Taking Bets On Speed!

Ok, no voting here. Hehe. Just a simple post taking bets on what I'll hit on the new mobo.

I did this when I blew up my Cobra engine with 28psi of boost on the stock engine (i was building a motor for it already). I took bets on how many cylinders were toast.

BACK to the topic:

I'm taking bets on how far I'll be able to overclock my setup with the new mobo when it arrives tomorrow. I have work to do, so it might not be until the weekend (I work from this computer).

Fyi, by max overclock I mean the speed and FSB for complete stability during:
- Demanding games
- Development and Graphics work (including MSSQL on W2K3, I'm running W2K3)
- No dangerous heat level

You can see the setup here and watercooling in the sig: http://www.ocforums.com/showthread.php?t=345473


A hint: I've hit 4.5Ghz in POST several times. Just the graphics locks up when switching to GUI mode (does this for 4.3 and 4.4). Common issue amongst the PCI Express boards that don't unlock...

The P5AD2-E that arrives tomorrow does unlock that.


I'll go first... 4.7Ghz

I'd like to hit this stable so I can use the A.I. overclocking tool, so the system idles at 3.6Ghz, but jumps to 30% overclock (4.7Ghz) with a 260mhz bus (1040mhz FSB).

OT, after 8,000 miles on the cobra the winner was 1: one piston let the ring land go and it ate oil, even though the top of the piston was fine. The fact that it had 7 cylinders that blew the head gasket out, and stretch a number of head bolts to just being finger loose I'm sure was a factor. Had plans to build a new negine anyhoot to hold the boost.

Next...

Would you care to share a little of your money with me? or maybe just the car....Please?

Anyway, I'm betting that you'll hit 4.4 stable..... maybe 4.6 for a screenie. Good luck on the 4.7 though...... that would be cool to see.

i dont think the new mobo will give much of a difference... ill guess 4.4 stable.

beau_zo_brehm said:

i dont think the new mobo will give much of a difference... ill guess 4.4 stable.

Click to expand...

Well, it's a known issue with the version of the board I have (non "-E") that the PCI-E bus does not unlock. When it actually unlocks with the P5AD2-E, it's all over the net. People hitting 288mhz bus speeds, with one guy at 300 (and PCI Express card and SATA drives).

You can read all of the issues I've ran into and suspect in the link above.

Any reason to think the CPU will even OC that high regardless of the PCI-E lock?

My 3.2 "540" should be here tomorrow, along with an P5GD1. I wasn't ready to do DDR2, hence the 915 mobo, but I didn't see there was a lock problem with the ASUS boards until today If it doesn't lock or unlock as the case may be, I will be RMAing straight away for an "E" mobo. Had I realized I could get some decent $$ for my oh-so-awesome BH-5 XMS, I probably would've gone that route in the first place, but adding DDR2 sticks with a mobo, CPU and vid card would've been a tad too pricey with Xmas right around the corner

Only problem is that my new 775 WB won't be here til Monday, so no water for a few days...

The PCI-E bus lock is imperfect by design. Intel don't want people to overclock, they want you to buy a more expensive CPU if you want more speed. A few mobo makers like Asus and Abit (perhaps others) have figured out how to "manage" the PCI-E bus, at least up to a point. For example, on my Abit AG8, the PCI-E has an "auto" and a "lock" setting. The auto setting acually works better. The lock setting is screwed up. Generally, this will allow you to be stable up to about 250 to 260 FSB or so. The PCI-E bus speed sort of floats on the auto setting. If you want extreme overclocking, then, you probably need to invest in the newer i925XE chipset.

Thanks for the info. How much of a float is there? I vaguely remember seeing someone saying they had it set for 90MHz in BIOS, but it was running 107 or 114. ie. the lock wasn't doing anything at all (Asus).

I'll be running air til my new WB/rad/pumps (2x MCP-350s) gets here (Mon-Tues), so I don't think major OCs are going to happen til then. I'd probably be really happy with 260FSB (4.1GHz), but if it gets there with relative ease, maybe I'll need to spring for the 925XE to see just how far it will go

I guess in the meantime, I can do some lesser OCs and see what's up with the PCI-e bus...

Ross said:

Any reason to think the CPU will even OC that high regardless of the PCI-E lock?

My 3.2 "540" should be here tomorrow, along with an P5GD1. I wasn't ready to do DDR2, hence the 915 mobo, but I didn't see there was a lock problem with the ASUS boards until today If it doesn't lock or unlock as the case may be, I will be RMAing straight away for an "E" mobo. Had I realized I could get some decent $$ for my oh-so-awesome BH-5 XMS, I probably would've gone that route in the first place, but adding DDR2 sticks with a mobo, CPU and vid card would've been a tad too pricey with Xmas right around the corner

Only problem is that my new 775 WB won't be here til Monday, so no water for a few days...

Click to expand...

ASUS will not RMA it for the "-E". I tried, many of times... Refuse the package and get your money back. Then get a P5xxx-E version somewhere else.

batboy said:

The PCI-E bus lock is imperfect by design. Intel don't want people to overclock, they want you to buy a more expensive CPU if you want more speed. A few mobo makers like Asus and Abit (perhaps others) have figured out how to "manage" the PCI-E bus, at least up to a point. For example, on my Abit AG8, the PCI-E has an "auto" and a "lock" setting. The auto setting acually works better. The lock setting is screwed up. Generally, this will allow you to be stable up to about 250 to 260 FSB or so. The PCI-E bus speed sort of floats on the auto setting. If you want extreme overclocking, then, you probably need to invest in the newer i925XE chipset.

Click to expand...

Correct, to a point. Asus and Abit have figured out how to unlock the PCI-E bus from the 925 chipset by applying an algorythm to slow down the bus rate based on the overclocking amount. It's not a true unlock, but more of a "variable scaling down" of the pci-e bus (source from some online article). Asus does it better then Abit as Asus actually tests and clocks down the Northbridge to allow you to run SATA drives at extremely high bus rates. The LGA775 Abit "top of the line" board did not unlock it (from the article I read) and caused no SATA drives to show after boot. Switching to an IDE drive resolved that issue to get hte bus as high as the Asus could go.

But the newer P5AD2-E mobo seems to actually lock it, as told from many people in Europe and Japan with this mobo. THey have posted screen shots where they are running at 280/290/even 300 bus and the PCI Express lock is locked at 100mhz. Seems they have figured it out with the newer revision of the board.

Also another note is if you are using an ATI xX00 PCI-E card. Nvidia is restricted to a 125 to 130mhz PCI-E bus speed (250-260mhz overclock) from a few articles I found. ANd I ran into that issue here as well.

If you use an ATI card, they seem to be able to take MUCH higher PCI-E bus rates. One article I found showed the test lab clocking an Asus P5AD2 (not the "E") to 278mhz with the ATI card! Which forced the PCI-E bus speed to 139mhz, and the ATI card still booted.

I'm not sure if it is benificial to overclock the PCI-E bus.


Btw, how do I know what chipset this P5AD2-E is? You stated "i925XE", and I know my existing P5AD2 is 925X. What's the P5AD2-E? I can't seem to find that on Asus' site. Basically, I'm hoping on having a board to support dual core cpus, or the next generation cpus. Guess I'll wait to run CPU-Z.

Ross said:

Thanks for the info. How much of a float is there? I vaguely remember seeing someone saying they had it set for 90MHz in BIOS, but it was running 107 or 114. ie. the lock wasn't doing anything at all (Asus).

Click to expand...

Btw, that was me. I've attempted to lock my PCI-E at 90mhz and booted with a 15% overclock (230mhz fsb) and the PCI-E bus speed was still 114mhz! Even though I set it to 90, 95, 100, etc. No matter what I set it to, it was 'fake'.

Are you trying to high-jack my thread? LOL

Scratch my question the chipset above... P5AD2-E just arrived, and it says on the box "925XE".

It's my understanding that the Asus P5AD2-E does have the new i925XE chipset.

With my Abit AG8 and 3.2 LGA775 running at 3.9 gig with the PCI-E set to "auto", the PCI-E bus speed was 112 MHz. Since that was 244 FSB, then if the PCI-E bus wasn't being managed, then it should of been 122 MHz, so the floating PCI-E bus speed does work somewhat with Abit mobos.

Here is a screen shot of my 3.8 CPU with the multiplier dropped down to 14X on that same Abit AG8 mobo. I was able to play Medal of Honor Pacific Assault for a couple hours at that setting with no stability problems. I don't know what the PCI-E bus speed was though. But, obviously it was somehow managed or else it wouldn't of ran. That's a FSB of 285 by the way.

Nice high bus speed.

I've made a post, and didn't get a SINGLE reply about bus vs. cpu speed. They ("they" being people on this forum and another overclocking forum) state that Intel boards respond better to overclocking the front-side bus then the AMD boards.

The post I made was asking: Is it better to run a very high FSB, or is it better to run a very high CPU clocK (cpu lock vs. cpu unlock)?

For example: You are running w/cpu lock enabled, which lowers your CPU lock down to 14x. This allows you to get to a very high FSB (1144mhz) before the cpu becomes unstable at very high freq (you are at 4.0Ghz).

With my goal of 4.7Ghz (that's 700mhz faster!), that's only a FSB of 1040mhz (260mhz bus). I'm hoping the CPU will be stable at that speed. My 560 prescott has a 18x multiplier, for others to reference in these calculations (18 x 260 = 4.7Ghz).


If I used CPU Lock, that would drop me to 14x. So then my goal would be the 300mhz bus speed (1200mhz FSB), so 14 x 300 = 4.2Ghz.


I'll play around with benchmarks, but batboy: what have you experienced?

show us some stuff
and post up
good luck

No, I was not trying to hi-jack your thread...just get caught up in conversations...sorry. I won't make any more posts that aren't relevant to the thread BTW, I was talking RMA with the vendor, not ASUS directly, but I don't want to refuse the whole thing cuz it's the proc and x700 Pro vid card with it, which I know I want to keep. Thanks for the responses and info though.

To make this post relevent: as for the clock .vs higher bus speed, it's going to be a compromise obviously, but I really don't know how large. If you mainly do mem intensive stuff, a higher FSB will give you more mem bandwidth at the same clock (lower multiplier). I generally will just OC max clock to see what I "can" get it to, but re-clock concentrating on max mem bandwidth/lowest latencies/sane temps for day-to-day stuff (nothing heavy duty).

If you're quest is for 4.5GHz, unless I am missing something, you'd need a 320+FSB on the 14x multiplier. I doubt that's a possibility (maybe on vapo?), but you'd have mem bandwidth up the wahzoo! If you could get 300FSB on the 14x and maintain decent temps, that would still be a 4.2GHz clock/big mem bandwidth. That thing would probably rock-n-roll! I look forward to your results...good luck with it.

Generally, you get the best performance from high FSB and low multiplier (especially if your RAM can run high FSB at the 1:1 ratio). But, if you are after max CPU clock speed, then high multiplier and lower FSB will probably make it a little easier to reach your lofty goal. Good luck.

Ross:

LOL. I was just ragging you about the off-topic post. Hehe.

You are correct that 4.5Ghz is a 320mhz FSB with CPU Lock enabled. But I was looking to achieve 4.5Ghz (or 4.7Ghz) without the CPU Lock, so it would be using the 18x multiplier with a 250mhz bus (4.5Ghz) or a 260mhz bus (4.7Ghz).

Humm... 278mhz bus x 18 multiper = 5.0Ghz.


Ross/batboy:

Thanks for the responses on memory bandwidth and the bus speeds and such... Now I have to decide, which am I using more of? The CPU clock or memory bandwidth. The applications I use are:

50% - Microsoft VS.NET (dev. tools), IIS (dev. and perf. testing), MSSQL Server/MySQL dev. and perf. testing), HTML/Graphics/Photoshop/etc.
35% - Gaming (latest top-of-the-line 3D games, right now Joint Ops).
10% - Highend Flash (highend as in 200 to 300MB streaming movies and media) and media encoding (WMV, Real, QuickTime, AVI).
5% - Misc blueprinting for houses and other drafting applications (and 3D tools).

Yes, I went to only one PC these days... I hated the "Server Room" sound from running 3 or more PCs around here.

Yeah, I know and deservedly so...it's cool These will be my first PC updates in over a year, so I was eager with anticipation. Long story short, after having to drive to FedEx to pick up the package myself(!), it's installed and on the stock BIOS (1004), 220+ FSB = nothing...the 10% lock was in effect. Flashed a 1006 and it loaded at 221 no problem, so let the OCing begin again.

Well, everything runs through the bus so you probably just want as much FSB as you can get for best general performance like batboy said. I run my PC daily kinda like you with Apache, MySQL, PHP, PS, etc. and find that I get much better daily performance with the best mem settings. In other words, I clock for as high as I can get the mem without having to lower the divider...whatever the clock speed happens to be, that's where I run it. I don't have the dilemma of having a selectable multiplier tho.

I think first you should do some playing around and see just how high the CPU will go, if the mobo starts crapping out on FSB, what your temps look like and if it's even stable enough to run daily. If you can get 300+FSB out of it without melting your case, I think it would probably be a no-brainer answer of using the 14x. If 300+FSB is out of the question, I think it's also a no-brainer: 18x as much as it will take while keeping the DDR running as close to maxed as possible without having to drop the divider.

There's no other way to say this...

KICK ***!

Talk about SNAPPY and speedy! Yes, I got it installed today (couldn't wait) and my gawd, what a difference. I think this was party due to I couldn't run my memory any higher then 640mhz before on the older P5AD2 mobo.

So who was the winner? Well, I've only fiddled with it for a few hours for benchmarks, stability, and temps and found a happy medium I think I like:

CPU: 4.2Ghz
FSB: 1200mhz
BUS: 300mhz
MEM: 800mhz (5-5-5-15, stock settings haven't had time to play with this yet)

Guys, this system rocks! The highest I could go with the CPU lock disabled was 4.4Ghz. At 4.5Ghz, the system just wasn't stable... And running some extremely high voltages to keep it there I didn't like. So I ran a few benchmarks and started out using the CPU Lock feature. Why I didn't do this, I have no idea.

How about 6.9GB/s on the memory?
3DMark05 went from 4820 to 5322!!
System temps are almost dead exact to what they were before (actually a tad lower!): 38C idle, 50C max with full HT loaded up (prime95 won't do that, used Cinebench).

Not to mention how snappy/zippy the system is. Voltages, I'm still playing with but I've gotten them down to a safe level and the system is still stable (ran Prime95 for an hour so far, multiple tests of mem and cpu as well). It "seems" to be ok with the voltage settings,


I'm sure there's more in the system (beyond 300mhz bus), but this is a good time to call in the night on a happy note. Ok, enough with the rants. Time for pics.

Btw, this will most likely be my last post to this thread. I have another one I run (in the sig) that I'll be posting updates to. Thanks guys for playing!

Awesome! It's great that you're getting 300FSB...I thought it might be a struggle, but figured it would be great performance if you could run it. Congrats!

Are you really running 1.6Vc or is CPU-Z just showing really high like mine? At 1.4Vc CPU-Z shows 1.5xV for me. It reads my DDR freq wrong too? My "100MHz locked" PCI-E bus is up to 114 at 250FSB (4.0GHz stock Intel HSF atm) on the P5GD1, so I might be following your footsteps with a 925XE myself

Sweet. Actually it wasn't much of a struggle at all. I saw where a few guys were running 300 bus speeds, with near-identicle hardware except the 3 people I found were all on air. Given, a really big aftermarket heatsink and fan for air, but still they were only on air! I knew I had them beat in the cooling department.

That's actually a pretty big concern of mine. Intel's new LGA775 specs now require the CPU's fan to cool (usually passing through the heatsink first, straight down) the VRMs and NB sections. Since I am water cooling, I do not have this air-cooling effect. Given I am removing the CPU's heat from the case completely with the WC setup so the case doesn't go as hot, the VRM and NB still aren't cooled with air though. This is why I would have liked to run a higher CPU speed then bus, as I don't have to push the voltages around other components as hard; therefore not needing as much cooling.

I'm thinking of getting some heat sinks for the caps around/on the VRMs and coils. Then for the NB/chipset I may water cool them if they go out or get flaky.

I left Prime95 running all night, no errors @ 4.2Ghz, 300 bus, and 6.9GB/s ram! Man, this is really nice.


Voltage, I'm not running that high. I actually have it set a bit high though @ 1.58v. I used to run at 1.55v @ 4.2Ghz with 233mhz bus speed on previous board. I was just turning up voltages trying to ensure it was stable. I'll lower it down to 1.55v today to see if it's stable.

Memory is at 2.1v, chipset is at 1.7v, FSB is at 1.4v. These is what others set in order to get 300FSB. I'll play with lowering them down over the weekend to keep a stable system. The bios shows these values as Pink, as a warning. Red values seem dangerous to me (i.e. the memory turns red in the bios when selecting 2.2v). I like the color-coded warning levels.

I had seen a lot of talk about 4GHz not really being that difficult on the stock HSF and I kept thinking...right, and it's probably 40°F in the room. Much to my surprise:

The boxed stock Intel HSF, 1.475Vc. In all fairness, I did have the case open and 120mm dangling by the wires over the HSF area.
And I'll link these since they are pretty large:
3.2 @ 4.0GHz
BH-5 XMS 3500...make sure to note the CAS timings I am going to HATE to get rid of these. I've hit the 6.35MB/s mark with them on the 2.8E/875P/285FSB!!

I wanted to try 1.525ish Vc, but it runs so hot on load right there (like 60C), I don't really want to turn up the Vc any more until my 775 WB gets here. I am pretty impressed with this so far. I am really contemplating ordering a Abit AA8XE to get around the 915P PCI-E float. I'd go for the P5AD2-E, but that's a smoking a price.

I know what you're saying on cooling things around the CPU with no HSF. I have my rad mounted on top of the case with 2x120s pulling in (1x80 rear and 2x80mm front = exhaust)...It's enough to get the majority of it and it hasn't been a problem on my 875 mobo, but I may need to revise something if it turns out to be a problem on the 915/925 board. Sinking everything would probably help...how much I don't really know. Can you stick a nice quiet 92 or 120 in the side panel over the CPU area? Something like the "OTES" cooling system on the Abit Fatal1ty would probably be awesome for cooling mosfets/caps during WCing. You can't see it in that pic, but there are 2 fans at the I/O port in the black shroud.

As much as the question has been asked (whether NB cooling is worthwhile), on my 875 Asus mobo (still in sig), I literally burned myself when I touched the NB once (unsinked). It's been WC'd ever since. At the very least, I would probably put an active HSF on it. I can't say this would or did do anything for OCing, but it was super hot and I wasn't really happy about it

Have you tried any faster on the DDR timings yet or is that as good as it gets at that speed? Just curious because I ordered 1GB of the same stuff this morning in anticipation of a 925XE upgrade sometime soon...