ocing doesn't boost speed at all!

just got my new hd last night, and windows is up and running and i'm ready to oc again. the settings i've tested are cpu and ram @ stock, cpu @ 225 htt with 1:1 ram @ 2.5-3-3-10-2T, and cpu @ 235 with ram set to run at ddr333 (so it ends up something like 190 mhz) at 2-3-2-10-1T. all of these seem stable.

to test these all, i've tried running the sisoft sandra 2005 memory bandwidth test, which showed little noticeable difference between the three settings.
i've run 3dmark2001se at defaults settings and there really hasn't been a drastic change between any of these settings, only a couple hundred points. i've run 3dmark05's cpu test, and scores have all been within 100 points of each other.

therefore, is it even worth ocing with such close results? ocing my vid card helps a lot, but ocing the cpu seems to do nothing. i used to be able to set cpu @ 240, with ram at 200 2-3-3-10-1T, but no more, and from looking at what i've just found, i bet it wouldn't have changed much. am i missing something? you'd think that even if the ram were a bit slower than spec, if the cpu was oced by 235 mhz that there would be some difference!

i can't run any in-game benchmarks yet as i now only have half-life and cs, which aren't exactly taxing on my rig. i will have hl2 soon, but probably only during the upcoming winter vacation.

should i be seeing improvments? is it even worth wearing out the cpu if it doesn't help at all?

also, i now run a sata hd, and i've heard that the via sata controller doesn't do well past 220 mhz. is this true? could this be affecting my oc? this is a nioce hd (a WD Raptor), and there's no way i'm changing it after having just spent $170 on it.

Ram won't make that HUGE of a difference unless you're going from something drastic like 200 @ 3-3-3 to 240 @ 2-2-2. You didn't mention how far you're OC'ing your CPU, what settings have you tried? Are you doing 235xstock multiplier? You should get a decent increase in CPU score, but not a huge difference. Keep in mind 2005 is 90% GPU.

225mhz? your complaing about only 25mhz over stock? That is'nt much.

I'm not to familar with your board, so i can't add any advice. Try running the lowest ram timings you can (2-2-2-11) at stock speeds (200mhz) and see what you get.

Please don't refer to your ram speed at HTT, its really confusing when you start running dividers. (Higher HTT speed do not increase performance what so ever)

i ran the kv8 pro up to 260 with no issues
try 240x10 that is a nice 20% oc and a big increase in speed

did i refer to my ram speed as htt? sorry, didn't mean to, i do know the difference.

my stock multiplier is 10x (which would make that 25 mhz htt boost into an extra 250 mhz for cpu speed). i have heard that with a64s you want to use the highest multi possible.

my cpu will not clock higher than 240 htt, at least that's what i've found. i've tried 240x10 with vcore up to 1.7 (board overvolts by .05 anyway, and i don't want to run 1.8 through my cpu without water cooling), with an fsb:ram divider making ram (at stock) run as ddr333, there for ending up with 2.4 ghz cpu and ddr400 ram. this couldn't even last a minute in prime95, and would freeze during any gaming environment, from actual games like Prince of Persia: Sands of Time and Half-Life and CS (original versions) to 3dmarks 2001, 03, and 05. would also casue random reboots and crasing to desktop. so, obviously, that wasn't stable.

ram will not stably run 1T higher than 205 or 206 mhz, and is rated for 2-3-3 timings at ddr400. stock voltage is 2.65, but board overvolts by .09, so it's more like 2.75. i can get away with 2-3-2, but whenever i try lowering the 2nd number down to two, no matter the circumstances, comp won't boot, and i have to manually reset the bios via jumpers on mobo.

cuddles, when you say it should give me a decent cpu score boost, you do mean more than like 60 points, right? also, cpu tests run at the absolute lowest IQ settings so that the cpu does all the work, not the vid card.

highest oc i've been able to run with the ram is 2.5-3-3-10-2T. how much do timings affect performance? i really want best perf possible in those cpu-dependant levels in hl2, hl2 deathmatch, and cs:s. how much of a difference does 1T make over 2T? remember ram won't allow 1T when it's running over 205 or 206 mhz, which, obviously, is pretty darn low.

i am aware that i need new ram, but for the moment i have spent all the money my parents will allow me to for at least 9 months, and i'll need a job to earn what i spend from now on (no more leeching my bank account, say my parents, and seeing as mum is basically in charge of all withdrawls of money...)

bump the CPU to 1.8v and see where you can go.

Run dividers and keep your ram at 1T if you can. Try feeding the ram more volts.. I dont think that board does more than 2.9v so you should be safe feeding it all the board can support.

EDIT

1.8v.. just keep an eye on temps and if they get too high for your liking drop back.

Quote:

---

Originally Posted by binkgle

just got my new hd last night, and windows is up and running and i'm ready to oc again. the settings i've tested are cpu and ram @ stock, cpu @ 225 htt with 1:1 ram @ 2.5-3-3-10-2T, and cpu @ 235 with ram set to run at ddr333 (so it ends up something like 190 mhz) at 2-3-2-10-1T. all of these seem stable.

---

in this case what u had it running before would be better since the ram is at 225 instead of 190 something. although because ur increasing the cpu by 100mhz ud prolly get a small increase in mark01

but when the ram is at 225 it's using 2T, and when it's at 194 or whatever, it's running 1T, and 1T is suppossed to make an enormous difference. i test how much of a difference tomorrow.

i have a zalman cnps7000a-Cu cooling my cpu, with arctic silver 5 thermal grease. the hsf is using its fan controller, limiting it to 1500rpm. i never saw much of a cooling perf improvement without the the fan controller, and certainly not enough to warrant to increased noise level (sys is already too loud, what with 2 120mm fans that came with the case which are spinning (i think) at their max speeds (as an aside, how could i control them? they make much more noise than any other fans, especially the intake one). i don't feel comfortable with cpu temps above 50 celcius, but i really don't know what's too much. what would you guys consider too hot?

my mobo has cpu temp reading issues (almost all abit kv8 pros do, it's a bios issue, abit hasn't corrected), but in general, with everything stock, my cpu idles around 37, and loads around 45, 47. still, this seems a bit too hot for stock speeds, even with air cooling. what do you think?

would lowering ldt multi down to 3x from 4x lower temp? it isn't suppossed to have much effect on performance at all.

EDIT: for better cooling, i *might* be able to get water cooling if i do enough babysitting/working at local indian retaurant. i am a total noob at water-cooling, so would be happiest with a recommendation for a good full kit that performs well, is quiet, is relatively cheap, and cools cpu and gpu (and preferably northbridge, too). or recommend a place where i could either build a kit and have them send me the parts (buy pre-made rig on the internet style, as one does from alienware, for instance) or a list of parts. must be easy to do!

ldt will not affect temperature.

47c seems a bit high, but my 2800+ at 2.4 with 1.75v gets about the same with 21c ambient and stock cooler on the bench.

1T adds about 25MHz of performance. That happens to coencide almost exactly with your performance.

see the edit i put in my previous post just as a side note i'd also like to bring up

Quote:

---

Originally Posted by craig588

1T adds about 25MHz of performance. That happens to coencide almost exactly with your performance.

---

so running cpu @ 235x10 with ram @ 194 mhz 2-3-3-10-1T should be the best?

it'll be nice to have hl2 and its timedemo feature... any ideas on pushing cpu further than 235x10? should i try lower multipliers?

ram is only rated up to 2.8 volts, so the higest setting i feel comfortable setting it at is 2.7 (seeing as board overvolts up to 2.79). highest bios will allow is 2.8. is it safe for ram to bring voltage higher than 2.8 volts? i have alifetime warranty, and, let's face it, mushkin would never know i'd raised vdimm higher than 2.8 if i had to rma it.

why would ram not allow 1T after 206 mhz? should i try loosening other timings? how much effect on performance do each of the main 4 timings have?

EDIT: sry for the double post

The ram will be fine at 2.8v.

i really dont recommend WC, save an extra 100 bucks and get a mach1 off ebay then have an AC guy regas it

OT: why do the stock phase units come with crappy gas?
binkgle: try running the ram at 2-2-2 and then find the max HTT it can go.
then find ur max cpu speed and find a divider that will allow your ram to be as fast as possible.
with a64s cpu speed takes preference over ram speed

take this for what its worth since I have never run your ram or mobo

270-275 FSB with a 4/5 divider and 9X multi and lowering ldt multi down to 3x if needed.

this should put the CPU around 2475 mhz the ram at 220 mhz

see if it will run that at 1T with low timings ,, I never use 2T

i'll see if ram runs 2-2-2 when at ddr333, but would prefer to not have to manually reset my bios (again).

if i were to run 300x8 (ldt multi 3x), would setting ram stock to ddr266 be ideal?

have any of you heard about the thing were a64s should use highest available multi? is this true?

EDIT: i tried running ddr266 stock with 300x8, but comp wouldn't manage to boot (would get past loading of bios, but would hang when trying to boot from hd). ram was @ 2.8v, cpu @ 1.75v. hmm...

Quote:

---

Originally Posted by binkgle

EDIT: i tried running ddr266 stock with 300x8, but comp wouldn't manage to boot (would get past loading of bios, but would hang when trying to boot from hd). ram was @ 2.8v, cpu @ 1.75v. hmm...

---

Dude thats really confusing.. are you saying you had your max mem speed set to 133mhz? then set HTT in bios to 300mhz?

I'd try using clockgen, I couldn't boot at 10x245mhz 1:1 but when I set it in clock-gen it ran 100% stable 24/7... computers are odd sometimes.

Quote:

---

Originally Posted by WesM63

Dude thats really confusing.. are you saying you had your max mem speed set to 133mhz? then set HTT in bios to 300mhz?

I'd try using clockgen, I couldn't boot at 10x245mhz 1:1 but when I set it in clock-gen it ran 100% stable 24/7... computers are odd sometimes.

---

sry i confused you. ram was set to ddr266 when cpu was at stock. with an htt of 300, that would come back up to ddr400 (i think). timings used were 2-3-3-10-1T. I'll try with 2T. And I'll try memtest86+, as well.

8x300 using 133mhz memory divider comes out exactly at 200mhz for the memory.

Did you set your HTT to 3x? If HTT total mhz is over ~1100 it can cause stability or corruption issues sometimes, depending on the board etc. Good idea to shoot for around 1000mhz or under.

HTT multi X HTT FSB (3 x 300 in this case) = HTT Frequency (900mhz).

yes, i did it with htt mulit (bios calls it ldt multi) 3x. don't know why there was the hang, but i know it didn't work, so...

It is the LDT multi, like i said, don't refer to memory speed as HTT it confuses us all, also don't refer to memory mhz in its DDR speed, also confusing (well when we are trying to troubleshoot your problems).

My rule of thumb is anytime your HTT speeds exceed 265mhz lower your LDT multi to 2 or 2.5x.

Quote:

---

Originally Posted by WesM63

It is the LDT multi, like i said, don't refer to memory speed as HTT it confuses us all, also don't refer to memory mhz in its DDR speed, also confusing (well when we are trying to troubleshoot your problems).

My rule of thumb is anytime your HTT speeds exceed 265mhz lower your LDT multi to 2 or 2.5x.

---

i can't do half multipliers. anyway, isn't 2x a bit low? there must be some perf difference between 4x and 2x! i'll give it a shot, though.

well, i tried 300x8 with 1.65 volts (overvolted to 1.7) with ram @ 200 mhz 2-3-3-10-1T and ldt multi 2x. i managed to boot, did very well in the sandra cpu arithmetic benchmark, then tried to run 3dmark05's cpu test. system hung when loading the first test. should i try with voltage set to 1.7v (effectively 1.75)?
anyway, how exactly does a cpu multiplier work?

Nice one binkgle, your getting there now.

Try a little more voltage on the CPU, 1.75 is about as high as I would go. Just to see if you get get the system stable.

If that dosent help try lower memory timings (keep 1T, but try 2.5,4,4,10.1t)

If thats still no use try lowering the HTT a bit, try 295..

Quote:

---

Originally Posted by Holst

Nice one binkgle, your getting there now.

Try a little more voltage on the CPU, 1.75 is about as high as I would go. Just to see if you get get the system stable.

If that dosent help try lower memory timings (keep 1T, but try 2.5,4,4,10.1t)

If thats still no use try lowering the HTT a bit, try 295..

---

i bet this is a prob with the cpu, not the mem, as the mem is running @ 200 mhz (stock) when cpu is @ 300x8. i'll run memtest86+, though, just to see.

i'm gonna test it with prime95, too. i think i'd call it truly stable if it can do about an hour, maybe even an hour and a half. what do you think?

EDIT: 300x8 lasted 3 minutes in prime95, and 295x8 was reaching the same, very high, load temp (61-63 celcius), so i didn't wait to see how long that would last. i think for this i'm gonna need water cooling. i'd want to cool cpu and gpu, and, if possible, northbridge, and would prefer to start off with a kit rather than lots of seperate parts that i buy. what do you guys think of the coolance exos or the zalman reseator? for both i'd get a better gpu block. also, how would i make sure that it wouldn't leak?

Hmm, getting better but i've seen 3000+ NC's hit 2.7ghz on 1.6v. I would'nt doubt if your ram is causing some problems also. Did you try using 2.9v on the ram?

Your HTT speed dosen't increase perfromance what-so-ever as i've stated allready. HTT is the speed at which the CPU talkes to its componets, and the HTT speed is allready talking faster than the componets can, thus not increasing performance.

should i run ram at 2.89v when cpu is doing 300x8, cause ram is running at stock 200mhz? i'm sure my cooling has something to do with it.

Its not going to hurt it, I'd try less cpu mhz and more memory mhz.

Quote:

---

Originally Posted by WesM63

Its not going to hurt it, I'd try less cpu mhz and more memory mhz.

---

could you recommend a specific htt x multi and htt:ram divider? i'm getting lost in all this testing.

Quote:

---

Originally Posted by WesM63

Hmm, getting better but i've seen 3000+ NC's hit 2.7ghz on 1.6v. I would'nt doubt if your ram is causing some problems also. Did you try using 2.9v on the ram?

Your HTT speed dosen't increase perfromance what-so-ever as i've stated allready. HTT is the speed at which the CPU talkes to its componets, and the HTT speed is allready talking faster than the componets can, thus not increasing performance.

---

So your saying that 9x300HTT 3x, 133mhz divider at 2700mhz will not increase perf. compared to 13.5x200HTT 1:1 ratio????

Quote:

---

Originally Posted by The Virus

So your saying that 9x300HTT 3x, 133mhz divider at 2700mhz will not increase perf. compared to 13.5x200HTT 1:1 ratio????

---

when he said htt he was talking about the htt multiplier, and meant that having it set at 2x rather than 4x really has no performance impact at all.

this thread died over the past few days.
question: would 300x7 provide a noticeable perf boost over 204x10? ram would be @ 200 2-3-3-10-1T (maybe/probably 2-3-2)

EDIT: wasn't stable/too hot (mid 50's celcius within a minute of starting prime95 blend torture test) when running 300x7 1.75 volts, ram @ 200 mhz. i guess i just can't oc very much (i.e. not at all) effectively. i'll try again when i get hl2. if you've any more suggestions, please, please, please post them here.

2nd edit: would taking the fan controller of my hsf and letting it run at 2500 rpm as oppossed to 1500 help? would there be a noticeable cooling perf difference? noise level?

shoul be able run 300x7 at stock volts easily??

as long as CPU and ram speed are constant performance should be within a couple %.

i've tried running superpi with everything at stock and i gets errors when performing the 32M digit test. why? cpu @ 204x10 1.50v, ram @ 204 mhz 2-3-3-10-1T 2.75 volts.
i'll try 300x7 again

300x7 errored while running superpi 1M digit test, and about 30 sec in prime95. vcore was 1.65. ram was at 200 mhz, 2-3-3-10-1T, 2.75v

Try cas 2.5 or 3, maybe 2T, volts up or down... if its not stable you have to turn something down.

how much of a perf dif is there between cas2 and 2.5? 1T and 2T? why is 1T so much faster?

anyway, it wasn't a mem error, mem was at stock. i'm sure that was a cpu limitation.

cas latency doesnt matter too much, about as much as cpu multiplier.

2T takes a big hit.

1T is faster because it sends commands once per clock cycle, 2T is once every other clock cycle.

ive gotten up to 2.55ghz recently because of my new OCZvx i love it. i got the same proc as you. your ram clock & timeings suxxors.

i seem to be stable now with cpu @ 240x10 1.62v, ram @ 200mhz 2-3-3-10-1T 2.8v.

i hate my ram, but it was only $100, which fit my imit of $800 when i recently overhauled everything about my computer except drives, even case (well, hd died last week, so I bought a raptor :) ) my ram is only one step up from value (it's mushkin blue line, rated for 2-3-3 @ 200 mhz), so it's my major ocing bottleneck. i won't be able to get more ram for a while, unless i get a job and stop leeching from my bank account (which my mumhas control over, so i can't do anything without clearing it with her). where did you get your ocz vx ram? do you use their ram booster thing? what speed and timings do you run at? where did you get it? would boosting its voltage help its oc? it cannot, and this has been proven with this type of ram, be run over 204 mhz with 1T stably, which sucks.

i thought cpu multi was important (200x10 is much faster than 200x9, isn't it?)!

cpu multi is important.

10x200 with 200mhz mem is faster than 9x222 with 183mhz mem divider(9/10, 200mhz mem)

what exactly is cpu multi? how does it work?

i'm gonna try to squeeze proc up to 244x10, therefore ram to 204 mhz, and 2-3-2-10-1T (which are the timings i use at stock).

okay, 240x10 and 235x10 (mem at 200 and 195 respectively) cannt pass even a minute of prime95. what is this? mem is at stock speeds and timings, with extra voltage! cpu isn't overheating. it loads in mid to high 40's (celcius). why isn't this working?

Because your CPU isn't stable at those speeds.

any way to improve the situation?

There may be, there may not be, but this isn't the thread for it. Read some more overclocking guides.

i'm away until monday, but if anyone has any ideas, please post here!

270x9 1.62v seemed ok. mem @ 221 3-3-3-10-2T or 2.5-3-3-10-2T, 2.79v (though i might pump in an extra .1v). sandra cpu arithmetic was 11062, 3969/5157, mem bandwidth benchmark was 3008, 3013 when ram was @ 3-3-3-10. when i lowered ram to 2.5-3-3-10, mem benchie went up to 3025, 3026, but cpu went down to 10401, 3970/5160.

i'll test further for stability (so far i've only done sandra and a 1M calculation with super pi), but which of these ram timing setings means better perf?

EDIT: as a reference, with cpu @ 204x10 1.52v, mem @ 204 2-3-3-10-1T 2.74v, sandra cu arithmetic was 8779, 3332/4329, mem bandwidth was 3137, 3137

anyone?

oh yeah, and both of the below 270xp setups failed prime95 in under a minute