Does anyone know if there isa simple utility or program out there to measure system performance or benchmarks? I'd like to compare my new computer to my old one but don't have time to run a million individual tests. There isn't some miricle all-in-one benchmarking utility out there anywhere is there?

http://www.sisoftware.co.uk - Download SiSoft SANDRA. Excluding 3D (OpenGL, Direct3D) tests, it is about as good as one can get for an 'overall Windows benchmarking package'. TomsHardware (and many other) sites use it, and have used it for years.... it also does not have a shady history.

Tests are synthetic though, and designed to isolate a particular component or sub-system (so each can be compared to another system, one by one). eg: Their CPU test results are 99% CPU, not 95% CPU and 5% RAM/chipset. It can generate (5 sided ?) spider graphs though to compare 2 systems very quickly in each area of performance in (5 areas ?).

'Futuremarks' 3DMark is also pretty popular, but a +5% increase in actual performance can give a +20% increase in 3DMark scores. (The scores rise in a non-linear fashion, relative to a linear performance increase, which is why high end cards 'look so fast' when comparing points). Mad Onion did a good job with 3DMark, but when they became Futuremark things went downhill fast. There are others who have noticed similar things with 3DMark helping to 'promote' the high end, or 'slightly higher' end cards in a false light. Many hardware review sites have ceased using it altogether.

View topic here: http://forumz.tomshardware.com/har [...] c&p=944859

A benchmark can not be honest if the underlying benchmark software is not honest.

Futuremarks PCMark is in the same light, it is OK but may show 'more than' linear improvements. The code and 'user scenarios' it executes is heavily Intel Pentium 4 optimized aswell. (eg: Encryption using SSE2 + Another task = Good relative performance on Pentium 4 systems with HyperThreading). Still it does show the improvement one could expect during such 'scenarios' but most users in a true blind test would wonder why the scores differ so much.

Some of the formulae Futuremark use to calculate the final scores could be argued as 'just plain wrong', and lead to incorrect conclusions when comparing performance, but I won't go there. (It is like how +3dB is twice as loud, regardless of the original volume, only the opposite)

The best 'benchmark' is the software you use.

I personally like the Half-Life 2 (Counter-Strike and Loast Coast only) benchmarks, so long as the same settings are used on each machine. (2x FSAA, 8x Ansio, HDR, Reflections, Model and Texture Detail, etc). As a non-synthetic benchmark they do a decent job. Half-Life 2 is more performance friendly towards ATI cards though, not that it is slow by any means on the nVidia GeForce 6800 GT/GS or 7800 series of cards.

The one single 'super benchmark' is 'the user'. If there are then any performance problems / bottlenecks observed (from a user perspective) they should be isolated, investigated and fixed.

Tip: Synthetic benchmarks are generally better for isolating problems to a component or sub-system in a very swift manner.

Be interesting to see what other people reply, generating an 'overall score' indicative of everything for PCs is hard. eg: Do you make the 'overall score ' 50% influenced by a systems 3D performance, or just 20% ?, How do you decide ?..... and that is why there are so many different benchmark programs, each designed to show off 'something' in its best light.

superpi

^ Good Answer
[shakes head in shame]