agent_x007 wrote:Ad. 1. Because 40%+ clock difference per core is noticable almost everytime (E5450 @ <2,4GHz vs. E5400 @ ~3,5GHz). Not sure what you plan to do, but 12MB cache (or 6MB per two cores to be precise), isn't enough to overcome that big of a clock gap (and E5400/E5500 shouldn't have much problems with ~3,5GHz clock, even on these boards).
You are comparing E5450 with overclocked E5400 (because stock is 2.7 GHz) so you probably assume that that 800 FSB is not problem to increase to 1066 to reach 3.5 GHz, ok that could be good idea.. About L2 cache afaik its not per core its shared between all of them and E5400 has only 2 MB and Xeon 12 MB, its big difference. Before i will check numbers, for me are all of these cpus to slow for real XP gaming, i have other machine to it, so i dont care too much about single core performance / clock even 2 GHz are still good enough for DOS and WIn98 and Windows Xp/7/ Linux im using to thing like Web browsing multitasking where is really good to have 4 cores.
A now numbers.. Here is comparision of CPU+MB+MEMM power which im used to use.. Geekbench 3, im selecting some average stock clock results. To be fair and count and downclocking on E5450 and overclock on E5400. wem should remove + 20% from E5450 results and add +30% (800 to 1066) to E5400 result, lets count that cpu power scale linearly with frequency.
https://browser.geekbench.com/geekbench3/comp … aseline=8685898
So number would be:
Single core E5400: 1417 x1.3 = 1842 vs. E5450 1605 x 0,8 = 1320 => OK E5400 is better for single thread is 40% faster
Multi core: E5400 = 2650 x 1.3 = 3445 vs. E2120 = 4900 x0.8 = 3920 => OK 5450 is faster but only by 13%
What about price E5400/E5500 are just 5$ on Ebay, 5450 are $20.
Xeon has additional small advantage in HW virtuallization which you can use for some task, im using it often for some testing even on these old machines.
Ok i have to admit that E5400 is better choice, than Xeons. Only thing im not use its noise and TDP, because E5400 is 65W chip, but its quite big overlocking, so it could be 80W, but also 100W, maybe someone now more precisely. Note that there is also Core 2 Duo E4700 - 800 FBS at 2.6 - 65W its very similar to Penitum E5400.
agent_x007 wrote:
Ad. 2. There are nForce3 boards with Phenom II support (and AGP slot).
Still not understand why i would want such combination, because afaik nforce 3 is not working DOS PCI sound cards and for Windows 98+ there are better combinations (i have Windows 98 on X99 chipset for example). There MBs with primary video cards selection to have additional AGP slot for other OSes is no really benefit.
agent_x007 wrote:
TDP does NOT scale with frequency as much as you would want it to.
However, I'm pretty sure most Quad Cores will have a much lower TDP on those boards regardless of what Intel says "on the box" though.
Reason is simple : Vdroop control and LLC does not exists on them, so having a 0,1V Vcore drop on load isn't unusual.
There is not exact formula for TDP decrease depends on max. frequency, but from Afterburner monitoring i would say that decrease is quite big, similar to bigger power consumation when cpu is overclocked. Maybe these older architecture are not soo good with it, but some ryzen MBs and CPU has imn options to decrease TDP from 65W to 45W or 35W, lots of X99 and i thing X79 too have possiblity to set TDP for CPU directly in the Bios and real effect are such lesser clocks (there are not cpu architecture mods depends on it, just different power state handling / limits), so it should work vice versa.
Im old goal oriented goatman, i care about facts and freedom, not about egos+prejudices. Hoarding=sickness. If you want respect, gain it by your behavior. I hate stupid SW limits, SW=virtual world, everything should be possible if you have enough raw HW.