• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

AMD althon 64 3200+

X

-(Xyphox)-

Supreme [H]ardness
Joined
Sep 9, 2004
Messages
6,518
Ok i have an AMD althon 64 3200+ and thinking of overclocking, i have an Thermatake heatsink/fan that keeps it nice and cool, i have never overclocked before. My pc runs nice and fast. is it worth the risk? and if decide to do it? how far above stock should i push it to?
 
I was able to get another 500mhz out of my 3200+ here is a pic.

fast.jpg


Did not take much of a vcore boost. The settings to do this while keeping the memory fsb arround 400mhz ddr is:

HTT 250x4
fsb in bios to 166mhz
multiplyer at 10x
Pic shows it all
This was done using the MSI K8N neo platinum
 
What type of cooling did you have? and was that an program you used or what?
I have an Abit KV8-Pro Mobo with an nice OC Guru, but kinda iffy on using it because i have never OC anything before... dont want to fry my CPU
 
the program he used to show your his chip specs is call CPU-Z

most of us o/c via the BIOS, I personally don't trust those Windows based overclock utilities; but then again I've never really used one, so I can't really talk about those

as for his cooling...I don't know

I personally use a MCX6400V, 92mm Vornado (w/ fanbus), on an A64 3200+ Newcastle; I only run 200mHz over stock core (218x11=2400mHz) no vMod

WOW!!! all those certs and you've never overclocked....weird
 
Hmm.. I'm surprised you guys didnt switch out the stock HSF to a more robust one. Do the other HSFs not improve the OC that much? I was looking at getting a 3200 and OCing it to 2500MHz, but (hopefully) on air. I figure it'd match up with DDR500 memory well.
 
probably b/c stock cooling is a good referance to start at; a HSF most people who buy the proc will have
 
cpu.jpg


I've got a 754 3200 running at 250x10 with no vcore bumps. Cooling with a swiftec MCX6400-V and a 92mm Panaflo fan turning at 3000 RPMs. I've got essentially no heat problems and haven't had a single crash at this speed (at least no OC related crash).

I'd reccomend not using the uGuru software. I've got the same board as you and I did all my OCing in the BIOS, I was later looking at the diagnostic part of the uGURU software and a lot of the numbers didn't match up with the real numbers.

OCing in the BIOS is pretty easy though, just slowly bump up the frontside bus until things become unstable and then back off a bit. Remember to lock your PCI/AGP bus at 33/66 MHz. Remember to watch how high you HTT bus is getting as that will cause instability if it goes too high. If it gets much above 1000 MHz drop its multiplier one number. Lots of other stuff too, but these are the basics. And really, so long as you're careful, you don't run too big a risk of frying your chip.
 
This is mine currently using a ThermalRight 948U. It's prime stable @ 1.7v. Would bench at 1.6v. There is also a huge issue with these chips and the Neo2 Plat board from MSI. The board will not boot if the temp is below -0C. A lot of people are being affected by this issue and neither MSI or AMD have commented other than "get a new hsf combo" lol.. i guess they're clueless to phase change or can't understand the term mach I / II. If anyone else if having these issues i'd love to hear some feedback by pm or in the post about whether or not it works or doesn't below -0C. Here are a few links if your having these issues:

http://www.xtremesystems.org/forums/showthread.php?t=45386
http://www.xtremesystems.org/forums/showthread.php?t=44530
http://www.xtremesystems.org/forums/showthread.php?t=45309

MSI’s forum:
http://forum.msi.com.tw/thread.php?threadid=65706

HardOCP forum:
http://www.hardforum.com/showthread.php?t=830416

I've email kyle as well as a few other people in the overclocking community to see if they can get a solid answer out of MSI or AMD as to why this is happening. I don't mean to flood the boards with info on this but it's annoying to upgrade and run on air when i have a fully mod'd mach I sit'n idle in my room :)
On with the screenshots! Let's see some more!

3200@2.65air.gif



P.S. Charlie on xtremesystems has one running @ 3ghz on air (lucky sob) :)
 
On the Response to the OC, This is my first Custom Built PC, i usually just order from Dell, But i know how to build them, just never did. But i never OC because of the risks and never saw the use, but i have been reading up on it, and i am going to OC my Althon 64.
On my Abit Mobo it has an Built in Guru Utility, Should i use that to OC it? or is it best to OC from the Bois?
And all those Cert's are for networking :D
 
If you're comfortable with the BIOS I say use that, OC'ing is half luck anyway... no need to introduce more variables into the equation. Otherwise the uGURU proggy isn't a bad option if you just want to experiment.
 
Impulse said:
If you're comfortable with the BIOS I say use that, OC'ing is half luck anyway... no need to introduce more variables into the equation. Otherwise the uGURU proggy isn't a bad option if you just want to experiment.
Click to expand...

Ok, Thanks for the info, i shale do that
 
KarateBob said:
I wonder why Kyle didn't bump the vcore up 0.05 or 0.1 volts, im sure the stock cooling could handle the small increase in heat.
Click to expand...

Well first off it was Sean writing that one, not Kyle. Kyle or Steve woulda gone into some extreme cooling as well I'd hope.

But at any rate, Sean does a pretty good job of representing the first time overclocker in that article, even if he didn't use a particularly good overclocking mobo.

With some better component selection and a decent HSF you could easily push 4000+ speeds with the lower clocked winchester parts I'd say.
 
So far from my experience, ram was the biggest limiting factor. As for my cooling, I am using water cooling because I wanted a quiet pc that I could leave on constantly, I have a overclocked XP system that is on air cooling and it sounds like a leaf blower while my water cooled pc can not be heard when it is on. I will be ordering some of the Corsair XMS-LL memory soon, so hopefully I can go 250mhz on the fsb and get a HTT of 1250mhz instead of the current 1ghz htt I have now.
 
Is there a correlation between FSB and HTT, or what? Otherwise, why would you need better memory to run a faster HTT? I'm curious because I just ordered an A64 3200+ Winchester with an MSI K8N Neo2 Plat, and I have Corsair PC3200 CAS2 ram already. I honestly have no clue whatsoever how overclocking is supposed to work now that the FSB is supposedly integrated into the chip, or something like that...I understood it fine the last generation, of course they have to go change everything.
 
The HTT is tied to the FSB in a similar way as the PCI/AGB bus. As you change the FSB the HTT bus goes with it. Unlike the PCI/AGP bus, you can't lock it and run it out of sync so as your FSB goes up, you may run into some instability from it. But, there is a multiplier for the HTT bus similar to the your CPU's multiplier. If the HTT bus gets to high, drop the multiplier and get it back down closer to stock. Stock being 200x5. If you can get your FSB to 250 MHz you can drop your HTT multiplier to 4 and everything is groovy. Also, from what I understand, the HTT bus is not being maxed out at 1000MHz, so if you need to run it a little slower its not going to effect performance too much.
 
Geno-san said:
I will be ordering some of the Corsair XMS-LL memory soon, so hopefully I can go 250mhz on the fsb and get a HTT of 1250mhz instead of the current 1ghz htt I have now.
Click to expand...

You do know there's a multiplier to keep the HT below 1k as having it above 1k simply doesn't prove stable on the vast majority of boards ('least not a whole lot over 1k)... And memory is in no way tied to this, it's just the way the HT works. It doesn't even mater in the scheme of things since the HT is rarely saturated and even at 800mhz it doesn't seem to affect performance...
 
Impulse said:
You do know there's a multiplier to keep the HT below 1k as having it above 1k simply doesn't prove stable on the vast majority of boards ('least not a whole lot over 1k)... And memory is in no way tied to this, it's just the way the HT works. It doesn't even mater in the scheme of things since the HT is rarely saturated and even at 800mhz it doesn't seem to affect performance...
Click to expand...

Yes, I knew this, my HTT is 250x4 right now, but for somereason the system would not post if i left the max memory clock setting in the bios at 200mhz, and when i dropped it down to 166, the system posted a bios of 208mhz 416mhz DDR, this still confuses the hell out of me, but the system works, does its 400mhz fsb and 1ghz htt so I cant complain, I would like to have 500mhz ddr and that is why I will be changing memory, I was able to test this with a stick of pc4000 and when I left the max fsb setting at 200 the ram posted at 250...
 
Geno-san said:
Yes, I knew this, my HTT is 250x4 right now, but for somereason the system would not post if i left the max memory clock setting in the bios at 200mhz, and when i dropped it down to 166, the system posted a bios of 208mhz 416mhz DDR, this still confuses the hell out of me, but the system works, does its 400mhz fsb and 1ghz htt so I cant complain, I would like to have 500mhz ddr and that is why I will be changing memory, I was able to test this with a stick of pc4000 and when I left the max fsb setting at 200 the ram posted at 250...
Click to expand...

The RAM divider of 166 is actually, a 5/6 divider of the CPU clock so that it ends up being a divder of 12 from your CPU clock of 2500MHz. The RAM divider is calculated as (10 * (6/5) = 12) 10 is the CPU multiplier, and the 6/5 is the simple RAM divider of 5:6 which is set when you put the RAM to 166MHz in BIOS.

So.. The RAM speed is 250MHz x 10 =2500MHz /12 = 208.33 MHz.
 
You mean even with all those networking certifications, you don't learn even once that no K8 platforms, particularly not the board you're using, use DDR2 memory? You might want to fix your signature there, buddy.

What'd you do, just randomly list some, "sweet sounding," components?!?
 
I currently have a 2800+ and Gigabyte K8NS Pro combo with HyperX PC4000 512MB x2.

Right now I can do 2.4 Ghz (9x267 HTTx3 1:1 memory) @ 1.6 v

the only thing I am missing is dual channel ddr (my Sandra scores are in the 4200mb/s range and Lavasoft shows my latency at #2 right behind #1 939).

All this is on air cooling (Zalman CuA heatsink)...

I've been reading all the OC info on the 939 3200+ and it looks to be around 2.4 -2.5 in potential..would I be gaining much if I chucked the 2800+ Oc set up?
 
First of all, after doing a lot of testing, I have found the following principle to hold true for Athlon 64s:

Increasing the Front Side Bus at a memory divider is pointless.

Basically, if you run your FSB at say, 250 and your RAM at 208, your performance would be the same if you ran at the same total MHz with your FSB & memory at 208.

That said, I have a .13micron Clawhammer running at 2.4 GHz w/Volcano HSF. It's not up all the way either, it's a very reasonable volume. It doesn't seem like this is a breakthrough, OC wise. And the graphs don't show that big a jump for having dual channel. Either way, Gentoo x86-64 just smokes. Hell, it's faster than my dual 2.8 Xeons.
 
Right, but you're assuming that the chip can hit its maximum clock rate with the multipliers available to it and the memory can keep up. For example, let's say your memory can't handle more than 208, but your CPU's multiplier is maxed out already, and you know that with a higher HTT, the chip can go higher at the given voltage; in that case, you're going to want to run the memory at a lower than 1:1 ratio, because even though the memory won't run faster, the extra 100 or 200MHz core speed on the CPU will definitely make a difference, particularly if, for example, you drop the memory ratio to 10:9, and raise the HTT enough that memory is now at 208 again--if your core accepts the increase in absolute core speed, then you've benefitted. This is primarily the case for any K8 with the exception of Athlon FX, because their multipliers are fully open in both directions, but for those of us that aren't rich enough and have a multiplier ceiling, this is an option some of us rely upon.

Luckily, my memory does 280 10-3-3 2.5 1T just fine, so I can keep my 1:1 ratio and I actually dropped my multiplier to achieve a higher memory speed, since my chip was maxing out at only 2.5. Too bad not everyone is that lucky, though (in terms of RAM). I can only do this with one stick, too--but this is because I'm running S754. With both sticks installed, I can't go more than about 248 with the same timings. I'm using PQI Turbo PC3200 (TCCD) on a LP UT nF3 (it's a dual channel 1GB kit, so just one stick still leave me half a gig, which I can deal with for now, since I do all my Photoshop on a different rig and I haven't played SWG for months now--that game lagged to hell with, "just," 512MB of RAM).

-Ed
 
The other benefit of the divider (other than processor headroom like explained in the above post) is the timings themselves. If your RAM would perform better at 200 or close to it (say 200-230) with tight timings then it's prefferable to use the divider than to run it at 260+ with totally ass timings and limit your processor OC headroom in the process.

It all depends on the particular, how high you're OC'ing, and the multipliers available to you (if it's a 3000 for instance your highest will be 9 so you'll need a fairly high multiplier to reach 2.3+ OCs).
 
You mean even with all those networking certifications, you don't learn even once that no K8 platforms, particularly not the board you're using, use DDR2 memory? You might want to fix your signature there, buddy.

What'd you do, just randomly list some, "sweet sounding," components?!?
Click to expand...
All certs do is get you a job (if you don't have a bachelors degree) or a possible raise (if your company cares about a particular cert) but it doesn't mean you know jack about anything, and neither does a degree.

I have A+ and Network+ certs, but who cares?
 
Astral Abyss said:
All certs do is get you a job (if you don't have a bachelors degree) or a possible raise (if your company cares about a particular cert) but it doesn't mean you know jack about anything, and neither does a degree.

I have A+ and Network+ certs, but who cares?
Click to expand...
That was my whole point. ;)
 
You know, I've started to wonder how high Kyle and co. went with their voltage on the "quasi-review" they did of the 3000, and if the K8N Neo2 has any Winchester core voltage issues... I haven't kept up with the quirks of the board as I decided to buy an A8V insteac but I know the A8N in particular will not adjust the vcore of the Winchester cores at all until you reach 1.6v, anything under that just provides stock or near-stock voltage.

A BIOS update (still in beta) is slated to fix this, I dunno if it's an issue with the Neo2 or other boards in general... It might be though, the A8V was one of the first boards with 90nm part support and they're still getting around to correcting all voltage issues, just a guess.
 
Right, but you're assuming that the chip can hit its maximum clock rate with the multipliers available to it and the memory can keep up.
Click to expand...

Yes, you're right. However I don't think the multiplier thing is an issue. Good RAM is always critical for good Overclocking. Although if you're on a budget and bringing over your PC2700 from your nForce 2 setup this may not be an option (which was my situation originally).

The other benefit of the divider (other than processor headroom like explained in the above post) is the timings themselves. If your RAM would perform better at 200 or close to it (say 200-230) with tight timings then it's prefferable to use the divider than to run it at 260+ with totally ass timings and limit your processor OC headroom in the process.
Click to expand...

I thought so originally too, but after alot benchmarking, I found that often the additional MHz afforded by looser timing was giving me more performance than a lower speed and tighter timings. Honestly, I believe that 2-2-2 being fastest (etc) is a myth that doesn't hold up to empircal data. But as always, it pays to do your own testing.
 
The DarkBlade said:
I thought so originally too, but after alot benchmarking, I found that often the additional MHz afforded by looser timing was giving me more performance than a lower speed and tighter timings. Honestly, I believe that 2-2-2 being fastest (etc) is a myth that doesn't hold up to empircal data. But as always, it pays to do your own testing.
Click to expand...

The A64 in general is less susceptible to performance drops due to looser timings (compared to a P4 or even an Athlon XP), but it all depends on the exact ratios that you're comparing... Anand's article on A64 memory a while ago provides a nice watermark on how loose or how fast you can/need to go in order to get better performance.

http://www.anandtech.com/memory/showdoc.aspx?i=2226&p=9

Easier than testing every speed step under the sun with all possible timings anyway... :p

Note: The setting with the biggest impact by far is the 1T/2T command, if you have to run 2T due to mem speed then just about any other option that enables you to run 1T will perform faster. The timing with the second biggest impact tends to be tRCD. Much more so than CAS or tRAS.
 
CZ100B said:
The RAM divider of 166 is actually, a 5/6 divider of the CPU clock so that it ends up being a divder of 12 from your CPU clock of 2500MHz. The RAM divider is calculated as (10 * (6/5) = 12) 10 is the CPU multiplier, and the 6/5 is the simple RAM divider of 5:6 which is set when you put the RAM to 166MHz in BIOS.

So.. The RAM speed is 250MHz x 10 =2500MHz /12 = 208.33 MHz.
Click to expand...

Ah, thanks alot dude, I looked and looked and never got a anser as clear and to the point as that one.
 
That's interesting, I wasn't aware of that, but then again my motherboard doesn't support Command Rate tweaking. Something to keep in mind in the future I suppose.
 
You must log in or register to reply here.
Back
Top