Help with overclocking E5450 (core2 quad q9560)

[unikinqay]

Hi all

I need your advice on how to overclock xeon E5450 modified to fit LGA 775 socket. I'm not looking for very high OC, maybe 3600 Mhz or 4000 at most. What should I do? I don't know much about overclocking.

My old rigs specs are as follows:

CPU: Xeon E5450 3000 Mhz (333.3 Mhz x 9)
http://ark.intel.com/products/33083/Intel-Xeon-Processor-E5450-12M-Cache-3_00-GHz-1333-MHz-FSB
MB: Gigabyte GA-EP45-DS3
https://www.gigabyte.com/Motherboard/GA-EP45-DS3-rev-10#ov
RAM: 8 gb(4x2gb) Crucial DDR2 800 Mhz (without cooling ribs)
CPU Fan: Scythe Mugen 5 PCGH edition
http://www.scythe-eu.com/en/products/cpu-cooler/mugen-5-pcgh-edition.html

http://shrani.si/f/1X/kE/1uCVQ3Pn/1cpu.jpg
http://shrani.si/f/1R/rI/193sFnTS/1ram.jpg
http://shrani.si/f/2z/87/4YWsJG2n/1ram2.jpg


Will this specs allow me to OC or should I just forget about it? I have GTX 1050 TI and am getting a feeling CPU is bottlenecking it's performance quite a bit. I'll be buying new PC soon, but would like to wait until Coffee Lake and AMD APUs come out and then decide what to get. So in a meanwhile a little performance bust to my old rig wouldn't hurt🙂

 

[BFG-9000]

A Xeon overclocks just like any S775 quad (duals are far less limiting on FSB) in that the limit is mainly how much FSB the chip will take, especially as your multiplier of 9 isn't the highest. However there are some peculiarities:

In order to avoid any errata and an error message about an unsupported CPU on bootup, it is best to flash a BIOS with Xeon microcodes (generally these remove the Pentium 4 microcodes to make room but all the regular C2D and C2Q microcodes are still there). If you are too lazy to mod the BIOS yourself by inserting the microcodes (and have no need for any Fernando AHCI or SLIC type BIOS mod), premodded BIOSes are around--just google the motherboard name + "Xeon microcode."

After that it's just like any other S775 overclock--if you have good RAM it's limited by only the CPU.
In order to minimize stress on the memory while determining the CPU's limits it's best to set the RAM multiplier at one of the lowest 2.0 settings (I used 2.00D).
Leave everything else at auto and enable CPU Host Clock Control, then gradually raise CPU Host Frequency from 333.
If things start to get a little unstable, raise the CPU Vcore a little then try raising the FSB some more.
Eventually you should reach a maximum stable FSB somewhere around 400-450 without going over 1.45v Vcore.
If you do get all the way to 4.05GHz @ 450, the chip will be using in the neighborhood of 200W depending on the voltage needed to get there.

At this point you will test the memory latency and find it is shockingly poor, which is normal with a 45nm chip on a 40-series chipset even if you have DDR2-1066 5-5-5 or DDR2-800 4-4-4 (unless you run the memory at 6:5 or 5:4 which is pointless without an IGP considering the FSB bottleneck). The magic setting is under Advanced Timing Control under the individual Channel A and Channel B timing settings, and called Static tRead Value. If you can get it down to ~8 or so then latency will be as good as the defaults on 30-series chipsets. This may require some voltage increase for the MCH core and DRAM. If you have to run the DDR2-800 at 5-5-5 it's not the end of the world as latency is only 3-5ns worse, and 70ns is as good as a modern Ryzen with integrated memory controller and DDR4!

By the time of the 40-series chipsets at the end of the Core 2 era, most of the other BIOS settings were well ironed out at defaults, so aren't really worth tinkering with. This is really vintage overclocking and nothing at all like modern turbo multiplier-only overclocking.