My new system

[Sorrento]

Also the noise factor weights a ton in favor of any watercooling setup.

[Thraxz]

I just ran a series of tests in superpi/sisandra and came up with conclusive evidence that bandwidth is not useless... it's more important than mhz by some and far more important than timings. (At least in superpi) I'll organize and post it later.

[NEOAethyr]

I love THunDA's and AG's sig's lmao...

Funny stuff.

 

Anyways the other board you had.

And did'nt have a diff vga card that got fried?

Something about caps right?

 

I think the answer is right there maybe, if tis not cooling related.

 

However, you shoud be going by diode temps.

It's the only way to look at it.

Socket is all cool when comparing stuff, but it does'nt cut it.

Watch your spikes.

 

 

Anyways the vga card, may of been using less power.

Or may of been feeding out less dirty power, or..., may of ran cooler.

Or......, may of produced less emi.

 

The mobo.

The old one.

Any mod from dfi will dononce this 100%.

However not all boards are created equal, they would admit that one.

However, not all board have the same exact parts on them.

 

For example.

Caps.

Some caps do better, call it performance wise, however they may or may not last as long ^^.

Others may last longer, maybe, but "performe" worse.

 

Geneerally if you are'nt runnig the rig into the ground, in a bad way, the caps should be fine.

I cannot say this for sure.

However I've never ever had an issue with a cap blowing out on me.

I'm not saying it could'nt happen from a mishap from manafacturing or whatever, but still.

Generally they all last for along time, even if they are really crappy caps.

 

I don't speak out agais'nt dfi's choice in caps, no.

They seems to place decent ones where they are needed on the cpu,and the rest, well it can depends but apperently they do pretty good even for memory, when they put bad name brand ones there.

 

I don't really know how to say it.

But a cap is a cap.

Unless it's a high end one.

 

In any case.

Cpu stability could be many things.

Psu, diff mobo even if it has all the same parts,resistors, caps, everything.

 

Heck, the cpu may not liek being that high right now, perhaps from the previous board blow out, or maybe because it was ran at higher temps for a peroid of time.

Who knows.

These things, very hard to figuer out.

 

Gel application...

Major on that one, I mean shoot, if the heatsink soluition is getting hot, and keeping the gel at a higher risk.

Then the gel won't last as long.

I know you just got the water cooling on there.

However, a diffrence in application, on how it spread afterwards, is sorta major.

Again, doide spikes.

 

 

I like how ag said it.

Water and air, allmost the same.

Heck, many times I've seen air do better then water.

Sometime way better.

Sometimes close, and sometimes water slightly better.

Why???

AG explained most of it.

I'd liek to add that the majority of all piece by pice and complete setups for water cooling, are not the best water cooling you can get.

The best involves specialty parts.

Such as special radiators, good tubing, the right water and mix, very good blocks(sometimes custom).

Pumps, pumps that they don't sell for pc cooling , same for rads.

Ie, an oil cooling, or transmission cooler, of the right size and the right type of umm, what do you call them...

Ones that are'nt tubes, because those suck.

The ones like car radiators in the, the big ones, those types of capulary's, probably the wrong name err...

Not saying you need big rads, ones that handle 1x or 2x 120mm's.

 

A nice copper block, not one that has some puny heatsink imbedded in them and the rest is plastic.

 

Big tubing, strong tubing.

Etc etc...

Layout...

 

I never built one though.

Still, I know theat it's possible to do better then any pc based set they sell.

Even though you may by parts that relate to pc cooling,m not all of the best is pc based, for sale as pc based is what I mean.

 

Even so, if you had the best of the best, custom everything or something.

It may not do that much better.

 

Just because you're running 10c lower does'nt mean you can get any more mhz.

You need to cool it faster, so those spikes, even if they don't reach as high(because of lower temps), they still reach high enough to stop the cpu in it's tracks.

 

Dirty power.....

Blah blah blah.

 

It's best you run the rig for a while it's max, it's stable max.

From there, after a while, try somethign with it.

Try lowering the vcore, or upping the speed, or whatever .

 

Remember that there is nothing wrong with something you've been testing that has been running good, and is still running good for fsb/mem and some decent cpu speeds.

Just because it can't reach something you've been to before, does'nt mean you can't ever again.

Then again it you might not beabel to.

 

I'd reccomend cutting off the ihs if you have'nt allready.

Do the small things.

Perhaps it's how the psu is hooked up to everything, even thoug it's runnign pretty decent, you may beabel to get more out of it by having everything on it's own lead if possible.

Ie, each hd on there own lead, same for burners, etc etc.

Psu temps, if for some reason they are getting hotter now..., seperating each device from all other devices will help a bit on this anyways.

Etc.

 

 

Blah.

It's either:

Emi

Power related, lead distro/layout or heat

Something is messed up very slightly, and needs time to recover to it's full potential

The new board is'nt enabling everything, stuff that's hidden(like ecc for the cpu perhaps)

The new boad is'nt as good, so it needs time, and the cpu needs time to cope

 

 

Many things , hehe, gotta love how pc's throw ya for a loop.

 

 

Btw AG.

I listen .

I never meant it to seem liek I was'nt, just mean everything as generic talk from tiem to time.

However, I do believe now, after experiencing what I have at these speeds and constant superpi'ing.

The diff is probably(very likely) very very min on those platform's when it comes to timings after certain speeds.

Benches mean squat, espeically everest and sandra, because for one, they can't bench everything, they can miss the inpact of certain timings.

 

Gaming wise and everything else(mostly I'm guessing) is very likely very min or no diff at all.

 

Mhz is king.

Even on the old nf2, why..., because fsb, fsb is the total amount the cpu can handle.

If the cpu is being limited, you can only get so much to it.

On the nf4's, well, they hab the min latency amoutn with total ease, I mean, small speeds and super low latency's.

The bandwith goodness comes from mhz then, because it's less likely that any bench will notice a diff in the timings.

Even in some program where the timings may make a diff, the diff would be min because you would likely be around it's peak anyways.

 

I never thaught that any program could reach some sort of peak of memory bandwith it could possibly use.

However I've seen it.

 

So...

Got the memory that does low timings only, do that.

Got the meory that does the high mhz only, do that.

Whatever the memory likes, peak that out, because the bus or cpu may love you for it.

That and whatever you can gain is good anyways.

Timings are good, but mhz is king...., for memory anyways

 

Do I sound allright now lol ? (can you hear me now.. lol)

 

 

 

Edit:

TGM...

I remember seeing a pic of your rad setup a while back.

You have the fans sucking the air out of it.

However, you can do better by pushign air into it, like air cooling on the cpu.

 

For one, the fans will get hot anyways, the moter is on top of the heat so it doe'snt matter what way you point them, the fans will just get hot if they do.

However, better cooling is to be had by pushing air into the rad.

 

If the fans are high performance, you coudl perhaps add spacers.

Liek say if they are 2x120's.

Add dead cut out 120mm's inbetween the fans and the rad, so the fans don't get hot, and add to the heat.

[TheGeekMistress]

no

 

there's nothing wrong

 

the problem that most of you don't seem to grasp yet is that aircooling is pretty much maximized these days...technology has brought us to the edge of the best cooling on air...ask the guys from DangerDen and Swiftech...they'll tell you the same thing.

 

And water...water will only run your gear at or near room temp (ambient water/air temps). It can wick away voltage related heat better than air, but not anything close to phase-change/subzero cooling.

 

 

This is just another one of those rumors or un-facts as I like to call them that people still believe in (like memory timings are important on A64's lol).

 

Watercooling is NOT going to give you more overclock. I have NEVER been able to get more than 5% overclock maximum going from air to water....because, as I said, aircoolers are at a maximum efficiency these days, and so is water. There just isn't going to be a difference in overclocking.

 

Temps...now there's a difference. And if your aircooled overclock was netting 70C and limiting you to 2500Mhz, then of course a good water setup will bring you down to say 40C and get you up to 2900Mhz etc.

 

But

 

 

I, and 99% of the rest of you, can hit 2900Mhz on air @ 60C, and switching to water will only give you maybe 2950Mhz, but you will be back down to 45C or such.

 

 

This is just the way it is.

 

To gain a couple more %, you cut off the IHS...but it still won't give you but a couple % increase maybe...but it WILL lower your temps a few degrees more.

 

 

 

 

 

this is the reality of water vs air cooling.

 

oh

 

and watercooling, if done right, is silent lol.

 

 

oh

 

and also watercooling the gpu can have a significant impact, but again, if you have a killer air cooler like a Zalman on your gpu...you won't see much more overclock on a gpu when going to water either

sorry AG, i have to take issue with that.

 

on my first DFI NF3 mobo, before it died ...i was running the 948U with a SFII on it. my max frequency was 2950mhz ...i switched it over to my water cooling and straight away i was running 3150mhz rock stable. then i pushed it a bit futher and i got 3200mhz. so in my own personal experience, what you'r experience with water cooling doesn't hold water *pun intended* hehe

 

TGM

[smolt]

sorry AG, i have to take issue with that.

 

on my first DFI NF3 mobo, before it died ...i was running the 948U with a SFII on it. my max frequency was 2950mhz ...i switched it over to my water cooling and straight away i was running 3150mhz rock stable. then i pushed it a bit futher and i got 3200mhz. so in my own personal experience, what you'r experience with water cooling doesn't hold water *pun intended* hehe

 

TGM

just goes to show every board and setup is different. NOT to mention user error

[KimTjik]

sorry AG, i have to take issue with that.

 

on my first DFI NF3 mobo, before it died ...i was running the 948U with a SFII on it. my max frequency was 2950mhz ...i switched it over to my water cooling and straight away i was running 3150mhz rock stable. then i pushed it a bit futher and i got 3200mhz. so in my own personal experience, what you'r experience with water cooling doesn't hold water *pun intended* hehe

 

TGM

I thought that was what he said: some years ago it did make a bigger difference, but as for now when air-cooling also has improved the difference might no longer be that great, but you still have other gains...

[TheGeekMistress]

I love THunDA's and AG's sig's lmao...

Funny stuff.

 

Anyways the other board you had.

And did'nt have a diff vga card that got fried?

Something about caps right?

 

I think the answer is right there maybe' date=' if tis not cooling related.

 

However, you shoud be going by diode temps.

It's the only way to look at it.

Socket is all cool when comparing stuff, but it does'nt cut it.

Watch your spikes.[/quote']the A64's aren't set up like the older A-XP's with their 462pin and opening in the center of the socket for a thermal probe. the only temp that is available to monitor on an A64 is the internal diode temp and it's doesn't act like the diode temp reading on the A-XP's.

Anyways the vga card, may of been using less power.

Or may of been feeding out less dirty power, or..., may of ran cooler.

Or......, may of produced less emi.[/qoute]that's a moot point as i have taken out my 7800gs that replaced my x800 and place an old FX5200 which takes nothing to power it, and the results are the same less than 100mhz greater overclock from air to water this time around.

The mobo.

The old one.

Any mod from dfi will dononce this 100%.

However not all boards are created equal, they would admit that one.

However, not all board have the same exact parts on them.

 

For example.

Caps.

Some caps do better, call it performance wise, however they may or may not last as long ^^.

Others may last longer, maybe, but "performe" worse.

 

Geneerally if you are'nt runnig the rig into the ground, in a bad way, the caps should be fine.

I cannot say this for sure.

However I've never ever had an issue with a cap blowing out on me.

I'm not saying it could'nt happen from a mishap from manafacturing or whatever, but still.

Generally they all last for along time, even if they are really crappy caps.

 

I don't speak out agais'nt dfi's choice in caps, no.

They seems to place decent ones where they are needed on the cpu,and the rest, well it can depends but apperently they do pretty good even for memory, when they put bad name brand ones there.

 

I don't really know how to say it.

But a cap is a cap.

Unless it's a high end one.

yes, there are lesser and greater performaning motherboards of the same model, but i thought we all already came to that same conclusion. i have a 2nd DFI NF3 skt754 mobo, i bought two b/c DFI discontinued the boards. after i do a freash OS install with this board, i'll prolly try the 2nd board for comparison.

In any case.

Cpu stability could be many things.

Psu, diff mobo even if it has all the same parts,resistors, caps, everything.

 

Heck, the cpu may not liek being that high right now, perhaps from the previous board blow out, or maybe because it was ran at higher temps for a peroid of time.

Who knows.

These things, very hard to figuer out.

 

Gel application...

Major on that one, I mean shoot, if the heatsink soluition is getting hot, and keeping the gel at a higher risk.

Then the gel won't last as long.

I know you just got the water cooling on there.

However, a diffrence in application, on how it spread afterwards, is sorta major.

Again, doide spikes.

again, the A64 is a whole nother breed compared to the A-XP and it's reading off the diode only and my temps have never exceeded 50C on or off water. right now at 3.1ghz i'm idleing at 30~33C and full loading at 40~44C; and at 3000mhz on air the idle temps were 38~39C idle and 50C load.

I like how ag said it.

Water and air, allmost the same.

Heck, many times I've seen air do better then water.

Sometime way better.

Sometimes close, and sometimes water slightly better.

Why???

AG explained most of it.

I'd liek to add that the majority of all piece by pice and complete setups for water cooling, are not the best water cooling you can get.

The best involves specialty parts.

Such as special radiators, good tubing, the right water and mix, very good blocks(sometimes custom).

Pumps, pumps that they don't sell for pc cooling , same for rads.

Ie, an oil cooling, or transmission cooler, of the right size and the right type of umm, what do you call them...

Ones that are'nt tubes, because those suck.

The ones like car radiators in the, the big ones, those types of capulary's, probably the wrong name err...

Not saying you need big rads, ones that handle 1x or 2x 120mm's.

 

A nice copper block, not one that has some puny heatsink imbedded in them and the rest is plastic.

 

Big tubing, strong tubing.

Etc etc...

Layout...

 

I never built one though.

Still, I know theat it's possible to do better then any pc based set they sell.

Even though you may by parts that relate to pc cooling,m not all of the best is pc based, for sale as pc based is what I mean.

 

Even so, if you had the best of the best, custom everything or something.

It may not do that much better.

 

Just because you're running 10c lower does'nt mean you can get any more mhz.

You need to cool it faster, so those spikes, even if they don't reach as high(because of lower temps), they still reach high enough to stop the cpu in it's tracks.

like i was saying to AG, on my first DFI NF3 before it died on me i was running 2950mhz on air and then i installed my water and straight away i went to 3150 solid. then pushed a bit i got 3200mhz on water compared to 2950mhz on a 948U with a SFII.

Dirty power.....

Blah blah blah.

i'm running an Enermax 651P with 36amps on the 3.3v, 5v, and 12v lines and i bought the OCZ line conditioners for my video card and my high rpm raptors drives.

It's best you run the rig for a while it's max, it's stable max.

From there, after a while, try somethign with it.

Try lowering the vcore, or upping the speed, or whatever .

 

Remember that there is nothing wrong with something you've been testing that has been running good, and is still running good for fsb/mem and some decent cpu speeds.

Just because it can't reach something you've been to before, does'nt mean you can't ever again.

Then again it you might not beabel to.

 

I'd reccomend cutting off the ihs if you have'nt allready.

Do the small things.

Perhaps it's how the psu is hooked up to everything, even thoug it's runnign pretty decent, you may beabel to get more out of it by having everything on it's own lead if possible.

Ie, each hd on there own lead, same for burners, etc etc.

Psu temps, if for some reason they are getting hotter now..., seperating each device from all other devices will help a bit on this anyways.

Etc.

i understand that sometimes after running a while that you may or may not get a little bit more out of the equipment, but it's like a kick in the butt when i've tried everything to even get to 100mhz over air when the last time i got to 3150mhz (200mhz over air) solid stable.

Blah.

It's either:

Emi

Power related, lead distro/layout or heat

Something is messed up very slightly, and needs time to recover to it's full potential

The new board is'nt enabling everything, stuff that's hidden(like ecc for the cpu perhaps)

The new boad is'nt as good, so it needs time, and the cpu needs time to cope

 

 

Many things , hehe, gotta love how pc's throw ya for a loop.

well i haven't changed any wiring or electrical components that aren't inserted to the mobo. it was basically a pull out of the mobo, place replacement mobo in and used air instead of water till i got my water cooling ready again. and this time i'm only cooling the cpu. last time my loop cooled the cpu, chipset, and vga. only thing else thats changed is the vid card and placing in an old low power FX5200 made zero difference.

Btw AG.

I listen .

I never meant it to seem liek I was'nt, just mean everything as generic talk from tiem to time.

However, I do believe now, after experiencing what I have at these speeds and constant superpi'ing.

The diff is probably(very likely) very very min on those platform's when it comes to timings after certain speeds.

Benches mean squat, espeically everest and sandra, because for one, they can't bench everything, they can miss the inpact of certain timings.

 

Gaming wise and everything else(mostly I'm guessing) is very likely very min or no diff at all.

 

Mhz is king.

Even on the old nf2, why..., because fsb, fsb is the total amount the cpu can handle.

If the cpu is being limited, you can only get so much to it.

On the nf4's, well, they hab the min latency amoutn with total ease, I mean, small speeds and super low latency's.

The bandwith goodness comes from mhz then, because it's less likely that any bench will notice a diff in the timings.

Even in some program where the timings may make a diff, the diff would be min because you would likely be around it's peak anyways.

 

I never thaught that any program could reach some sort of peak of memory bandwith it could possibly use.

However I've seen it.

 

So...

Got the memory that does low timings only, do that.

Got the meory that does the high mhz only, do that.

Whatever the memory likes, peak that out, because the bus or cpu may love you for it.

That and whatever you can gain is good anyways.

Timings are good, but mhz is king...., for memory anyways

 

Do I sound allright now lol ? (can you hear me now.. lol)

 

 

 

Edit:

TGM...

I remember seeing a pic of your rad setup a while back.

You have the fans sucking the air out of it.

However, you can do better by pushign air into it, like air cooling on the cpu.

 

For one, the fans will get hot anyways, the moter is on top of the heat so it doe'snt matter what way you point them, the fans will just get hot if they do.

However, better cooling is to be had by pushing air into the rad.

 

If the fans are high performance, you coudl perhaps add spacers.

Liek say if they are 2x120's.

Add dead cut out 120mm's inbetween the fans and the rad, so the fans don't get hot, and add to the heat.

no, i have a push/pull setup so i have 4 120's on my rad. and i think you've got it backwards, using the pull method if only using fans on one side of the rad is the best way to go. b/c you don't get the heat build up off the fan's hub and there's no dead zone for a hot spot to occur under the hub using the push method.

 

TGM

[TheGeekMistress]

i'm gonna go eat some jelly beans, watch a movie, and get my head out of this frustrating situation

 

TGM

[Thraxz]

CPU-----MEM--------timings--------bandwidth----superpi time---

2.7-------180------2-2-2-5-1T--------5700-----------33.0xx------

2.7-------246------2.5-3-3-5-1T------6850-----------31.6xx------

2.7-------225------2.5-3-3-5-1T------6500-----------32.1xx------

2.7-------246------3-4-4-8-1T--------6650-----------32.2xx------

2.7-------246------2.5-3-3-5-2T------5800-----------32.3xx------

2.7-------270------2.5-3-3-8-2T------6200-----------31.7xx------unstable

2.7-------270------3-4-3-8-2T--------6100-----------32.1xx------

2.8-------255------2.5-3-3-5-1T------7100-----------30.4xx------semiunstable

2.8-------255------2.5-3-3-5-2T------6000-----------31.1xx------

2.8-------255------3-4-3-8-1T--------7050-----------30.8xx------

 

 

You tell me what YOU think.

[ReelFiles]

What do you consider semi-unstable? Either it's stable or it's not, imo.

[TheGeekMistress]

CPU-----MEM--------timings--------bandwidth----superpi time---

2.7-------180------2-2-2-5-1T--------5700-----------33.0xx------

2.7-------246------2.5-3-3-5-1T------6850-----------31.6xx------

2.7-------225------2.5-3-3-5-1T------6500-----------32.1xx------

2.7-------246------3-4-4-8-1T--------6650-----------32.2xx------

2.7-------246------2.5-3-3-5-2T------5800-----------32.3xx------

2.7-------270------2.5-3-3-8-2T------6200-----------31.7xx------unstable

2.7-------270------3-4-3-8-2T--------6100-----------32.1xx------

2.8-------255------2.5-3-3-5-1T------7100-----------30.4xx------semiunstable

2.8-------255------2.5-3-3-5-2T------6000-----------31.1xx------

2.8-------255------3-4-3-8-1T--------7050-----------30.8xx------

 

 

You tell me what YOU think.

looks to me like i once suggested, that a balance of tight'ish timings and higher mem bus are going to give you the best superPI score. i would say the 2.8ghz / 255 @ 3-4-3-8-1T (1T helps alot), is your best config. you can't get any better from your data. now that thats out of the way, the one single thing that will lower your superPI score by lowering latencies is to increase the memory controller frequency which is internal to the cpu, so to do that, you increase your core frequency. but, 2.8ghz might be your limit. but that's the only thing that's left to lower your superPI score with your config.

 

TGM

[Thraxz]

There are varying degrees of unstable... can't run superpi... can't run 29089257083027 hours of 2347523490527390 prime and can't boot. This gave a single superpi error before running it three time in a row without a hitch... the unstable one took about 5 tries to get ONE run through. As a good scientist, I actively noted any variances/complications in the test procedure to help ensure it's validity. That being said, the numbers adhered largely to the trends already in place in that point so I felt it good to include the numbers with said note.

 

Now stop being a wankery philosopher about something as trivial as describing a memory setting as semistable or stable and look at the numbers involved.

 

looks to me like i once suggested, that a balance of tight'ish timings and higher mem bus are going to give you the best superPI score. i would say the 2.8ghz / 255 @ 3-4-3-8-1T (1T helps alot), is your best config. you can't get any better from your data. now that thats out of the way, the one single thing that will lower your superPI score by lowering latencies is to increase the memory controller frequency which is internal to the cpu, so to do that, you increase your core frequency. but, 2.8ghz might be your limit. but that's the only thing that's left to lower your superPI score with your config.

 

TGM

 

This is stuff I already know. I was referring to the corrolation between whatever variable and the output superpi time. It is uncontested that CPU speed is king of superpi hill and was not the issue here. You'd do well to note that the most direct parallel in question is between the bandwidth and the superpi time... not MHz or timings or command rates. It also appears that MHz is closely behind bandwidth in the superpi time importance. It appears that timings/command rate are a far far distant third.

 

The bold part is totally wrong. Your super pi times go down because your processor is processing faster... period.