2500+@3200+ fraud MIPS

[DPA]

alright tech gurus, MHz mathemeticians, gather 'round

this ones gonna be fun to figure out...

I got my 2500+ in today, shoved the FSB to 205MHz so that it'd be slightly higher than 3200+ speeds, and it ran fine, and i gave it a little more voltage just for good measure.

seemed to be running fine that is...

SISoftsandra comes into to verify the fair game.. and im defrauded!

Even with my MHz at 2552 i cannot touch a 3000+

I have 8408MIPS and 3400MFLOPS

where as the 3000+ has 200 more MIPS & MFLOPS that me.. WTH?

i updated my Nforce drivers, and checked all my temps, nothing is burning past 50C

what do you guys think?

This is also confirmed on another 2500+ sitting next to me running 9x whereas im running 2K SP4

any of you got a 2500+ doing the same thing?

[Xtreeme]

Sandre refrence benchmarks are typically off. Trust me -same core-same features, same mhz='s same chip, and same power. A real 3200+ would bench identical in your rig. The bench is just off. By the way the 200fsb dont do jack for the 3200+ really. After 333fsb you get like next to no gain. Its all about useing a lower multi is all and keeping pace with intel's fsb which is also 200mhz.

[Ultima]

I wouldn't say it makes no difference at all.

Since I have my nForce2 motherboard, and run my XP1800 @ 2.2Ghz/400Mhz paired with DDR400, my system boots up quite a bit faster then when set to 2.25Ghz/333Mhz and synced ram.

Agreed, most difference can only be seen in benchmarks, but it is noticable.

[DPA]

well my point is that even when the multiplier is raised and the FSB stays, i still cannot touch even a 2800+ when clocked like one, i mean im barely tagging a 2400+ out.

As far as i can tell the sandra cpu arithmetic bench should not be off as i have used it as a pretty good checker for my CPU speeds. I can't really tell if im doing any worse/better than by 2700+ [OC'd to that level] or not being as my system has not been under controlled testing and is constantly recieving a tweak or new part here and there.

oh well, ill keep an eye on this and see what i can't get to show up.

Im going to use AMDs N-bench2 to check out the true throughput of my CPU.

thanks!

[Xtreeme]

Get the results? I ran my at dif oclocked lv. And each time the PR rating and the overall score form Nbench 2 matched up pretty dang close. By like a few mhz dif. AMD is pretty well on with their rating on the lower mhz bartons speed.

[Phideaux]

You really can trust all these benchmarks. Most are synthetic anyway.

I reallly think that the Barton's pr rating is high. I don't think that the extra on-die cache makes that much of a difference and neither does the 3200's bump to 400fsb. Hell, most of us have been running at higher fsb anyway.

[DPA]

well i tested several times with Nbench2, and according to the comarison of a 1Ghz duron, i am just about exactly 340% better, which is right on target with AMD's PR.

I know one thing about the 400 FSB, its that i get quite a bit more bandwidth out of it than 266FSB. 2300MB/s -> 3100MB/s

gaming is faster than ever, and im thoroughly pleased with my barton core.


now i just have to find out if my friends 2600+ is going to OC higher and more efficiently than his 2500+... close game so far.

thanks all!

[Xtreeme]

Of course bandwidth is higher-its the mem speed after all! BUT does it help cpu that much? No. I benched at 12.5x 177 and at 11x200. The 177 makes it a tid over 2.2gig BUT the benches are were they should be. They are just about dead =. So no the 200fsb didnt do much for the 3200+. It was just so they could also say "we have a 200fsb" just like Intel. And therefore make full use of ddr400-see. But not-really. End result I found, I too see that the pr is right about were it should be with N-bench2!