Influence of L2 cache on FSX? (Or: Q9300 vs Q9450)

[Guest simjunkie]

Nick, would you work with me & my new Q9650-getting things dialed in? I can start another thread somewhere else. I've got it somewhat stable (so far 12 hours prime small FFTs) at 4GHz. I would really like your advice though if there's any polishing up I can do on my settings.-jk

[Guest Nick_N]

jkThis really is not a good time.. I dont even have time for bjoern right nowI saw you were hitting 445 and the voltages you posted looked good. I would not push the NB any higher than 1.52 without special cooling450 would be the sweet-spot as I recall a latency switch occurs there however your memory may hit a wall on that.The drill is lowest CAS, lowest tRD @ highest FSB @ 333 CPU Strap and memspeedwork voltages to stabilize, monitor temps and do not exceed memory manufacture max spec445 may be your wall because of the memory.. I think the proc may do 4050 as its probably the best quad on the market right now.4 sticks of memory is always harder on a clock than 2. Work the problem from the 450FSB mark, backwards.. if you already tried then more than likely the memory is simply stopping you. You can check that by setting 450FSB and raise memory timing to 6-6-6 to see if it boots, if it does then your hitting a memory wall at CAS5 on those sticks.If you find a memory wall is present then more than likely the fix would be to upgrade the memory to something designed to do the job at 2x2GB at CAS5 and a higher memspeed.. see this post:http://forums.avsim.net/dcboard.php?az=sho..._id=43731&page=

[Guest Nick_N]

>Nick, would you work with me & my new Q9650-getting things>dialed in? I can start another thread somewhere else. I've>got it somewhat stable (so far 12 hours prime small FFTs) at>4GHz. I would really like your advice though if there's any>polishing up I can do on my settings.>>-jk>>I did notice your Vcore seemed a bit low in that post1.35-1.40 max on that proc with a max temp under hard test load of 75c and shut down all the CPU internals too under the CPU list in the BIOS.. all DISABLEDalso set: CPU VOLTAGE DAMPER to ENABLEDsee if that helps

[Guest simjunkie]

I totally understand Nick, just thought I'd ask. We're all looking forward to GEX Eur and I know it's a tremendous effort.Things are actually good right now. When I said 12 hr prime that was when I stopped it, not because of any errors. My goal all along has been 4.0GHz and I'm there. Just need to find lowest stable voltages. Memory's pretty good too: 2x2G sticks of the Dominator 1066.I'll just keep reading because I know in spite of how busy you are, you'll be around.-jk

[Bjoern 1,153]

>I suggested the larger L2 higher mult proc because clock per>clock it

[Guest Nick_N]

>>I suggested the larger L2 higher mult proc because clock>per>>clock it

[Bjoern 1,153]

I guess the "L" made all the difference.I'm at 3.06 now and everything seems fine.Vcore is 1.24V, temps under load are 64

[Guest Nick_N]

>I guess the "L" made all the difference.>>>I'm at 3.06 now and everything seems fine.>Vcore is 1.24V, temps under load are 64

[Guest D17S]

There's still hope. Running that FSB above 400Mhz was always going to subject its users to stability issues: for the ram, the board, the CPU, even for those simply adjacent to the conversation . . . . but 400 will run. That's what all of these P35s (X38/43/35/38/et al) were made to do. At least learning the hard way is still learning. Seems your choices at this point is to trade the board for one that has a better chance of running a 450Mhz FSB. It will still be iffy. Ram will be pushed and the system's balance will always be questionable. For all this iffyness, you'll get a 7.5 x 450 = 3.3Ghz CPU.Or trade the CPU. Get the Q66 that will allow the board to run at its design spec of 400Mhz and a 9 x 400 = 3.6 CPU. The contention that a 45nm quad running at 3.3 will "wipe the floor" with a 65nm quad at 3.6 in FS gameplay is simply rediculous. Pushing that 7.5 multi with a 450FSB will continue to plague its users with repeated outbursts of clearly definable unstable behavior. Buyers beware.

[Guest Nick_N]

He doesn

[Guest D17S]

A 45nm Penryn at (430 x 7.5 = ) 3.2Ghz will be even Less able to compete with a 65nm at (400 x 9 = ) 3.6Ghz. Notice with the penryns, Intel increased the default FSB for those CPU. This had - Absolutely Nothing - to do with increasing performance. Once again, this had - Absolutely Nothing - to do with increasing performance. So what's the deal then? It had - entirely - to do with allowing Intel to Decrease the CPU's multiplier. OK, so why did they do that? 2 reasons. 1) Their lay customer base would "buy" that a "faster FSB equaled better performance."2) Their savvy customer base would NOT be able to get that "free ride" those old high multipliers provided.We can see this strategy successfully playing out right here. 1) "Faster/bigger/cooler is better" pre-sold the Q93. It had a uselessly bigger cache, ran with a uselessly faster FSB and temp is not a performance issue . . . but it pre-sold the CPU. I really tried, but the decision was made before the post was posted, It only needed a little encouragement.2) Now that the CPU is sold, the lower multiplier is capping the CPU's performance potential by means of the P35's (38/43/45/48's) topped-out FSB. If one has any doubt that Intel does not understand how this works, reconsider. Remember, they build these chipsets too. (Note the P58's FSB (Nehalem) will run at 133Mhz.)I have no doubt these Q9XX0s will go to 3.6, even 4.0 and beyond. However Intel purposefully limited these CPU's performance potential by lowering their non-adjustable multipliers. "Want more?" Intel asks. "Spend another $$$ our partner's super-deluxe mobos with our latest-and-greatest version of the P35, upgraded voltage regulation and other such stuff (. . . that - might - go above 400Mhz before it refused to even boot" they quietly muttered. "Gentlemen, the party is Over. You want speed? Pay Us.") Or a more savvy builder might consider their (pre-gen) 9x multi'd quad CPU and run a standard P35 (or a 975) mobo at its rated 400mhz FSB.There on the other hand, we need to give Intel a break. After all, they need to make money too . . . but let 'em have someone else's!

[Garys 209]

Sam why cant you just let this thread go. Your like the guy who has to have the last word after the other guy has already turned his back and walking away. Gary

[Guest D17S]

These are simply observations. Any construction beyond that must be sustained as the responsibility of the observer.

[irocx 292]

>Sam why cant you just let this thread go. Your like the guy>who has to have the last word after the other guy has already>turned his back and walking away. >>GaryGary, you're kidding correct? The "other guy" has made 12 posts since the "call" was "officially" made to "let this thread go" and that is being generous considering their post edits, what's been stated on record, and the fact that that "call" has been made by them several times now. Yes, it does seem to me the "last word" has been "gotten in"....... and them some. Is it halloween yet. Regards, Kendall#1: E8500 @ 3.95 - HyperTX2 Gigabyte P35-DS3L 4GB Ballistix Tracers PC6400 EVGA 8800GT - XG's 174.74 Seagate 250GB 7200.11 CH Yoke/Pedals/Saitek Throttle Dual Monitor: Dell 2405/1905 #2: Dell 8400 3.2 H.T. 3GB PC4200 - X800XT Diamond Xtreme/Logitech X-530's

[David Roch 1,959]

Sorry to lock this thread guys.Nick and Sam, your knowledge and help are really appreciated here but please, let's keep cool and stay away from personal attacks that are deemed inappropriate in this forum no matter who is right or wrong.Thanks.