Ballistix PC-3200 overclocking

[imported_bluedart]

I am trying to get my system stable with a descent overclock here are my specs:

MSI K8N NEO 2

AMD 64 3000+ 90nm (winchester)
2x512 Crucial Ballistix
UATA 160 GB 8MB buffer Western Digital 7200 RPM
Enermax 420W noisetaker power supply

First, just a note, I started out with a 2x74 GB raptor, and ran into data corruption problems. Then, I switched to the Ultra ATA with more success.

I am currently able to run at 2.5 GHZ (280 FSB) with 3-5% over Vcore asynchronously. The Ballistix are holding me back though, when attempting to overclock for max bandwidt I run PRIME95 the system will fail when running at 250 FSB @ 2.8 Vdimm with latencies at 2.5-4-4-10 even 3-4-4-10!) I also cannot overclock for tight timings either. Most reviews I have seen are able to achieve DDR433 at 2-2-2-5, I cannot even achieve 2-2-2-10 at 2.8 Vdimm @ 205 FSB. Prime95 and Rise of Nations fail in minimal time. For some reason Rise of Nations will fail when my memory is overclocked and being stressed beyond it's capacity.

I beleive these modules are the same as the PC-4000 ballistix modules, but rated at DDR400 2-2-2-6 on Crucials website. I think the modules that AnandTech recieved were hand-picked because they were able to get to 278 FSB @ 2.85 Vdimm with 3-4-3-10 settings.

I was wondering if anyone else has had difficulty in overclocking Ballistix PC3200 ram? Did I just get a bad batch? How do I RMA these modules and ask for better ones? I have scheduled a tour at Microns testing facilities here in Utah, and will ask their head engineer the question while attempting to obtain a job there .

If anyone has some suggestions, please feel free to comment.

Thanks,

BLUEDART

- Has anyone ever tried a diamond heat spreader/heat sink? Much better than copper....

 

[Rigs]

Looks like your ballistix overcklocks as poorley as mine,
I can't get over 207 at 2-2-2-10.
Just out of interest, what is the manufacturing date on yours?

 

[Elfear]

I'm having the same issues with my Abit AV8. It won't clock much over 206MHz at tight timings and won't go over about 225 or so at any timings. I just thought it was incompatibility with my board but it looks like others are having the same problems.

 

[Duvie]

It may be board incompatabilities...

Dont trust prime95 to diagnos ram issues...use memtest86 and boot off of the disk and run it for awhile...

 

[jettech]

I am in the same boat with crucial but only one piece of my memory is the problem. That is one stick will run @ddr 440 2 2 2 and the other will not post at @420 2 2 2. I've had my timmings all over the place and the crucial is always the road block but i get much tighter timings at higher clocks with one over the other.

My poor performing stick:
batch CL11145.G4
p/n 47072

As i'm typing this i'm running prime95 so i'll post the batch for the good stick later if anyone is interested.
currently running
2520mhz
1.38 vcore
52c cpu@100% w/stock cooling
htt 280
fsb multi = 8x
mem divid 3/5 + ht 3x = ddr 460 2.5 3 3
2.8v dimm
with just the one good stick of ballistix.

I've had the processor as high as 2.65ghz and both sticks of memory running stable but i'm trying to find a combo that will give me a fsb of 1000, ddr 450 @2 2 2, and cpu > 2.5ghz.

A little background on this memory follows for those interested.
Before i bought this memory i contacted crucial to see if they offered "dual channel kits" in ballistix and they told me they didn't need to because thier memory yada yada yada. So I bought me some ballistix basically because of the anandtech memory test showing ballistix running ddr 466 at 2 2 2 and my chat session w/crucials tech person. After I discovered the different characteristics between the two sticks I contacted crucial again to ask why one piece is so different from the other and their response was "it's not different. Both sticks will run as advertised, ddr 200 at 2.2.2". They refused to even discuss the issue saying they do support overclocking. Evidently i'm pretty naive but the issue for me is not how well they overclock but rather that they perform diferently. It sounds like a quality control problem to me. Go figure.

This has been my first real attempt to oc anyting that doesn't roll on wheels and what a journey. I want to say thanks to the people that post here. I've read a lot of very good information here.

 

[Elfear]

I wonder if Anandtech's sticks really were hand-picked then. Either that or they got extremely lucky getting two sticks that performed so well together.

 

[cronic]

i have two of these sticks and they prefrom great. i have them runnin 2530 with fsb of 1150. thats 11 x 230. they are set to 1t, 2.5,2,3,5, timings. i am getting around 6800 in memory bandwidth tests. i am going to try another beta bios and see if i can push them even harder.

 

[jettech]

oops! correction to above post.

fsb mulit = 9x (280*9=2520 duh!)
mem divid 4/5 (166 in bios) ( i'm not smart enough to figure out why the math doesn't work but cpuz is reporting my memory at 229.x mhz and i assume it's correct)

also, I checked the other stick and it's from the same batch so no joy there.

if i figure this deal out i'll post a posible solution.

 

[imported_bluedart]

Rigs, here are the numbers from my Ballistix:
How do you tell the date of production?
#1
CL11145.G3
47072

#2
CL11145.G3
47072

Thanks for all the response to this topic... so I guess nobody has had any luck with Ballistix like Anandtech has had...

Nobody is getting even close to 250-280 FSB... That makes me feel really confident of the quality of Ballistix Oh well I guess I will have to get some better memory based on Samsung's TCCD chips to get a descent overclock eh?

 

[Rhoel]

I only have one stick of ballistix in my system, but I'm really happy with it. I'm at 250@ 2.5-2-2-8 1t. After much tweaking, I think its the cpu holding me back, not the board or the ballistix.

cpu-z

 

[BonzaiDuck]

Just throwing in two-cents worth.

I see people all over the place getting memory rated at DDR400 and trying to push it higher. Some-a-time, they do, and some-a-time -- they don't.

Many are obsessed with low-latency memory, which has its benefits at the stock (800 FSB) settings. But there have been too many reviews here, at THG and elsewhere confirming that looser timings and higher speeds yield a net improvement in performance over lower speeds and tighter timings.

I don't want to get into a pizzing contest, but it just seems to me that if you want to push the FSB higher, perhaps running a 1:1 divider, you'd be better off spending a little extra for matched dual-channel modules that are "certified" reliable at higher speeds, like DDR433, DDR500, DDR533, etc.

For instance, Mushkin DDR433's will run at 2-2-2-5 at speeds above DDR400.

It could just be a problem with these Ballistix. Here's what: try running them at their stock timings and stock speed; torture-test with MEMTEST86 for the good part of a day. With a little luck, they will fail and you can get a replacement under warranty. But you KNOW they ain't gonna give you warranty replacement if the failure is a matter of trying to run them above spec, and you get into murky waters even with looser timings and higher-than-rated memory speeds. Remember, they warrantied them to run at DDR 400 or whatever with a set of stock latencies. That's the extent of their warranty commitment.

 

[lauy]

hi everyone

I have now come across more people who have problems with ballistix modules! So its not just ME! But difference with me is that I actually bought a matched pair of PC4000 sticks and I still can't even get a good overclock. I can't even get 250mhz at 2.5-3-3-8 or even as loose as 3-4-4-10. The max i've been able to run at stably is around 230mhz at 2.5-2-2-5, if I go higher then it starts to freeze in games and fails memtest.

I've had this problem for a while now and nothing will let me push it further. I was thinking that it may be my motherboard (Asus A8V deluxe rev2) with a max vdimm of 2.8v.....but surely it should be able to run at the rated speeds anyway.

Well, all I have to say is that I will never be buying crucial products again since I will have to now go buy some other sticks if I want to overclock further.

 

[BonzaiDuck]

I got OCZ EL "Gold" DDR500's (not even revision 2's) with a latency spec of 2.5 ,4 ,4, 7. I can run them at DDR480 and timings 2.5, 4, 4, 6 at VDIMM 2.75V -- a tenth- or twentieth-Volt lower than their "support" rating.

I was looking with some interest at these Crucial Ballistix DDR's rated between DDR500 and DDR550, but I'll stand pat with the OCZ's.

I'm sorry I'm not as adventurous as some of you folks. Maybe you DID search the reviews before buying, or maybe those reviews DID hype the Ballistix, or maybe you DID get a faulty module. But I tend to put off my purchasing decisions until after I find reviews, customer-forum posts, and these over-clocker forums with some consistently good news about products.

I'm a parasite -- watching your mistakes. :disgust:

 

[jettech]

I think BonzaiDuck makes some very good points about what to expect from memory; if it says it will do ddr400 at 2 2 2 and it does that, there really isn't much to complain about. Also, I know at least I would be better off if I had just bought ddr 500. However, whatever the memory will or will not do, if you buy two peices, even if they are not matched pairs and especialy after the manufacturer says they don't need to be matched, they should perform equally at all levels. I'll get off the soap box now.

bluedart,

i've managed to get this rig stable with both sticks and here are the specs, maybe the'll work for you:

a64 3000+ cpu running @ 2448
Crucial Ballistix 2x512 @226.3 2.5, 3, 3, 8 by cpuz
MSI K8N NEO2 Platinum Settings
max mem = 166 (cas =2.5, ras =3, Min ras =8, Trp =3)
HP mode = manual
Aggressive timing = disabled
Dynamic oc = disabled
Spread spectrum = disabled
HT freq = 4x
Cool & quiet = disabled
CPU ratio = x9
HTT = 272
AGP freq = 67
CPU VID = 1.450
CPU Volt = by vid (works out to 1.42v as reported by cpuz)
Memory voltage = 2.85 (i was at htt 270 @ 2.80 and needed 2.85 to get 272; ballistix is rated for 2.8 so 2.85 shouldn't hurt it to bad)
AGP volt = 1.6

Hope you find this info useful.

Again thanks to all who post here.

 

[Zebo]

Originally posted by: Elfear
I wonder if Anandtech's sticks really were hand-picked then. Either that or they got extremely lucky getting two sticks that performed so well together.

I can clock exacly the same as anand even a little better....

 

[Zebo]

I just pulled out some old mem tests notes I did with Ballistix all memtest86 stable...(i let you know when I get my MSI board if they stay same)

1T

2-2-2 I get up to 225 Mhz

2.5-2-2 I got 247 Mhz

2T

3-4-4 305Mhz

So I guess I got good luck on my mem...bad luck on my chip... win some loose some..

 

[Slug]

Check this thread:

Thread

 

[imported_bluedart]

FROM ZEBO: I can clock exacly the same as anand even a little better....

If you can clock BALLISTIX as good as or better than anand's, why don't you run your setup 1:1 at over 2.5 GHZ it would probably be faster than running it like you are now w/ 5:4 ratio.

I wish I had your stix... Mine are very poor... now I am running at 250 FSB mem 200:166 with ram @ 208 2-2-2-10. Still not completely stable.

I wish I would've just payed a little extra and got the OCZ 3200 EL PLAT REV 2 modules.

Oh well... Anyone want to trade?

 

[Zebo]

No it's not. 1:1 makes 2t and crap timmings which is way slower Than LL a async I'm running.

 

[imported_bluedart]

Okay... I was just wondering how far you could go at 1T. I figured you could even surpass Anandtech's top 1T FSB of 278... But I guess not.

Thanks...

 

[imported_bluedart]

Alright guys.... Here's my last attempt at overclocking with ballistix and my 3000+... I have a problem though...

My system halts at the following screen while booting:

NVIDIA RAID IDE ROM BIOS 4.60

Detecting array ...

This is while attempting to run:

CPU: 2.2 GHZ @ 247 FSB
RAM: 166 ratio (asynchronous) @205 2-2-2-10 1T 2.8V

Has anyone else had a problem locking up at this screen?

I also did some calculations in excel and found this configuration to be interesting:

8x multiplier = 2.5 ghz @ 312 FSB on the 3000+ CPU while leaving the Memory at around 200 by changing the Memory ratio to 133:200.

What do you guys think about this configuration?

 

[CrucialLabs]

Interesting at best....I guess a few people fail to realize there is more then memory that would cause a system to be unstable and not run properly at higher a freq. I am not trying to start a flame war, just sighting some facts. When you are overclocking you are affecting the whole system not just the memory. I am not trying to offer an excuse for what you all are going through just trying to offer some solutions you might be able to use to get what you are looking to do. Yes we guarantee our memory to hit standard timings. You will need to loosen up your timings to hit higher freq. To stay will stock timings and ask any memory to hit phenomenal speeds can not be done.

BL6464Z402 PC3200 Ballistix memory timings: CL-2-2-2-6 Voltage 2.8v

These timings are not written in stone but they will have to be adjusted to allow for higher speeds. To find out what works for you may take some time and effort. Each motherboard, memory, and processor can have slight differences which can affect the result. The higher the speed the looser the timings need to be.

BL6464Z505 PC4000 Ballistix memory timings: CL-2.5-4-4-8 Voltage 2.8v

If you are looking to get speeds to and over 250MHz and stick with a little tighter timing I would recommend using the PC4000 modules.

I would also recommend disabling and performance settings in the BIOS. Performance settings like Turbo or fast just cause more problems then they are worth when you are trying to overclock the system yourself. So I would start off with the stock timings and make sure the system is stable and there are no issues then slowly move the FSB up til you start having instabilities then begin changing the timings one increment at a time. Make sure you write each change and effect down before changing something else. If you don't then you could be taking 2 steps backwards.

Just trying to help.