Overclocking QX9770

[aigomorla]

Originally posted by: GuitarDaddy
I've yet to see one proven prime stable for 12/24hrs any higher than 4.2ghz.

How about a 10 hour? my final clocks are 4.25 @ 1.392Vcore :T

 

[bryanW1995]

ok guys, I have a confession to make. I bought 4 vapochills last night and stacked them on a Q6600 b3. well, actually one of them is on top, then there's one for each side of the q6600 except the part on the bottom (facing down towards the video card. my cpu won't boot to windows at over 5.2 ghz, is this normal? I spent good money for this cpu and I'll be damned if I'm going to settle for only 5.2 ghz. oh yeah, all of this is running an an ip35-e with 17gb of rdram.

 

[Owls]

Originally posted by: bryanW1995
ok guys, I have a confession to make. I bought 4 vapochills last night and stacked them on a Q6600 b3. well, actually one of them is on top, then there's one for each side of the q6600 except the part on the bottom (facing down towards the video card. my cpu won't boot to windows at over 5.2 ghz, is this normal? I spent good money for this cpu and I'll be damned if I'm going to settle for only 5.2 ghz. oh yeah, all of this is running an an ip35-e with 17gb of rdram.

If you read the freakin AT guide you would know you'd have to buy 2 vapochills per RDRAM, 3 alone for the PSU, and when you overclock the vapochill you need another vapochill for the vapochill. How could you be so stupid?! RTFM!

 

[BigMoosey74]

So back to the 1.79V....did it burn up yet? Lets just use this as an extreme case to see how long these things last over 1.5V

 

[aigomorla]

Originally posted by: BigMoosey74
So back to the 1.79V....did it burn up yet? Lets just use this as an extreme case to see how long these things last over 1.5V

I think you missed my post where i said i lost a Wolfdale. E8400 ES. @ 1.57Vcore in a span of 1 month loaded.

 

[AuDioFreaK39]

1.7875V on a one thousand, four hundred and twenty dollar processor.

Five words for the OP: LMFAO!

 

[PCTC2]

Originally posted by: aigomorla

Originally posted by: BigMoosey74
So back to the 1.79V....did it burn up yet? Lets just use this as an extreme case to see how long these things last over 1.5V

I think you missed my post where i said i lost a Wolfdale. E8400 ES. @ 1.57Vcore in a span of 1 month loaded.

I was once like that. I was enthusiastic and I ran the MAX OC on my E6400 @ 3.733 GHz for MONTHS at 1.55v. I'm surprised it lasted through those months, but now it is at 3.2GHz 1.375v. Listen to these guys when they say be patient and do it right. Plus, WTF do you need a 4+GHz PC for trading? I'm sure the stock would be more than supple for your purposes.

EDIT: I'm looking at a QX9770 now maybe for a new P5E3 Premium and my watercooling. Just go all out like Aigo and buy a QX.

 

[JohnVM]

Okay well, the sentiment is pretty clear about our overclocking job so far, and I believe Idontcare basically expressed out situation best. It's not like we're running around over here trying to be all cavalier and intentionally doing a shit job of overclocking because we think its funny -- we just are not experienced at doing this (we're finance guys not hardware experts) and want this system to run as fast as possible. In reality, the trading profits of squeezing an extra 10-20% out of this chip can make up for the cost of it in a single trade, so it's worth it for us to try, even at potentially dangerous voltage levels. I know people have referred to this as buying an enzo and crashing it, but it's actually a purely economic decision for us with rational thought processes behind it -- it's worth the risk and the chip is easily replaceable if it fails.

That said, it may not be necessary to "overclock like a jackass" the way we have been in order to achieve quality results. Long story short -- we really don't know what we're doing when it comes down to squeezing the most (or even a lot) out of this chip, and at no point have we ever tried to pretend or put forth the idea that we do. What would be *super* actually, would be if someone with a similar setup (from what I've seen Yoxxy may be the closest) would be willing to help us (in depth) with this overclock, which hopefully would: a) prevent (further) retardation damage by us, b) allow us to get the most (realistic) performance from this setup.

If anyone here is a really good overclocker and is in the Boston area, we'd love your help -- you could just come hang out and we could order pizza/beer/anything else you want (cohibas? patron platinum? fine cognacs?) and make a night of it, or we could just pay you, or some combination of both. Or, if anyone here is really good and would like to help online, we'd love that too. PM me if you're interested in any of this.

So to all the people who are really being assholes about this whole thing at least just try to put it in perspective. You can disagree with our methods so far, which I'm are highly flawed in many ways, but it'd be much more useful to post some helpful suggestions instead of just flaming the shit out of us for trying to get a nice overclock on a nice system despite not knowing a whole lot about overclocking. For the record, we did read the Anandtech guides on it, but were obviously still a bit unclear about precisely what to do (and we still are unclear now). To all the rest of you who have helped out so far -- we sincerely appreciate it and would appreciate continued help as we still need it; you guys have been really great.

Thanks a lot to all the helpful people so far.

 

[Markfw]

NOTE TO OTHERS IN THIS THREAD. ITS TIME TO GIVE THESE GUYS A BREAK, AND GIVE ADVICE, OR NOT PARTICIPATE. THIS IS POSTING AS A MOD.

Note to John, You guys really need to back down the vcore though, and look at other things to up, to get a good OC. I am not on the bleeding edge, so PM Yoxxy if you think he can help.

 

[Owls]

I must apologize for my posts. Seeing a good chip like that and on top of that a nice setup being misused kinda just messes with our hearts more than our minds (think classic cars). I (maybe we) understand your position though, from the beginning it seemed like as if you guys were throwing money at the problem instead of just taking a breath and stepping back.

Paying someone or having someone help set you guys up would probably be the best bet, you'd learn from a pro and see how to do things with your own eyes. For me that was always the best way to learn.

How fast can that QX run at 1.45-1.5v?

 

[aigomorla]

1. Consult RubyCon regarding Insane IO

2. Consult n00b 0f l337 in regards to your vapochill. You might be able to contact him faster on XtremeSystems.org/forums. You may want to have him upgrade it. As it is, its not going to keep up with your cpu @ the voltage im guessing you want to hit. He's in the jersy area i believe.

3. Get RID OF NVIDIA CHIPSET. You want FASTEST cpu and IO, that doesnt involve the fastest GPU. Jump on the DFI LT X48-T2R Platform.

4. Tell Nol EXACTLY what you did, and what sound you heard, on your VAPO to see if he can fix it or diagnose it. (he might give you a good stern lecture like we did on reading the manual, dont worry i get it from him too from time to time.)

Originally posted by: Owls
How fast can that QX run at 1.45-1.5v?

YMMV my friend. But @ those voltage, mine does around 4.6ghz :T

The OP may not have a good overclocking chip from the sound of things tho.

 

[nomagic]

Please excuse my ignorance. I feel ashamed to ask, but my curiosity got me. What kind of finance machine needs to be overclocked so much at the expense of stability?

Isn't it a consensus that for finance machines (or any mission critical machine), stability is far more important than speed?

If serious business is involved, wouldn't you want to rent or buy a super-computer? Even building a computer cluster would be a better choice depending on the application. Why build a enthusiastic machine for finance use?

"Reliable solutions from a trustworthy tech company" <- my 2cents...

 

[Markfw]

Originally posted by: nomagic
Please excuse my ignorance. I feel ashamed to ask, but my curiosity got me. What kind of finance machine needs to be overclocked so much at the expense of stability?

Isn't it a consensus that for finance machines (or any mission critical machine), stability is far more important than speed?

If serious business is involved, wouldn't you want to rent or buy a super-computer? Even building a computer cluster would be a better choice depending on the application. Why build a enthusiastic machine for finance use?

"Reliable solutions from a trustworthy tech company" <- my 2cents...

Good question !!!!

 

[PCTC2]

I would like to apologize for my post. I was just reminiscing about my first OC.

I would've loved to have helped you (I was in the Boston area) but I have now moved to La Jolla, CA. Sorry dudes. But I would definitely say keep the vCore down to 1.4 and just try a few things. I would agree with Aigo except I don't mind nVIDIA chipsets. I think it'll be fine for you.

 

[aigomorla]

Originally posted by: nomagic
Isn't it a consensus that for finance machines (or any mission critical machine), stability is far more important than speed?

If serious business is involved, wouldn't you want to rent or buy a super-computer? Even building a computer cluster would be a better choice depending on the application. Why build a enthusiastic machine for finance use?

"Reliable solutions from a trustworthy tech company" <- my 2cents...

*sigh* i brought this arguement up, so lets not continue with this:

Here is the reference: Pg.2
http://forums.anandtech.com/me..._key=y&keyword1=QX9770

 

[Idontcare]

Originally posted by: aigomorla

Originally posted by: nomagic
Isn't it a consensus that for finance machines (or any mission critical machine), stability is far more important than speed?

If serious business is involved, wouldn't you want to rent or buy a super-computer? Even building a computer cluster would be a better choice depending on the application. Why build a enthusiastic machine for finance use?

"Reliable solutions from a trustworthy tech company" <- my 2cents...

*sigh* i brought this arguement up, so lets not continue with this:

Here is the reference: Pg.2
http://forums.anandtech.com/me..._key=y&keyword1=QX9770

There are multiple financial transaction situations where latency is binary in determining whether you get the deal (or close the other end of the deal) or not.

Fo my system, which ultimately settled in at 3.73GHz for 24/7 100% loadings (B3 QX6700 on vapoLS) time was money. Yes downtime is money too, but you'd rather be up and making money at least some of the time versus being up all of the time but just a smidgen too slow to make money any of the time.

When you guys think "financial" your brains are really wired to envision "banks"...hedge funds, forex, securities traders, i.e. all the day-trading speculators out there don't need 99.99999% uptime. But we do need time to solution.

This is the program I use predominately - http://www.metaquotes.net/

All I can say is this is one of those things where if you haven't worn the shoes then you really can't expect yourself to have the insight or the experience to know the rationale or the motivations...Be it financial speculative trading or running a vapoLS rig.

@John & Justin - NOL is in your neck of the woods in terms of getting a custom-built vaporphase rig ($5k min) but since you already own a vaporphase LS I recommend PM'ing Jinu117 at XS forums and talk to him about getting your vapoLS upgraded (modded) to handle higher overclock heatloads of quads (which is precisely what is limiting your clocks at the moment).

Jinu117 mods them in the neighborhood of $250-$500 (so I'm told, never modded mine as I hit 4GHz just fine with stock vapoLS) depending on how much of an overhaul the system needs (just gases vs needing a new compressor, etc).

You really ought to be able to hit 5GHz 24/7 stable with that chip under phase cooling.

I don't live in Boston (I'm north of Phili) but am happy to help as much as I can thru the forums or by PM.

Edit: Meant to make this recommendation earlier - you need to buy a cheap quad CPU like the Q6600 to cut your teeth on (a sacrifical CPU you can trash) while coming up to speed with vapoLS rig.

I used a 65nm celeron (smithfield? forgot the codename) which took a lot of abuse from me for a couple weeks while I got comfortable tweaking my BIOS and knowing how to cleanly work around the chillicontrol limitations of the vapoLS rig.

A Q6600 would have been better but they weren't available for me at the time, for you guys it's $200 of insurance that I'd recommend and when overclocked it will "mimic" the heatload of your QX9770 albeit at lower clocks of course.

 

[Rambler358]

Okay, first post here so go easy guys! I have a QX9770 which I got a very good deal on. The mobo is a EVGA 780i FTW, and my system is water cooled. My PC, aside from normal PC activity - email, web surfing, etc. is mainly used for gaming and 24/7 Folding.

I've gotten 3.8GHz stable (IBT 9-passes) with a BIOS CPU voltage of 1.400v, and the FSB/VTT at 1.25v. I'm trying for stablity at 4.0GHz, but it doesn't look like I'll be able to do it with IBT (Intel Burn Test). It can pass OCCT and Prime95 though. My concern is the BIOS CPU voltage - it needs 1.50v and a FSB/VTT of 1.325v to get OCCT stable. At idle, my vcore is around 1.46v and under Folding load it's around 1.38v. Core temps stay under 50*C when folding. But is my CPU voltage (BIOS 1.50v) too high for 24/7 use and this CPU? Thanks in advance for any info!

 

[OCGuy]

Originally posted by: Rambler358
Okay, first post here so go easy guys! I have a QX9770 which I got a very good deal on. The mobo is a EVGA 780i FTW, and my system is water cooled. My PC, aside from normal PC activity - email, web surfing, etc. is mainly used for gaming and 24/7 Folding.

I've gotten 3.8GHz stable (IBT 9-passes) with a BIOS CPU voltage of 1.400v, and the FSB/VTT at 1.25v. I'm trying for stablity at 4.0GHz, but it doesn't look like I'll be able to do it with IBT (Intel Burn Test). It can pass OCCT and Prime95 though. My concern is the BIOS CPU voltage - it needs 1.50v and a FSB/VTT of 1.325v to get OCCT stable. At idle, my vcore is around 1.46v and under Folding load it's around 1.38v. Core temps stay under 50*C when folding. But is my CPU voltage (BIOS 1.50v) too high for 24/7 use and this CPU? Thanks in advance for any info!

Coulda made your own thread, we wouldnt have bit you!

Yes anything idling above 1.4v is the standard around here for being "too high." Intel warrants its 45nm off 1.3625V, meaning they are confident the chip will last 3 years + at that voltage.

1.4v Northbridge
1.4v VTT/FSB TERM
1.4V Vcore

Thats as high as I went on my 45nm quad........

 

[Rambler358]

Thanks for that info. Looks like it's back down to 3.8GHz for me then. It's just that I've been getting conflicting info on CPU voltage, but consistant info that FSB/VTT should never go above 1.40v and preferably be kept below that. Some have said to stay below 1.60v on the CPU in the BIOS, and others like yourself have stated 1.40v. Reading Intel's specs for this CPU, it does seem the 1.40v mark is where I need to be and not higher if I plan on keeping this chip for a while. Thanks again.

 

[OCGuy]

Originally posted by: Rambler358
Thanks for that info. Looks like it's back down to 3.8GHz for me then. It's just that I've been getting conflicting info on CPU voltage, but consistant info that FSB/VTT should never go above 1.40v and preferably be kept below that. Some have said to stay below 1.60v on the CPU in the BIOS, and others like yourself have stated 1.40v. Reading Intel's specs for this CPU, it does seem the 1.40v mark is where I need to be and not higher if I plan on keeping this chip for a while. Thanks again.

Its what actual voltage goes through your chip that matters. If you put 1.6 through a 45nm quad, that is a timebomb. Whoever told you that is trying to give you a $1400 keychain.

 

[bryanW1995]

yeah, even on the 65nm units 1.6v was too much. on a 45nm chip that's suicide.

 

[Rambler358]

Originally posted by: Ocguy31
Its what actual voltage goes through your chip that matters. If you put 1.6 through a 45nm quad, that is a timebomb. Whoever told you that is trying to give you a $1400 keychain.

Thanks again Ocguy. Yeah, I definately wouldn't put it at 1.6 but some have also said I'd be safe with 1.5. I'm just going to stay at a 3.8GHz OC and keep BIOS CPUv @ 1.40v. There's not too much of a difference between 3.8 and 4.0 from I've seen anyway. And although I paid no where near $1400 for this CPU, it was still enough for me not to want this chip as a keychain.

 

[CTho9305]

Originally posted by: Idontcare
{snip}

You seem to be a great person, but you of all people should know better - I'm going to laugh my butt off if your OC'd CPU ever gives an incorrect result costing you millions of dollars

 

[Idontcare]

Originally posted by: CTho9305

Originally posted by: Idontcare
{snip}

You seem to be a great person, but you of all people should know better - I'm going to laugh my butt off if your OC'd CPU ever gives an incorrect result costing you millions of dollars

Even stock clocked chips/systems can give you errant results.

OC'ing increases the probability of encountering an error but not overclocking does not reduce that probability to zero.

In my business no one should ever rely on a single-instance of computed results without verification. I overclock to get results that are "time-to-results" sensitive. I verify by duplication of runs, both on the originating system as well as on independent systems.

I spend the majority of my time developing methods which are increasingly insensitive to those critical time-to-result situations. But in an imperfect world the imperfect solution can sometimes be plenty good enough.

And for the record I don't need errant computations to lose a million bucks, I seem quite capable of doing that with my seemingly correct calculations to-date

 

[v8envy]

So is the target software single threaded and you're using a quad to simply run multiple instances? In that case I'm guessing you might get better single core performance out of an E8600, and more machines.

If the app can be distributed you could just... get more cores. Lots more cores. Think cluster here.

And how does computational latency compare with network latency? Is the goal to reduce a 5 minute computation to 4 minutes 30 seconds, or is the goal shaving tens of milliseconds off of a practically instant computation?

Independent currency arbitrage twitch trading seems like a great way to get ulcers and a heart attack in record time either way.