A Trip Back to the Past (LGA775)

[Berfs1]

Hello everyone, this is my first post (not sure if posting in right area, sorry), and I would like to start by telling you all of my recent overclocking experiments on the LGA775 platform. So, as everyone says, go with the highest multiplier that is possible, and then tune the base clock (FSB in this case) afterwards. And they are correct, or not... From my overclocks with a Pentium D 945 that I tested about a week ago showed some odd overclocking patterns. At the 17x multiplier, it would go up to around 4870 MHz, and I could not push it any farther. However, with the 15x multiplier, and a higher FSB (328.03 MHz), I was able to hit 4920.44 MHz, link: http://hwbot.org/submission/3880017_berfs1_cpu_frequency_pentium_d_945_4920.44_mhz I think this can reasonable conclude my earlier suspicions when overclocking a G3258, that certain multipliers do indeed work better, however I do not understand the science of it. About the G3258, I was able to hit a 47x multiplier, however the 46x multiplier would crash, EVEN AT SAME VOLTAGE! So, with my tests, I can confidently state that you guys need to test each multiplier and FSB combination before saying that is your best frequency! Oh I should also note that I only overclock for frequency, and thus the stability is extremely questionable (as in it will crash immediately under load). In addition, I primarily use my modded Gigabyte GA-P35-S3G, a wonderful and exceptional overclocking board despite the voltage limitations. This board is capable of hitting 500 MHz FSB on air (I have gone past 520 with an E6550). I am currently retesting my Pentium E6500K to see just how far I can get. Yesterday, I broke my previous personal best of 4341 MHz with a 4413 MHz submission. After I wake up, I shall push it even further with more tweaks to see how far it goes... Please let me know if you have any questions or comments, and thanks for reading!

 

[ClockHound]

Hello 2007! Thank you for making contact with the future.

No one is available to take your post just now - we are currently distracted by our multi-core soap-bar touchscreen 'smart' phones using 'apps'. We suggest you dispose of your old hobby tools and wait to be instructed when to deploy RGB lighting in your glass case.

;-)

Congrats on keeping the spirit of this forum going - CPUs and Overclocking. Makes me want to pull the old 6750 rig out of the archives and push the clock all over again.

 

[TheGiant]

From my overclocks with a Pentium D 945 that I tested about a week ago showed some odd overclocking patterns. At the 17x multiplier, it would go up to around 4870 MHz, and I could not push

so you can run pentium D, teh power room Prescott 2 cores room heater at nearly 5GHz?

well please if its doable experiment with some actual i5 or ryzens if it is doable

 

[TStep]

Kudos for being an old school tinkerer. It's been forever since I've tinkered w/ 775 stuff, but have you ruled out an FSB hole in your mobo/hardware combo? I had experiences where I could not run any multiplier in a small range of fsb speeds.

You can run at the 286 fsb w/ 17x but not 287. You know your cpu is capable of another 50mhz w/ 15x multi. Have you tried a 15x multi at 287 thru 328 fsb to rule out the fsb hole?

Also I never bought into running the highest multi possible. Unless you cannot lock your pci/agp bus speeds to keep them in spec, I was trying to run at the fastest fsb that maxed out my ram speed via the available ram dividers, then maxed out the cpu via increasing the multi. That way everything off cpu was as fast as it could be.

 

[SPBHM]

I remember the multiplier affecting OC and maybe performance on the Athlon XP nforce 2 days

 

[Charlie22911]

I remember the multiplier affecting OC and maybe performance on the Athlon XP nforce 2 days

Yup, I cross flashed a budget DFI board with the LanPaty bios to push 250X10 on a XP-M 2500+.
That was thanks to help from some guys and a bios modded at DFI-Street. Those were the days...

Also higher FSB is generally recommended first IIRC thanks to the memory controller being in the chipset, the CPU talks to RAM via the FSB.

 

[Berfs1]

Hello 2007! Thank you for making contact with the future.

No one is available to take your post just now - we are currently distracted by our multi-core soap-bar touchscreen 'smart' phones using 'apps'. We suggest you dispose of your old hobby tools and wait to be instructed when to deploy RGB lighting in your glass case.

;-)

Congrats on keeping the spirit of this forum going - CPUs and Overclocking. Makes me want to pull the old 6750 rig out of the archives and push the clock all over again.

lmao it's going to be put in a video soon

 

[Berfs1]

Kudos for being an old school tinkerer. It's been forever since I've tinkered w/ 775 stuff, but have you ruled out an FSB hole in your mobo/hardware combo? I had experiences where I could not run any multiplier in a small range of fsb speeds.

You can run at the 286 fsb w/ 17x but not 287. You know your cpu is capable of another 50mhz w/ 15x multi. Have you tried a 15x multi at 287 thru 328 fsb to rule out the fsb hole?

Also I never bought into running the highest multi possible. Unless you cannot lock your pci/agp bus speeds to keep them in spec, I was trying to run at the fastest fsb that maxed out my ram speed via the available ram dividers, then maxed out the cpu via increasing the multi. That way everything off cpu was as fast as it could be.

Oh, the PCI bus was never a problem (I ran it at a constant 100 MHz), however I can guarantee you the motherboard is completely stable even over 510 MHz FSB! It all depends on the actual CPU, in which I have a legit question: If my motherboard can do 500 MHz easily with one CPU, and another CPU cannot, will I need to raise the FSB frequency to increase the BCLK range of the worse CPU? Or would it not help? Thanks again for sharing your experiences!

 

[Berfs1]

so you can run pentium D, teh power room Prescott 2 cores room heater at nearly 5GHz?

well please if its doable experiment with some actual i5 or ryzens if it is doable

lmao it's definetly not stable, however I am currently re-attempting a Pentium E6500K overclock at the moment, and it seems to be quite a bad overclocker! (~4400 MHz with AIO and leafblower, and ICD thermal compound, at 1.6V)

EDIT: I should also note that I actually overclocked another person's i5-6600K once, you are looking at a max of ~4898 MHz on air cooling, a pretty bad one too (no offense to the owner), and with not so amazing thermal paste. When going for maximum frequency, I believe the LGA775 era still packs a hefty punch! Good luck beating the legendary Celeron D CPUs with an i7-8086K! (max overclocked frequency of course)

 

[naukkis]

Oh, the PCI bus was never a problem (I ran it at a constant 100 MHz), however I can guarantee you the motherboard is completely stable even over 510 MHz FSB! It all depends on the actual CPU, in which I have a legit question: If my motherboard can do 500 MHz easily with one CPU, and another CPU cannot, will I need to raise the FSB frequency to increase the BCLK range of the worse CPU? Or would it not help? Thanks again for sharing your experiences!

Dram controller have also multipliers, so you can't know if memory controller is running out of it's max stable frequency. Dram controller is clocked via straps, usually had 200, 266, 333 and 400MHz straps so chipset memory controller runs about same frequency with different fsb frequencies. So with 200 strap 250fsb clocks memory controller same as 500fsb would do with 400 strap. Some better mb's let you use what ever multiplier wanted to memory controller to clock memory controller as liked but those that don't can have fsb holes where memory controller is clocked too high and isn't stable, and higher fsb which engages lower multiplier to memory controller would still work fine.

 

[ClockHound]

lmao it's going to be put in a video soon

240p or it isn't valid! ;-)

 

[Berfs1]

Dram controller have also multipliers, so you can't know if memory controller is running out of it's max stable frequency. Dram controller is clocked via straps, usually had 200, 266, 333 and 400MHz straps so chipset memory controller runs about same frequency with different fsb frequencies. So with 200 strap 250fsb clocks memory controller same as 500fsb would do with 400 strap. Some better mb's let you use what ever multiplier wanted to memory controller to clock memory controller as liked but those that don't can have fsb holes where memory controller is clocked too high and isn't stable, and higher fsb which engages lower multiplier to memory controller would still work fine.

I have extensively tested the motherboard and multiple kits of RAM (another vid on that coming soon), and I was not RAM limited (my RAM can do up to ~1300 max), however with what you were saying about the live auto adjusting if the system wasn't stable, my MSI P43 Neo F did that, of course until it died (I suspect it is a dead motherboard although cannot say for sure, kept getting 00 post code, and I know my CPUs weren't dead).