box is together! -- ram performance, etc questions

The forum for non-component-related silent pc discussions.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

box is together! -- ram performance, etc questions

Post by Nate » Tue Apr 27, 2004 2:52 am

In this post you guys helped me understand a lot and to pick out the components for my new box. Well, the components are here (except for the case - the components are plugged together without a case!) and I have a few questions about how the ram is performing.

I ended up going with the OCZ Enhanced Latency Series Platinum Edition (2-3-2-5). What settings do the 2-3-2 numbers correspond to? Here are my results after intial testing...

Well first, the numbers listed here will be in the format of: row-active delay, ras-to-cas delay, row-precharge delay, cas latency time. This is the order the settings are in my NF7's bios. Also these tests were done with the multiplier on 9.5 (so the cpu isn't being pushed), DDR at 2.9v (does it need to be so high?), chipset at 1.6v (what is chipset voltage? what should it be set at?), and cpu at 1.8v.

First off I tried 5-2-3-2 with 200 fsb. This would not post. Should this work considering the ram claims "2-3-2-5"?

After some thought, I tried 5-3-2-2 with 200 fsb. This posted and passed memtest. Is this what the ram is rated at when it says "2-3-2-5"?

Now for some (further) overclocking! I tried 5-3-2-2 with 205 fsb. This would not post. Doh.

I relaxed the timings a bit and tried 11-3-3-2.5 with 210 fsb. This posted and passed memtest. I read somewhere that a row-active delay of 11 was the best setting for the NF7, is this true? At any rate, I don't think it has much affect on whether the setting will work.

Finally, I tried 11-3-3-2.5 at 215 fsb. This would not post.

So, do you think this is typical of the ram I have? Should I RMA the ram? I thought I'd be able to get a higher fsb than 215, this ram is supposed to be great stuff. Am I doing anything wrong? Should I be running the ram at 2.9v?

Which of these settings is optimal? It comes down to higher fsb and loose timings or lower fsb and aggressive timings.

I want to figure out how far I can push the ram, then I'll work on how far I can push the cpu, then I plan on using 8rdavcore to dynamically overclock as needed. This way I get a quiet HTPC in the living room that turns into a killer gaming machine on demand. :)

I'm currently typing at 2400mhz (12x200) and 5-3-2-2 timings. The CPU temp as reported by MBM is 41c (is this accurate for an NF7?) at idle, I haven't stress tested it with Prime95 to see if we get in the scary zone yet. The cooler is an SP-97 with a 92mm Acoustifan, passive sink on the northbridge. Does 41c seem kind of high for this overclock and 1.8v cpu at idle? The northbridge doesn't seem too hot, the ram got pretty dang hot earlier (but not too hot to touch). The southbridge gets damned hot however, I think I'll get a passive sink for it.

I'll stress test right after this post.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Tue Apr 27, 2004 3:15 am

The 3 corresponds to ras to cas delay , since your memory uses CH-5 chips , it 99.9% can't work with ras to cas = 2.

Chipset voltage is the voltage fed to the northbridge.
Increasing it helps reach higher fsb , but overheats the nb.

Row active of 11 is almost always better for nforce2 motherboards.

Since you are trying for max fsb , I'd recommend pumping memory and chipset voltages to their maximum (1.7 & 2.9).
Then , with 2.5-4-4-11 (or worse) timings find your motherboard's maximum fsb.
After that , lower the timings to find out the limits of the ram.

What are you using to cool the northbridge?
If it is a zalman , you might want (just for testing) to try pointing a fan at it , to make sure that your problem isn't bad cooling of the nb.

Generally , I prefer tight timings over high fsb.
In my case , the options were 223 @ 2-2-2-11 and 239 @ 2.5-3-3-11 , the first performed better in everything.

I'd suggest using some benchmarks to find out what is best.
My list (for memory) , consists of superpi , hexus pifast , sandra memory unbuffered , 3DMark01 and WinRAR.

EDIT: a few more things

Disable "fsb spread spectrum" and "agp spread spectrum".
Keep "cpu interface" enabled at all times.
Flash your bios to the newest version , I recommend looking for tictac modded bios (based on version 22).
Try using different dimm slots , some may work better than others.

Also , try running your memory at it's default specs (timings , speed and voltage) , with chipset voltage @ 1.7 , to determine if it's faulty.

Finally , I see you have a mobile cpu.
Does that have a 133 default fsb?
If it does , you might need the L12 mod to get very high fsb (don't worry , it has to be the easiest mod ever).

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Tue Apr 27, 2004 11:48 am

Last night after I posted I ran Prime95 at 2400mhz and it rebooted. 2300 had calculation errors. 2200 rebooted. At that point I went to sleep. :(

Thanks for the reply apocalypse80!

Ok, since the 3 in the OCZ ram's rating is ras-to-cas, the ram does work correctly at its rated timings.

The nb is covered by a passive Zalman. It doesn't get very hot to the touch. The sink is a bit of a strange shape, I don't have a fan that would fit it. I can point a house fan at the components though.

What bios do I want to flash? A quick search turned up tons and tons! Can you link me to a good one? :oops: I have the regular NF7 (not -S).

Yep, I do have a 2600+ XP mobile cpu. I believe it does have a fsb of 133 by default. Does this mean I need to do a mod to get a higher fsb? Whats the logic here? It'd be great if this was the problem!

My plan for tonight is to use your suggestions... I'll start by blasting the components with a house fan (they aren't in a case). I'll flash my bios, up the chipset and DDR voltages to max, disable "fsb spread spectrum" and "agp spread spectrum". Then I'll relax the timings to 2.5-4-4-11 and see how high I can go. Then I'll tighten up the timings and see where it is stable. I'll get some benchmark programs and see which performs better.

I'm jealous of your 220+ fsb apocalypse80! :evil: :P :?

EDIT:
I found this guide an interesting read. Scared me when he started talking about RMAing the border if it can't hit a high fsb. But there are some things I need to do before I come to that conclusion. Can't wait to get off work and go play! :D

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Tue Apr 27, 2004 1:16 pm

I think that the zalman on the nb should get very hot , but then again you aren't using a high chipset voltage so I'm not sure.
If everything else fails , remove it and reassemble (you might have bad contact).

For modded bios , look here http://s6.invisionfree.com/underclock/i ... howtopic=9
I'm using the D22 alpha 1 with S-ATA bios 4247 , with great success.
I'd recomend using the bootable floppy method to flash your bios.

Here are the pinmods
Image

The "L12 mod" , is the "FSB 166" one (forget about the rest).
You take a very thin wire , around 3-4mm long , bend it to a "C" shape and insert in the socket , shortcircuiting the pins shown in the pic.
Insert the cpu , boot the rig , go into the bios and "load optimised defaults".
Congratulations , you cpu is now detected as 166fsb.

The NF7-S overclocks much better with 166fsb cpus :wink:

For me , modded bios + L12 mod = +19fsb (from 220 to 239).
It didn't prove too useful (I'm only using 220) , but it was fun to do anyway.

Anyway , overclocking is the endless process of trial and error (much like silencing) , so good luck on your OCing.
And don't forget , cpu interface = enabled , you'll take too big a performance hit by leaving it disabled.

I pretty much agree with everything on that guide except ;
I didn't need southbridge cooling (I don't have any) , so you might not need it either.
DO NOT use bios D10 if you use S-ATA drives , there is a very big chance of hdd corruption.

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Tue Apr 27, 2004 3:52 pm

I don't know, have you touched your southbridge? Mine is hotter than anything else on the mobo.

I've got Arctic Silver 3 on the nb, but its not a problem to put the stock one back for a bit.

The guide was saying some of the higher bios' aren't as good for OCing. Should I go with D10 since I have the regular NF7 (no SATA)? If no one says otherwise I'll try D10.

Thanks for the tips and pic of the pin mod! I didn't know I had to hit load optimised defaults in the bios for it to take. Need to find a tiny little wire. :)

How are your temps? What are your voltages, sink, fan, etc? I'm using an SP-97 that I lapped, Arctic Silver 3, and a 92mm Acoustifan which doesn't push a lot of air but is silent. With cpu at 1.8v and 2400mhz (currently unstable) the temps were 41c idle. Starting Prime95, temps went up to about 49c before the computer rebooted (not from overheating). Is this normal for my setup? I think I would be happy with a max temp of 50-55c or lower under full load (probably only happens when playing games). I'll use 8rdavcore to run a cooler and quieter (temp controlled fans) speed at idle.

I'm not sure if I put the AS3 on too thin. As soon as I finished lapping the sink I cleaned up with alcohol and applied AS3, then used the edge of mu driver's license to take off all but a very very thin layer. Same with the cpu. The layer was extremely thin. Wonder if I should put it on thicker when I reinstall after the pin mod.

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Wed Apr 28, 2004 3:20 am

Well, been playing around without much luck. I flashed to D10, disabled APIC (not in the bios, but as described by Mickey Padge here), performed the L12 pin mod (and verified the mobo saw my 2600+ mobile as a 2500+ @ 11x166), disabled everything I could under "Advanced Chipset Features" in the bios (fsb/agp spectrum, etc), sat a passive heatsink on the southbridge, and set up a house fan to blast the components. Then I tried to push the fsb as high as possible with loose timings, using memtest #0, #1, and #5 (briefly) every step of the way to be sure I'm stable. I then worked my way to more aggressive timings, adjusting the fsb until I found what was stable in memtest. Here is the relevant information on what I have so far...

I'll put the timings in the format: cas latency time, ras-to-cas delay, row-precharge delay, row-active delay. Is this the standard format everyone uses? Relevant settings were 7x multiplier, 2.9v ram, 1.75v cpu, 1.7v nb.

timings = fsb speed
3-4-4-11 = passed at 220
3-3-3-11 = passed at 220, 225 failed with just a few errors
3-3-2-11 = failed at 215
2.5-4-4-11 = failed at 215
2.5-4-3-11 = failed at 215
2.5-3-4-11 = failed at 215
2.5-3-3-11 = passed at 210, failed at 215

I tried...

200x12 (2400), 2-3-3-11
Passed, memory bandwidth 2945/2791.

200x12 (2400), 2-3-2-11
Passed test #5, but consistently failed #1 with one error.

200x12 (2400), 2-3-2-5
Passed, memory bandwidth 2976/2831. This is prime95 stable with temps topping out at 53c during the torture test.

210x11.5 (2415), 2.5-3-3-11
I did see it throw errors in memtest once, but I came back to these settings again later and couldn't get it to give any errors. Memory bandwidth 3109/2918. Got a reboot during prime95 on the 3rd torture test.

All these numbers were with the passive Zalman heatsink on the nb. The passive sink was barely warm to the touch and the backside of the nb was hot but wouldn't burn my finger so I believe it was cooling sufficiently. I replaced the stock nb cooler and at 210x11.5, 2.5-3-3-11 the back of the nb feels as hot as with the Zalman. With those settings and the Zalman, prime95 rebooted during the 3rd torture test. Same result with the stock cooler. Could this reboot be caused by the cpu voltage being too low? I'll moved it up to 1.8v and will see how it does.

This ram is rated 2-3-2-5, it is interesting that it fails at 2-3-2-11. Do you think it would be worth it to retry the above timings with a row-active delay of 5?

Do the memory bandwidth numbers look correct?

Why am I unable to get a higher fsb? 210 with decent timings seems pretty lame. Is it the motherboard or the ram? Should I RMA the motherboard? RMAing the ram would be expensive (15% of $257 = $38 ). The ram does do its rated 2-3-2-5 at 200mhz, but I think it is what is stopping me from passing memtest at a higher fsb. :( I thought I was buying super good ram!??

The only thing I can think that I haven't done is to try a different bios. I'll give D20 a shot tomorrow, but since the ram isn't passing memtest doesn't that mean that the ram is what is failing? Can changing the bios affect that? Is there anything else I can do or try?
Last edited by Nate on Wed Apr 28, 2004 3:27 am, edited 1 time in total.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Wed Apr 28, 2004 3:27 am

The modded tictac bios has all the benefits of bios D10 (working 10x and 10.5x , no 222fsb wall) , plus all the corrections that abit did since the original bios.
There is no reason to prefer D10 over D22 alpha 1.
I haven't tried any other tictac bioses , so they might be better , but D22 alpha 1 was enough for me.

Yeah , my sb gets extremely hot (the MCP-T is hotter than the MCP) , but it doesn't bother it.
The hottest thing on your motherboard should be the Vcore mosfets , but again they don't care until you go over 1.9Vcore.

The error that most people do when mounting coolers on the nforce2 nb , is that they put a very thin layer of AS.
The northbridge isn't flat , so it needs more.
My method is , put some AS on the chip , then mount the HS and remove to check the imprint , if it isn't perfect , add more AS.
For the cpu , usually , thinner is better , but do check that you get a good imprint.

My temps are pretty good , but I use water.
I get low to mid 40s at load , mid to high 30s at idle.
I also have a second waterblock on the nb.

Abit's temperature sensor isn't very accurate , since it uses a socket thermistor.
I guess your temperatures are normal.

I use nf7vcore and clockgen (for testing) , they work nice , but it's too much of a hassle to use them all the time.
To get lower idle temperatures , enable the "cpu disconnect function" in the bios (it's only available on later bioses).

EDIT

The bottleneck is the ram , the motherboard can do 220.
However , 210 is indeed too low.

Can the ram do 200mhz , 5-3-2-2 , with only 2.6V?
If it doesn't , then it's faulty.

Trying again with 5 makes sense , since it makes a difference for your ram.
Another thing to do , start with minimum memory voltage and begin raising the fsb with tight timings , when it becomes unstable , up the memory voltage.
There have been some rare cases , where ram didn't like high voltage.

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Wed Apr 28, 2004 3:36 am

We must've posted at the same time. :) I sooo need to pass out, I'll have to try the 2-3-2-5 at 2.6v tomorrow.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Wed Apr 28, 2004 3:43 am

Welcome to the wonderfully frustrating world of overclocking.

Sh*t , almost forgot , make sure you set "AGP frequency" at 66mhz , NOT auto.

silvervarg
Posts: 1283
Joined: Wed Sep 03, 2003 1:35 am
Location: Sweden, Linkoping

Post by silvervarg » Wed Apr 28, 2004 7:01 am

I would also check with another CPU load program to make sure the crashes doesn't come from CPU overheating. E.g. CPUburn.
You could also run the test that doesn't use RAM in prime95.

If you still get a reboot then it is likely that you need a thicker TIM layer on your CPU. If it passes the test it should be your RAM that is the problem.

trodas
Posts: 509
Joined: Sun Dec 14, 2003 6:21 am
Location: Czech republic
Contact:

Post by trodas » Wed Apr 28, 2004 10:15 am

Hmmm, after reading how badly the "super mega hyper" OCZ ram performs it making me feel good I go with the Mushkins hi-performance and can run 2-2-2-11 with 200MHz w/o a glitch at 2.77V (at 2.5V it never worked, random address errors, memtest 3.1a test 1, and since Mushkin on website states that the ram should run at the clock at 2.5-2.8V, it's okay) :wink:
Using Epox 8RDA+ mobo, very happy with it.

About the wire mod. Too late, but I have suggestion - some hi-quality audio cables have cooper shielding. The cooper wires into this shield are thin enought to do not stress the ZIF socket too much (remember, the socket are designed ONLY to take the CPU pins, nothign more, so when you stuck there something with comparable size as the CPU pin is, you might asking for troubles :roll: ).
What do you think, apocalypse80, good or bad idea? :twisted:

Mine maximal temperature are 56 degrees (Vcore 2.2V, 200x11, Prime95), but I use watercooling and for me, 200x10.5 seems to be the limit of my CPU (old unlocked Barton (not mobile yet!) 2500+) with current ArcticSilver usage. My friend, who have borrowed the mobo and CPU before managed working 200x11 (then 3200+ rating) when - and only when - he do a exceptional work on the extremly thin layer of AS on the core - remember the instructions, Nate. The heatsink should be "painted" with the AS, but all AS should be then cleaned prior installing the HS on the CPU :wink:
No layer on HS!
It might as well limit your clock! (as it doing on mine)

NB cooling. I can't understand how the Zalman passive cooler can be so cool, when mine hit 58 degrees (measured with this http://deltatrak.com/infrared_thermo_8.shtml ) on it's center after power-on and 10 minutes of running memtest. Okay, my timings are tight AND I use two 512MB modules as doublechannel, but I suppose that everyone do the same, right? :wink: :P
Quite frankly these temps scared me enought to NOT try anything over 200Mhz, and I preparing FIRST before i try 201Mhz something that COOL the NB down a lot :twisted:
It will feature this cooper plate: http://ax2.wz.cz/show.php?p=badthingsha ... 7&d=1&v=v2 and I already laped my NB: http://ax2.wz.cz/show.php?p=wc&id=52&c=7&d=1&v=v2 and bought 8x8cm big heatsink: http://sales.goldmine-elec.com/prodinfo.asp?prodid=3556 and then, I will beat you all :lol:

Well, nope, because my mobo is revision 1.1 and therefore can't adjust the chipset voltage and therefore my highest FSB will probably be notably lover - frankly, i go as high, as the Mushkins manage the 2-2-2-11 and not a Mhz higher :wink:
...and Im still not ready to run them at 2.9V yet, because the temperaturemeter shows that after the mentioned 10min of testing, they hit 41 degrees. God know, what temp they have now, closed case and more that 10 hours of running :?

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Wed Apr 28, 2004 12:30 pm

I put slightly more of AS3 on the cpu after I did the pin mod. I am pretty sure I did it right, and temps are the same as before I did the pin mod. The pin mod went pretty smoothly, the wire I used was tiny.

AGP frequency is 66mhz.

silvervarg, I'll give CPUburn a shot.

I haven't done any measurements, but I still think the Zalman works great. I have some little "temp nurses" that I can use to monitor its temps.

Well, I tried 2-3-2-5 at 2.6v (what the ram is rated at) and guess what? It wouldn't pass memtest! Going to look into getting a replacement without paying $40. In the meantime I'll try a new bios.

alglove
Posts: 363
Joined: Fri Feb 06, 2004 11:21 am
Location: Houston, TX, USA

Post by alglove » Wed Apr 28, 2004 1:49 pm

Nate,

I read a review about this memory a few days ago. Basically, it said that it runs very well at its rated speed and settings, but that it does not like to have its timings changes and is a poor overclocker. Since you actually have the memory yourself, it will probably make more sense to you.

http://www.xbitlabs.com/articles/memory ... 200eb.html

Nate
Posts: 37
Joined: Wed Apr 07, 2004 4:07 pm

Post by Nate » Wed Apr 28, 2004 3:40 pm

Thanks for the link alglove! The memory they reviewed was 2.5-3-2-8 and what I have is 2-3-2-5, but it was still an interesting read. The OCZ memory actually beat the Corsair 2-2-2-5 in 3dMark03. Not by much though, and in the other benchmarks the OCZ was slightly behind. When they talk about its inability to be overclocked, I think they mean run at cas 2 since they said they got it up to 220mhz (and I don't think any CH5 memory can run with ras-to-cas 2). The memory I have is already rated at cas 2 and 2-3-2, the problem is it just can't handle a higher fsb. I would be happy with a 220 fsb.

I decided to go ahead and RMA the ram, newegg should get it by noon on Friday and I should have the replacements early next week. :? Pray for quality replacements! :!: I wish I had a chance to try a new bios but I wanted to get the replacements ASAP. The ram should be able to run at stock timings and voltage with any bios!

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Thu Apr 29, 2004 6:36 am

@ trodas
Any very thin wire will do , as long as it's not insulated.
However , to my knowledge , the L12 mod helps only on Abit NF7 & NF7-S , so I see no reason for you to do it.

Nate's zalman should be cooler than yours , he has much better case ventilation + a fan right next to it (the cpu fan).

@ Nate
The RAM tested by xbitlabs is different than yours so you shouldn't compare it in any way.
I don't know what chips it uses , but most likely , they aren't CH-5.
Also they're made for mhz rather than timings = not for AthlonXP systems.

bomba
Posts: 320
Joined: Fri Jan 16, 2004 12:36 pm
Location: USA

Post by bomba » Thu Apr 29, 2004 7:47 am

ch5 chips supposedly have internal voltage regulators that will limit/negate the benefits of increased voltage. Seach for CH5 AND voltage regulator and you'll find more info.

I built a NForce2 Barton 2500+ system with 2x 256 Buffalo CH5 sticks. This system was completely stable at 333MHz FSB (PC2700) 2-2-2-5 settings. I also built an Intel 865PE P4 2.4C system with the same RAM, most aggressive stable settings were 800MHz FSB (memory @ 400MHz PC3200) 2-3-2-5. If your OCZ sticks are CH5, they should be capable of these settings at stock voltage, if not RMA!

trodas
Posts: 509
Joined: Sun Dec 14, 2003 6:21 am
Location: Czech republic
Contact:

Post by trodas » Thu Apr 29, 2004 10:24 am

apocalypse80 - I know that I don't need to do it, especially my Barton 2500+ is recongized properly as 166x11 one, but just a tip for others :wink:

bomba - what is the deal having 166/333Mhz system at 2-2-2-11 (11 is faster that 5, and stress the mem more then :wink: ) ... :?:
I don't get it.
We talking about 200Mhz (400MB when the DDR thing come to play) and higher clocks :wink:

Image

Image

bomba
Posts: 320
Joined: Fri Jan 16, 2004 12:36 pm
Location: USA

Post by bomba » Thu Apr 29, 2004 1:53 pm

trodas wrote:bomba - what is the deal having 166/333Mhz system at 2-2-2-11 (11 is faster that 5, and stress the mem more then :wink: ) ... :?:
I don't get it.
We talking about 200Mhz (400MB when the DDR thing come to play) and higher clocks :wink:
Sorry my post was not helpful. Though I did not OC the Barton 2500, I did run the CH5 memory at 2-3-2-5 @400MHz/stock voltage in an Intel system w/ full stability. System would not boot w/ RAS-CAS of 2 even with increased voltage.

Some forums have reported that the CH5 chips internally regulate voltage to 2.77V, i.e. CH5 chips see only 2.77v at all voltages >2.77v. Seems like what Nate needs are those hard to source BH5 (or BH6) sticks to OC w/ tight timings.

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Thu Apr 29, 2004 2:26 pm

Not all CH-5 sticks have voltage regulation , very few actually , most work just fine with obscene voltages.
Not even the ras-to-cas = 3 is a law for CH-5.
There have been reports of (very rare) CH-5 sticks working with 2-2-2-x at 200+ fsb , but only on nforce2 motherboards and not showing any improvement in performance (like they should).

Anyway , Nate's ram is faulty.

Sure BH-5/6 would be better (I love mine).
But they are extremely expensive now , for not so great gain , plus they require very high voltage to really fly.

BTW , maybe I should upload some of my pics , god knows I have some good ones...

trodas
Posts: 509
Joined: Sun Dec 14, 2003 6:21 am
Location: Czech republic
Contact:

Post by trodas » Fri Apr 30, 2004 10:06 am

Hmmm, I have BH-6 and I would like BH-5 more... :?
It's simply fact, that the BH-5 are 5nS ones, where BH-6 are 6nS ones and therefore the maximum overclock is likely to be lower... :cry:
Yes, they are extremly expensive, but also worth every penny, because the speed is hard to beat with anything on Earth 8)

Todays I got angry (one test went bad and overclocking AGP from 66 to 78Mhz did not gain speed-up as it should!) on my machine and desplite my overheating chipset, I pumped my FSB 5Mhz up. No big deal, yes, but not too. For me, 200 x 10.5 working stable and 200 x 11 not (only Prime95, tough). So, its 2100 stable and 2200 unstable. Now Im at 2152, witch are pretty close to the line :lol: Luckily, it working and folding well.
I hope Steve send me the cooler soon, so I could pump the max FSB I can get stable out of my mobo soon.

Perhaps after a week will be time to set bets how far I can go :lol:

Anyway, Im still not very comfortable with overvolting the memory too much. Im now at 2.77V and the next and last option - 2.9V - sounds a bit too much to me. I would like not get over the 2.77...
...also I did not believe much that the maximum memory clock could change too much, when I overvolt it, but apocalypse80 keep convincing me othervise... :roll:
Maybe I then sometimes try the 2.9V... :oops:

apocalypse80
Posts: 90
Joined: Mon Feb 09, 2004 4:16 pm
Location: Greece

Post by apocalypse80 » Fri Apr 30, 2004 11:08 am

From what I've seen , in most BH-5 chips , +0.1V = + 8-10mhz fsb.
BH-6 , from what I hear behave much like BH-5 , so with +0.13V you should get +10-13 mhz....
That is , if your motherboard can handle it.

Are you convinced yet? :lol:

TheOnlyOne
Posts: 16
Joined: Sat Dec 27, 2003 9:23 pm

Post by TheOnlyOne » Fri Apr 30, 2004 12:37 pm

If you are looking into getting a different mobo i would look into the DFI Ultra Infinity. It's one of the best out there for overclocking and almost everybody is hitting 220FSB+ as long as your processor and ram can take it. If you look on the net you will see some people hitting 270FSB which is awesome for an AMD XP. If i remember right the highest vcore in bios is 2 and 3.3 vdimm. Good luck.

trodas
Posts: 509
Joined: Sun Dec 14, 2003 6:21 am
Location: Czech republic
Contact:

Post by trodas » Sat May 01, 2004 4:57 am

apocalypse80 - almost :lol: The thing is, that what more limit my overclocking is the cooling of the NB right now :oops:
The damn small Zalman ZM-NB32j is with fanless case simply not enought good to cool my NB at hi-end clocks.
However I did a little thing today. Feeling endlessly depressed from the lack of the big heatsink yet (slow Steve and post and waiting and waiting... :roll: ), so I get bored and tried to play a little with my unfortunete mobo.
Barton 3200+ 200x11 2-2-2-11 DDR nForce2 ultra 400 chipset DOUBLE CHANNEL - 1511MB/sec
Barton 3200+ 210x11 2-2-2-11 DDR nForce2 ultra 400 chipset DOUBLE CHANNEL - 1572MB/sec
Barton 3000+ 215x10.5 2-2-2-11 DDR nForce2 ultra 400 chipset DOUBLE CHANNEL - 1591MB/sec
Barton 3000+ 220x10 2-2-2-11 DDR nForce2 ultra 400 chipset DOUBLE CHANNEL - 1629MB/sec
Barton 3000+ 225x10 2-2-2-11 DDR nForce2 ultra 400 chipset DOUBLE CHANNEL - 1662MB/sec
...and then the chipset it as hot, as i can't leave the finger on it for more that sec, and there is no way to convince me point my laser temperaturemeter on it - forget it, I better DON'T want to know, what temps I got then... I then tried 240Mhz, but the Epox did not liked the 3-4-4-7 timings I give to the ram, so it not booted. Yet.
More experiments ONLY with better cooling.
Wanna know why? Mostly at end of 5 or at start of the 6th test, the memory start from clocks over/including 220Mhz, showing a little errors ever at 2.9V - the voltage DID NOT MATTER in my case - the errors are the same and are IMHO related to the chipset, that getting dangerously hot and I don't want kill it. Yet :twisted:

After all, reaching ATM "only" 215Mhz with such small passive cooler on unmodded Epox 8RDA+ should perhaps be a world record? :lol:

http://ax2.wz.cz/show.php?p=badthingsha ... 7&d=1&v=v2

Post Reply