Is SiSoft Sandra something we should trust?

I have an XP 2400+ mobile at 2.21GHz. This is almost equal in clockspeed to an XP 3200+, but my FSB is a little lower. So why on earth is Sandra loving my CPU so much?








Does anyone else have numbers that are this far off?

It's all synthetic tests I think, take it with a grain of salt. That's my experience with it; it can show some interesting things about hardware but it doesn't always apply to the real world.

Which version of SiSoft Sandra are you using?

Also to answer your question, well I've heard a lot of people talking about that we can't trust Sandra, others were arguing about that we can... etc. The thing is that it's not "inventing" it's scores. So everytime when you get the "newest" Sandra available, then for sure, those scores, which are predefined, and saved already in it, those were tested and benchmarked indeed at those systems. I don't see why they should change or screw the scores up.

When benchmarking there are a lot of other circumstances, and factors what are changing the scores. Just like motherboard, memory, hdd- not only the CPU. So just becuase your CPU is lower FSB then that 3200+, and it's clockwise the same, that doesn't means it's performance should be exactly the same. Your CPU, being a mobile one, and even overclocked (nvm that it has lower fsb then the other, it is still oc'ed)- it simply beats the shit out of that "test athlon 3200+" what they were using to get the predefined scores.

That's why I was curious which version are you using, because these scores get updated at every new release, so it's critical that always to have the newest version. Otherwise the scores can indeed be screwed up, and/or get outdated.

Benchmarks can't be really compared as is, but are just great enough to see for yourself how your rig performs. Comparing scores you can't, until you don't have the same OS (with the same background applications, services running), same motherboard (which is impossible when comparing AMD vs Intel), same memories.

Tomorrow I'll bench my rig with Sandra, if I don't forget it. :winkgrin

Newest, Sandra Lite 2007.


Yeah, I can't see a motivation for them messing with scores either. I was thinking that perhaps they are just bad at what they do?


Interesting, I didn't know that...


Much obliged!

OK- Tomorrow I'll run the benchmarks, when it's cold in the morning Downloading...Just like archnaid also said, these are synthetic benchmarks, so can't just relay on that you will beat every Athlon 3200+. But nvm hence due to those scores you're alright- also are you satisfied with your rig? That's what all matters.

EDIT: pfrr, only 11MB- finished

Haha, yes its a small download. And no, I wish I could get more out of my CPU, but it doesn't like me. (It's a freaking 2400+ mobile 35watt with the IHQ *or whatever* stepping! Come on!!!)

@ 3.6GHz, 240x15, 1.39v.
2x512MB DDR400 @ 384MHz [2.5-3-3-7]

Nice benchies, so how does yours compare to a stock 3.6?

Performs better, due to higher FSB.

EDIT: I'm very satisfied with my memory benchmarks. Why? Because my sticks are Geil Value 2.5-4-4-8 (cheap) sticks. And now running at ~384MHz instead of 400MHz, running at 2.5-3-3-7 - performs better then DDR400 sticks running 2-2-2-5. According to Sandra.. well in RL not sure about it, but nvm

I am not sure how FSB metter when benchmarking CPU. It should no metter at all. Of course it metters when talking about overall performance, but it shouldn't metter in CPU benchmarking. If it does it is evil benchmark.

Yeah, you're right... then it's kinda evil. Because few months ago I was playing with an EE, also with older version of Sandra... and... made tests at the same CPU speeds in 2 ways: a.) with higher FSB and lower multiplier; b.) with lower FSB and higher multiplier - the overall MHz was the same - so it shouldn't make difference. But it was... and it was kinda huge difference.

The processor can't process data that hasn't reached it yet. Access to data is a big part of a CPU's performance - so what if it is capable of, say, more FLOPS if the FSB can't keep up with it? On the otherhand, a CPU that is capable of keeping up... performs better.



Anything else would be purely synthetic and not apply to most real-world scenarios. IMO a big part of a processor's performance is it's ability to interconnect to RAM and other data streams. A 533MHz FSB Pentium 4 doesn't perform as well as a later Pentium 4 even if the coreclocks are the same. A synthetic test might show they're very similar, and there would be some difference due to changes in the core, but in practice there wold be more difference than the syn thetic test would indicate.