed. almost gets the idle cpu temp to 25c where the cpu fan shuts off :p
 
(cf)Eclipse said:
you listed about 3 or 4 different speeds at different voltages. which one do you want me to put up? :p
Click to expand...

put up 2.7ghz 300x9 at 1.7v on water. Thats my stable for this chip. Leave the results for my other chip[ up there.
 
I tried doing that CPU "burn-in" technique, with my processor at 2500mhz, and 1.5 volts from 1.6. At first my CPU was only prime stable for 1 minute, and now it is prime stable for 5 at 1.5 volts.
I am going to "burn-in" another 5 hours and see what happens then.
 
myth or not...it does seem to do *something* since now I'm able to run at 2250Mhz @ 1.36v whereas before I had to be at least 1.42. I've been burning it in about 12 hours a day for the past 4 days.
 
Here are the steps toward the burn-in, in a condensed fashion.

1. Lower vcore to minimum stable levels where Prime95 would probably run for only 2-5 minutes.
2. Use a program like CPU-Burn-In overnight. After the first burn, you should have increased your stability in Prime95 so that it runs stable for well over 10 minutes. The point of using CPU-Burn-In is to keep stressing the computer even if it produces errors, unlike Prime95 which stops on the first error.
3. Repeat the process and lower the vcore even more, until it is within what you want
4. Now that you have lowered your voltage, you can push the processor higher to see if it responds any better, it should. While you are pushing it higher, do a burn in for 4-5 hours with each 1 mhz increase on the FSB.

This is what Sentential wrote, only edited, so it is his work!
 
dmonkey said:
you can push the processor higher to see if it responds any better, it should.
Click to expand...
no, it shouldn't ;)

not to be an ass, but burn-in for cpu's really is a myth. there are some flukes, and some people who seem thoroughly convinced that it does work.. but if you know much about how cpu's work, the method of burn-in should just make the chip slower :rolleyes:


but, since you two seem to have gotten higher mhz out of it, i'm not gonna argue you, just state my beliefs :D
 
i can run stock speeds at 1.1v stable....but it takes 1.55v to get 2.5ghz....the lower voltage didnt affect my idle at all but lowered load 3-5C and 1.55 takes my idle up 1-2C and load up 3C

btw i use to only be able to take is down to 1.3v so cpu burn in must be true for me :p
 
(cf)Eclipse said:
no, it shouldn't ;)

not to be an ass, but burn-in for cpu's really is a myth. there are some flukes, and some people who seem thoroughly convinced that it does work.. but if you know much about how cpu's work, the method of burn-in should just make the chip slower :rolleyes:


but, since you two seem to have gotten higher mhz out of it, i'm not gonna argue you, just state my beliefs :D
Click to expand...

My CPU voltage is down from 1.65 to 1.55V...sometimes it'll let you nudge the clock a bit higher but it will usually let you get away with a little less voltage than you needed at first.
 
You must log in or register to reply here.
Back
Top