Archived

This topic is now archived and is closed to further replies.

Guest simjunkie

Q9650 & chance at 4GHz vs Q6600 @ 3.6GHz

Recommended Posts

I have a Q6600 running at 3.6GHZ (9 x 400) and am thinking about getting a new Q9650, E0 stepping, that has been shown to easily run at very close to if not at 4 GHz. With best case senerio, how much would that improve things for FSX?-jk

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

Tell ya what, I'm goin for it. I think the Q9650 just plain beats the Q6600 anyway and it will get me through the initial Nellie shakedown period and well into next year.-J

Share this post


Link to post
Share on other sites

At identical slider settings, CPU clock v FPS runs about 1:1. So 4.0/3.6 = 1.11. Meaning you'll get about an 11% increase in FPS with identical sliders. The penryn claims a very slight clock for clock advantage. So it might be wise to give it 15% as a generous concession. Or you could trade back some of that 15% FPS increase for another notch of AG or a bit of additional traffic. It won't be magic, but it Will help. The question is (as always) how much is a user willing to spend for that last 10%. That's not a judgment in the process of making, only an observation of the facts at hand. Speed costs, How fast do you want to go? Increased FSB won't help at all. Memory speed assistance is subjective, at best. You're on the right track cuz clock increases are the best bang for the buck. The real situation is that there's not much bang for Any buck to be had.

Share this post


Link to post
Share on other sites

If the money isn't a big deal, I'd go for it.Sams right it won't be a huge performance increase, I can speak from experience between 3.6 an 4.0 on a Penryn...but it will be a little bump.RhettFS box: E8500 (@ 3.80 ghz), AC Freezer 7 Pro, ASUS P5E3 Premium, BFG 8800GTX 756 (nVidia 169 WHQL), 4gb DDR3 1600 Patriot Cas7 7-7-7-20 (2T), PC Power 750, WD 150gb 10000rpm Raptor, Seagate 500gb, Silverstone TJ09 case, Vista Ultimate 64ASX Client: AMD 3700+ (@ 2.6 ghz), 7800GT

Share this post


Link to post
Share on other sites

I don't expect a huge difference. I'm after improvements and not miracles. But I would be much happier with a native 3.0GHz penryn as opposed to a 2.4GHz Q6600. With my settings I hang at 30 fps and dip down from there. If I can reduce the "dips" a little then I'll reach my goal with this chip.-jk

Share this post


Link to post
Share on other sites

I upgraded like you plan, and would do it again, but in truth FSX needs much more still.Too many areas remain that bring the framerate below acceptable.Its just an FSX problem IMO.Example: Fly Tampas Hong Kong looks amazing and frames are very highthen I go to MS's Tokyo and watch a slide show if settings are up.Instead of giving JS 10 million for 2 bad commercials, money might have been better spent improving FSX :)

Share this post


Link to post
Share on other sites

LOL! What crap, huh? Too much $$$ in advertising, not enough in the actual product.I got it in and it's at 3.6GHz (9 x 400) @ 1.312v BIOS (cpuz 1.28v idle, 1.24v load). Only just over 2 hours prime stable but I'm going to a higher clock and will Prime a minimum of 12 hours before I settle on stable enough for simming. I flew at 3.6GHZ (same as the Q6600) and no doubt at the same airport I fly out of (San Diego intl) I stay at 30fps (locked) for a higher percentage of time. The real trick I think will be to get my memory settings dialed in for low tRD, I'm shootin for tRD6 at around DDR21066, CAS5. -jk

Share this post


Link to post
Share on other sites

It's in. It's running at 3.6GHz @ 1.312v BIOS and 1.24v CPUZ under Prime95 load (loadline calibration disabled). Don't know how stable it is yet, only primed for 2 hours but no errors yet. More later.-jk

Share this post


Link to post
Share on other sites

3.99GHz, 1.325v BIOS, 1.248v CPUZ under load and close to 3 hours Prime95 stable. Hottest core under 60 deg.Gotta run Prime 12 hours to get some more confidence but these results so far are better than I thought! -jk

Share this post


Link to post
Share on other sites

Just finished a 12 hr Prime95 run. 4.02GHz (9 x 445), 1.325v BIOS (1.248v load),VTT is 1.25v, NB 1.55v, 62 deg highest temp on hottest core. 333MHz strap and Memory's running DDR2 1069 5-5-5-15, tRD7. Now just gotta go in and reduce voltages to lowest I can get.This chip rocks. -jk

Share this post


Link to post
Share on other sites

>Prime a minimum of 12>hours before I settle on stable enough for simming. >I hear you. Last night I was flying a short flight, KORD to KSTL, and my overclocked sys hardlocked near Joliet on climbout. Nothing more aggravating than losing a flight like that, and I thought "!@#!@ I was just about to press the ; key" to save the flight.I guess 1.26v @ 3.8 ghz isn't quite enough voltage for an E8500. Or it could be the memory too, I'll figure it out.If I had some sense I would have Prime'ed those settings for many hours as you recommend...RhettFS box: E8500 (@ 3.80 ghz), AC Freezer 7 Pro, ASUS P5E3 Premium, BFG 8800GTX 756 (nVidia 169 WHQL), 4gb DDR3 1600 Patriot Cas7 7-7-7-20 (2T), PC Power 750, WD 150gb 10000rpm Raptor, Seagate 500gb, Silverstone TJ09 case, Vista Ultimate 64ASX Client: AMD 3700+ (@ 2.6 ghz), 7800GT

Share this post


Link to post
Share on other sites

I make sure I run at least 12 hours of the small FFT and the blend test too. I also run Memtest86+ for at least 12 hours. 24 hours is what most would probably consider "stable". I just hate to put that much stress on my parts for that period of time so I run 12 hours. I used to crash out of the sim a lot before I did those longer prime runs.-jk

Share this post


Link to post
Share on other sites

>I guess 1.26v @ 3.8 ghz isn't quite enough voltage for an E8500. Or >it could be the memory too, I'll figure it out.I had to give mine 1.280 in the Bios which shows 1.272 idling in Vista with Loadline Calibration disabled for 3.8. It probably needs 1.36 or more for 4.0 but I've haven't tried it yet.3.8 is plenty fast for now. Regards,Jim Karn

Share this post


Link to post
Share on other sites

I'm at 1.325 in the BIOS to get 1.240 - 1.248 under load in cpu-z. I wonder if I should raise the Vcore up a notch.-jk

Share this post


Link to post
Share on other sites

>I'm at 1.325 in the BIOS to get 1.240 - 1.248 under load in>cpu-z. I wonder if I should raise the Vcore up a notch.>>-jk>Turn on the CPU VOLTAGE DAMPER which should stablize the vcore and deliver what you select in the BIOS with very little drop in WindowsLoad line droop or Vdroop is a normal function of an Intel system. It helps reduce the output surge or spike that results from fast load changes and its fine under normal conditions however in a clock it can present issues which is why they give you the DAMPER setting in the BIOS to reduce its influence on the CPU. Max vcore on that slug would be 1.35-1.4v and I would not exceed 1.4 with a P Quad even if the temps are cool. ENABLE the CPU VOLTAGE DAMPER and disable all of the CPU power manaagement fucntions and you should find a much more stable environment for clocking and be able to get that last 5GHz you were looking for, perhaps more.. but watch the temps1.35 should do nicely, perhaps a touch more. Enable the damper first and disable all the CPU power management junk under the CPU menu in the BIOS @ 1.32v Vcore and see if it allows a bit better stability for a touch more clock before trying a higher Vcore.You should then notice 1.32v in the BIOS = 1.32 (or close) in Windows

Share this post


Link to post
Share on other sites

'disable all of the CPU power manaagement fucntions'Nick, does that include 'execute disable bit' and 'TM monitor'? I already have C1E and EIST disabled as well as virtualization tech and vanderpool. All those disabled?I was thinking of using the loadline calibration like I had on the Q6600 but I heard from Anandtech that my asus board had a problem with that and 45nm cpu's. But I do have the latest BIOS and I'll go in & turn it on and set it as you advised. I don't like all that vDroop.Thanks! :)-jkEdit: OK, loadline works really well. Just set 1.318 BIOS and now very little change under load, but things started getting too warm under load (high 60's C). Set 1.300 and it reads 1.288 cpu-z under both idle & Prime95 load, but still a little warm (mid 60's). Only around 0.01-0.03v droop nowI ran small FFTs and blend for 12 hours each over the last two days at 1.248v cpu-z full load(1.325 BIOS setting) before I stopped the test (no errors) but that was without loadline enabled. So I'm thinking now: how does 1.29v BIOS for around a cpu-z reading of around 1.27v full Prime load. Then stress test for 16-20 hours.Does that sound too low?

Share this post


Link to post
Share on other sites

What are you using to check the temp of the proc?Are you sure you are looking at the core temp and not the case temp of the proc? As long as you have verified you are looking at the core temp diode reading, which is usually 10-13c higher than the case temp though software like CoreTemp, then your next step is to locate the Tjunction Max on that slug.If tjM is in example 100c then your max temp under a prime load for hours should not exceed 80c and that would be totally acceptable for a STRESS TEST check. You will never heat the proc up using FSX or normal use like you will with prime or a stress test software such as ORTHOS. Therefore as long as you remain 20c under tjM for the slug in such tests.. you are fine and have found the max stable point where the proc will function without perf loss. I usually keep it under 25 but that

Share this post


Link to post
Share on other sites

I use Realtemp v2.70. I also have Everest 4.6 If you think that's more accurate. I just finished 8 hours prime @ 1.272v Vcore as read under load in cpu-z, no errors. Max temp on hottest core read 58 deg in Realtemp. But the Tjmax in Realtemp shows as 95 deg, which is 5 less than what the real Tjmax for this slug is, correct?So I add 5 deg to my readings to get 63 deg max, is that right? Do you think I should use Everest to monitor temps?other readings in BIOS are Vtt @ 1.26v, CPU PLL @ 1.59v, NB @ 1.52v -jkEdit: I can't believe how good this chip seems to be. 4GHz under 1.3v? WOW!!!!

Share this post


Link to post
Share on other sites

>I use Realtemp v2.70. I also have Everest 4.6 If you think>that's more accurate. I just finished 8 hours prime @ 1.272v>Vcore as read under load in cpu-z, no errors. Max temp on>hottest core read 58 deg in Realtemp. But the Tjmax in>Realtemp shows as 95 deg, which is 5 less than what the real>Tjmax for this slug is, correct?>>So I add 5 deg to my readings to get 63 deg max, is that>right? Do you think I should use Everest to monitor temps?>>other readings in BIOS are Vtt @ 1.26v, CPU PLL @ 1.59v, NB @>1.52v >>-jk>>Edit: I can't believe how good this chip seems to be. 4GHz>under 1.3v? WOW!!!!Yes.. its a great slug which is why the 500 dollar price tag. Intel knows thatRealTemp and CoreTemp typically show 2 different values for tjM and Everest I would not trust completely. Lets assume its 95.. and I thought it was 100 but I would need to verify that to be sure@ tjM 95, 70-75c is your max 4-core prime stress test load temp... completely safe and fineso if you are trying to get to 450Mhz from 445, you have room to work in VcoreCheck it in CoreTemp and see what it reads for tjM and the cores

Share this post


Link to post
Share on other sites

.... and do keep in mind that a 4 core stress test will run that proc about 10-15c hotter than your most demanding applications. There can be some exceptions however you should never, ever see the same temp running FSX an other typcial load type apps as you do under a 4-core stress test load assuming the HSF is efficient and working correctly.

Share this post


Link to post
Share on other sites

and the way you are running that slug beats the living snot out of the Q6600 which is why when I saw your memory ability and you mentioned you wanted more.. I said if you can get into the 450Mhz range to enjoy the upgrade.;-)

Share this post


Link to post
Share on other sites

I'll go for 450-455 next and try to leave it at that, I gotta start flying! I have a TRUE too but right now I'm using the OCZ Vendetta 2 which runs with the TRUE according to my temp readings over the time I've had it. I also modified the Vendetta's push-pin mount to use the bolt-through/backplate parts from the TRUE. Haven't lapped anything yet, that's a great Saturday project soon. My Q66 was lapped and my TRUE was too but the Vendetta seems to cool just as well & it weighs less.Another great thing is that according to my UPS power usage meter, in my clock I'm using around 45 less watts with the Q9650 than I used with the Q6600, at least at idle.Cheers!-jk

Share this post


Link to post
Share on other sites

I could tell you were on the ball with this... good jobWith most Asus BIOSes 450 marks another change on the NB so anywhere between 450 and 463 would be the hard sweetspot for your rig.

Share this post


Link to post
Share on other sites

and one last tid-bitFSX is a memory subsystems worst nightmare... just because you are stable with prime does not mean you will be 100% stable in FSX. If you have reached the limit and edge on that memory you may see some initial lockups or glitches that require a trim. Reduce FSB (last resort if @ 450), raise vcore a touch, NB voltage, or mem voltage a touch and with 4GB a touch of clock skew or even some VTT tweaks may be needed however usually if its stable like you are testing, its good.. just be aware that if the memory is at its limits it may show fine in prime and other stress tests and glitch in FSX after or during a long flight in hard hitting scenery.You can tell by process of elimination if you are hitting the memory wall or the CPU wall.. this is one of the reasons I always purchase memory with the overhead built into it. That way I know what wall I am hitting quickly and do not need to push the limit... worth the investment when you see the end results and spend less time in setup over just plain bliss and happy times in FSX.Good luck

Share this post


Link to post
Share on other sites

> Last night I was flying a short flight, KORD to>KSTL, and my overclocked sys hardlocked near Joliet on>climbout. Nothing more aggravating than losing a flight like>that, and I thought "!@#!@ I was just about to press the ;>key" to save the flight.That's why I never O/C and fly without AutoSave. Google on AutoSave.zip for Flight Simulator. (I don't think there's an FSX version, though.)

Share this post


Link to post
Share on other sites