Archived

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

Noel

Best available guide for o'clocking P5E3 Premium (BIOS 0401)??

Recommended Posts

I am setting up a new mainboard and memory. There are lots of BIOS o'clocking settings on this X48 board that I am not familiar enough with to use both effectively and safely. Can anyone recommend a comprehensive guide on using some of the various controls?For example, it appears "Common Performance" (shows up when you enable Transaction Booster) must be MCH Read Delay or tRD, but nowhere can I find out what these esoteric control truly mean. Thank you kindly in advance,Noel

Share this post


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

I have just upgraded to p5kse with dual core 6750.This board has the older p35 chipset,however the principles should be the same. The first thing to look at is the multiplier; This should be set to the max for the cpu you are using. If not you may be able to speed things up by increasing the multiplier. WARNING MAKE SURE YOU HAVE SUITABLE COOLING. In my case the multiplier was already at the max (8)and so I went to the next step and increased the front side bus freq. Currently it is running at 400, up from 333. This increased the cpu from 2.66 ghz to 3.22 ghz. Am waiting to get some more cooling before pushing on. If you search wikipedia you will find a number of sites that will give you ideas on how and how much. Have fun.:) :)

Share this post


Link to post
Share on other sites

On the tRD front, it's the biggest memory performance enhancement easily available. At a 400Mhz FSB, it decreases latency by 2.5ns per increment. Once you're setup, get Memset. Menset's 'Performance Level' setting is tRD. You can change it in real time. Get Everest to monitor your memory latency. Keep reducing Memset's Performance Level it 'til the system locks, then back off a notch. With the x48, you'll likely be able to shave off 10-15nm. To put this in perspective, your out-of-the-box latency will be ~ 60ns. A drop to 45ns is a full 25%. See if you can see any difference at all with this definitive 25% increase in memory performance. Put on your imagination hat, cuz it's gonna take a poised eye.This experiment will describe why I'm not overly excited about Nehalem's memory subsystem optimizations. If 25% is virtually impossible to see, then 50% finally becomes minutely observable? We're getting extremely close to a zero latency here. It seems to me that latency is not a major player. Quickpath and that onboard memory controller are production ehnanCers, not performance enhanCers.

Share this post


Link to post
Share on other sites

If 25% is virtually>impossible to see, then 50% finally becomes minutely>observable? This is where we, subjectively, differ. I decided to go this road after seeing near 100% perfection in FSX when I had latency down to around 52ns using Everest. But, it IS subjective, tho focusing on all the elements of performance that seem to matter.After reading and experimenting, with the mushkin DDR3 1600, I am finding decent performance (in Everest) setting a WHOLE HOST of detailed timings, complements of mushkin support dude Greg. The modules are rated 7-7-6-18 2T, 1.85-1.90v, and they will do 1600Mhz, but surprise, at COMMAND RATE of 1T, tRD 6. It's iffy at 1600Mhz, but if I dial back, to 1580Mhz, she's stable. Everest latency is down to 48.7ns with totally low stress settings voltage wise.I have FSX installed, but my entire new platform is totally fragged, and so needs to go thru some defrag and a few other tricks to optimize everything before we can experiment some more.I did fire FSX up briefly last night after installing and configuring everything, and noticed the dang sound popping again! This never happened with the first P5E I used. It did with the RMA'd replacement. The main common component at this point is my QX9650. I have a hard time believing my CPU could be having issues, cuz it clocks to 4.3Ghz no problem still, and runs Prime95 error free. I will have to dl latest audio drivers and see what gives.

Share this post


Link to post
Share on other sites

Not much disagreement there. And BTW, that mid 40ns latency is just about where Nehalem's Quickpath / onboard memory controller is now. If anyone's looking for Nehalem's latency miracle, look to the agreement on subjective performance assessment that is occurring right here. It's going to be a - subjective - miracle. We are Actually going to get whatever clock they can drag out of the same 45nm core as the Penryn, + whatever data path optimizations they can engineer. Another 20% is my guess. The Q9xxx is going to end up looking like the right choice for the next couple of years. I'm liking my Q66 better and better . . . if I could just get my sound settled down. It appears my sound problem only occurs at FSB speeds above 375. 400 just plays havoc with my audio. All the P35s (X-yx, et al) are certified to 400Mhz, but that speed is frying my audio . . . and it's getting worse. Locking PCI-e at 100Mhz did not help. The RMA board did the same thing. An addin audio card did the same thing. This must be some kind of a buss <> buss comm link problem. It feels like a power problem, but my Kill-a-Watt shows only ~ 300 watts at load. That Silverstone 600 should be fine there. All high-load video runs are fine. Try turning down your FSB to 350 or so. Get your clocks with da multi and da "divider." See if that helps. Remember: Nehalem only needs a 133Mhz FSB. By jove, I think they're on to something.

Share this post


Link to post
Share on other sites

if I could just get my sound settled down. >>It appears my sound problem only occurs at FSB speeds above>375. 400 just plays havoc with my audio. All the P35s (X-yx,>et al) are certified to 400Mhz, but that speed is frying my>audio . . . and it's getting worse. Sam I had this issue with the RMA'd P5E, and dang my new P5E3 Prem WiFi. Even worse! It was the frickin video driver. I had dl'd the latest WHQL and once that was removed and residuals cleaned out and some olders ones installed all is well again.

Share this post


Link to post
Share on other sites

Well maybe, but this is coming up from scratch. I finally decided to take it down to an X'ed out, formatted HD. Vista 64, then SP1, then the drivers. There was never a Vdriver installed to begin with. It actually appears to be CPU related as well. I dropped the multi to 8, then ran the FSB at 400 (CPU at 3.2Ghz). The audio was fine. Started ramping up the FSB. At 425, the audio fragged again. (CPU at 3.4). I then reset the multi to 9 and the FSB to 375 so the CPU was still at 3.4. Audio runs fine. The FSB is OK at 400 with the CPU at 3.4, but NoT OK with the CPU at 3.6. Go figure. With the current 9 x 375 set, I can run Everest stress test (CPU at 100%) with WMP running in it's full-screen'd stonner-scope mode on the 42'er and the audio is clean as a whistle. Whah dude. What a show. I just wonder if CPUs might show a little wear and tear after higher volt O/Cs? All was fine, then it started getting worse and worse. It's OK at 375 x 9 right now. The ram-bois now have 900Mhz available and I can get latency's in the high 50s (not that it matters, but it's fun to play with). I thought those fancy-pants X models were super-duper. Even this plebeian P35 gets tRD6. That tRD is magic. Com' on. Show us some stuff. ^^ BTW, what driver did you end up with. It's certainly worth a try.

Share this post


Link to post
Share on other sites

Hi Noel,Putting a system together with the exact same board as you. Have a 9770 with the Corsair - CM3x2048-1600C7DHXIN-4GB. They have 7-7-7-20 listed at 2T running 1600. Hoping these run at 1T. These were $675.00 from NE and they did not have the Muskin modules when I got them. Will probably Exchange them if they do not run 1T or lower then 7-7-7-20. Going to start out on air before I switch back to H2O so will try to give you info on how I make out.Thanks,Ted

Share this post


Link to post
Share on other sites

>I just wonder if CPUs might show a little wear and tear after>higher volt O/Cs? All was fine, then it started getting worse>and worse. It's OK at 375 x 9 right now. The ram-bois now have>900Mhz available and I can get latency's in the high 50s (not>that it matters, but it's fun to play with). I thought those>fancy-pants X models were super-duper. Even this plebeian P35>gets tRD6. That tRD is magic. Com' on. Show us some stuff. >>^^ BTW, what driver did you end up with. It's certainly worth>a try.169.21 WHQL, but for 32bit . . .Yeah, one wonders. What I'm having trouble buying into is that my CPU might be showing wear n tear (leading to audio crackles--NOT!, or inability to tolerate a higher FSB, despite the multiplier down-adjustment), and yet it still clocks to 4.3+ with perfect prime95 tolerance. Doesn't make any sense to me! Yet here is Derek's (Corsair) take about stress on CPUs, which I'm buying into:"The 45nm chips are more of an anomaly than any other Intel made CPU. The FSB gives out before the thermals do. In other words, the fail of overclock is not due to great thermal output or voltage need. It has been found that the 45nm needs far less voltage to get where it is going, but far greater VTT and PLL voltages and it has also been shown that if you raise VTT too high, drastic issues of CPU longevity can result, leading to CPU damage. There are new electrical characteristics in the 45nm processing that are just now being documented because they are anomalistic. To put it bluntly, we just do not know the long term results or even the characteristics of the electrical reactions. Suffice it to say, 45nm product is being damaged with low voltages and high FSB's. Triply so with quad 45nm product which leads me to believe that the effect is both cumulative (multi-core additions) and degenerative.I personally would not go to the 500Mhz FSB. I would go to 400 --> 425 --> 435 --> 445 --> 450 and stop there. Keep in mind that 450Mhz will give you 1900Mhz DRAM but even that is not a given. Most are reaching 400, some not. Many are reaching 425 but more are not. Most are not reaching 450 but some are."Sam, right now my primary goal is to NOT have to rebuild another system for a while. I am using the ASUS Oclock profiles to retain a fastest/stable and a moderate/stable option. My CPU on this P5E3 Prem is not stable beyond an FSB of about 400 at voltages I am comfy with. I'm even seeing an NB voltage bump over 1.40 puts alot of heat. And there is comment that quad cores and 4GB adds NB stress.Here are my current two profiles:1. For low end computing: FSB 333, DDR3 at 1333, 6-6-5-12 1T, tRD of 5. With a CPU clock of 3.33Ghz, I get latency score of 54.5ns. My ram is volted at 1.65v for this slovenly pace.2. For FSX: FSB 394, DDR3 @1580, 7-7-6-16 1T, tRD of 6. With a CPU clock of 3.94Ghz, I get latency of 49.5ns vDIMM 1.84 in the BIOS (allegedly that will trans to 1.86 with a voltmeter on a P5E3 Prem), NB 1.34, vCORE 1.375 works. If I go much higher on my FSB I will need more voltage so why bother. Briefly, I did go to 420FSB and latency drops down to 46.7ns +/-. But I just don't have the stomach nor time to have to keep this up. I'd also like my QX to last long enough to benefit from better software use of multicore. We can only hope.So far this is looking great. Performance in FSX is excellent. As I said before, my DDR2 story prior to breakdown was also excellent. From what I can tell here, my components should be experiencing less stress than with the DDR2 whacked to where it was. QX9650 w/ Retail HSF|8GB Muskin PC-6400|ASUS P5E|EVGA 8800GT @700|Seagate SATA 2 x 4|Seagate Cheetah 15K.x|XP Pro|Vista 64--soon to be installed

Share this post


Link to post
Share on other sites

Just finished getting this squared away using BIOS 505. Works like a charm, if you still need assistance, let me know.Rob.

Share this post


Link to post
Share on other sites

I thought those>fancy-pants X models were super-duper. Even this plebeian P35>gets tRD6. That tRD is magic. Com' on. Show us some stuff. From my understanding, these fancy-pants X models overclock really well, and mine certainly does, plus gives ultimate flexibility when combined with optimal memory settings. You pay (thru the noz) for this, but I think it's true.My current burned in main memory settings are: 7-6-6-18 1T @ 1596Mhz, tRD 6. I have NB volts to 1.39, memory to 1.88v, and she runs great at 4.18Ghz. This is a stable and comfy configuration, I think. Runs cool with my cooling solution. I'm sure it would go beyond this, but the (Head + Stability):Performance ratio doesn't improve to the point there is a purpose, beyond braggin' rights as you say. And always, braggin' rights don't last long in this biz. FS is running gr8, and I need to try SP2 to see how that behaves. Near as I can tell, I can't just increase tRD anywhere. She won't boot with tRD of 5 as it stands now. I don't know if megavolts would change. In any case, it ain't worth it for braggin' rights! She does run at 420FSB, but I'm afraid that takes too much stress . . .

Share this post


Link to post
Share on other sites

Well actually it turned out to be a very informative ride . . . but what a ride it was! It was especially interesting to dig in and start to understand how this stuff works. I still envision that little NewEgg warehouse along side the Burlington Northern spur line . . . trying to sync a freight transfer through that single loading dock door. See, I told ya that unless I could get this so my (inner) 10 year old brain could understand it, I wouldn't have a chance. Now I get it, well kinda.So the clock got the rig to 4.0+. That really will help. On to the rest. I can get 60ns with my old 800/44412 at a tRD of 6. In my little world of understanding, that latency number is the whole, - EnTiRe - enchilada. Everything else (2000Mhz bandwidth, CALWI timings, et AL) is - ONLY - only about making those onload/offload events consecutively happen as often as possible. That is IT. The FSB can hit WarP speed. Memory buss can run beyond the Speed-of-Thought. But unless that bit gets transferred, NothIng happens. It's the RATE at which these Bits get Transferred (RBT) that counts. (Yes, I made that up@! Like it?) AKA "Latency." Synchronicity did Not end up being just a generic term, but The primary property that allowed this event to occur. All of a sudden, tRD started to make sense. Now look around, tRD is used for this and that and all sorts of things. Sometimes I just don't trust education, especially self-proclaimed. So, I understand maxo rigset decreased that latency number to ~ 45ns. If that is so, then this entire drill was about ~ 15ns from plan ol' DDR2-800. And this cleared up the stutters and the fuzzies and fixed all manner of system deficiencies on a 3.6Ghz quad rig. I think I remember an initial comment about this reflecting, "Something else is going on here." But in any case, I don't care How much further up the river Louis and Clarke got, it was still a trail yet blazed. Pizza for all!

Share this post


Link to post
Share on other sites

>another system for a while. I am using the ASUS Oclock>profiles to retain a fastest/stable and a moderate/stable>option. My CPU on this P5E3 Prem is not stable beyond an FSB>of about 400 at voltages I am comfy with. I'm even seeing an>NB voltage bump over 1.40 puts alot of heat. And there is>comment that quad cores and 4GB adds NB stress.>FYI with my P5E3 Premium I am now running 400 x 9.5 with 1.30v on the cpu. I have not touched PLL voltage at all.Without doing that voltage bump on the cpu, 400 x 9.5 was not stable on my machine.I run DDR3 at 1600; again no voltage tweaks on memory either.I am considering using the XMP profile for my memory but have not (that profile goes 7-7-7-20 2T at 1.9v, and the 1.9 seemed like a lot to me).RealTemp reports temp on the cores to max out in Prime95 or FSX hammer-sessions at 62degC at the present clock which seems to me to be quite nice. 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