Tom's Hardware Wants You: CPU Tests For 2011

How about a folding at home test on the CPU / GPU? People build farms for FAH or similar distributed projects.

David

Dump iTunes until they introduce multi-threading.

would love to comment, but you don't read uk comments so I won't bother. Ta for the thought though.

what about adding the boot / shutdown time of the system and the launch time of applications ?
that is user experience accurate as well imo

The games list is missing a good RTS title. SC2 is a natural choice since it has been proven to make CPU's work quite hard (and is extremely popular, so there's a good reference base).

I'd also like to see Microsoft Security Essentials as an alternative to AVG.

Well I really would like to see real-life situations like multitasking. I often multitask on my computer, and a test set up as follows would be relevant for me:
Start up a game (Battlefield Bad Company 2, or something like that)in windowed mode
Have an mp3, Flac, other music-file playing in the background
and then start to unzip a VERY big file

Time the unzipping and show the result.

Our Cpu's get more and more cores, so multitasking gets more and more relevant in my opinion.

The test must not be held back by a slow harddisk, so the game could be running on one disk while the music file and unzipping could be reading/writing to/from another.

Real Time Strategy games stress the CPU far far more than most other games. As a result this is one situation where CPU performance is vital (video encoding, for example, just takes longer). Not enough CPU grunt has the potential ruin your experience of a game as the fps number plummet into single digits. As a result, it makes sense to include number for games like Supreme Commander 2, where there might be hundreds of units on screen at once, as well large numbers of physics calculations. If you want to really test CPU performance, this would be a great thing to include (PS, not Starcraft, its not taxing enough)

I'd just like to see people no longer comparing apples to oranges as even this site has been guilty of in the past. Comparing Phenom II architecture to i7 architecture is the same as comparing Core2Quad architecture to i7 architecture. It's stupid, useless and misleading. In addition, stop using Futuremark since it has been proven that it optimizes better for any CPU identifying itself as an Intel. Once that was proven, sites should have stopped using it.

I'm still wondering how much software is out there that doesn't run optimally on AMD CPUs thanks to Intel's compilers. If it's widespread, is this going to detrimentally affect Bulldozer?

That list seems fine. But could you chuck your test render for 3DS max 2010 on the benchmark page so we can download said scene and render it to compare CPUs to our CPUs.

I would like to see workstation performance reviews, like Xeon and Opteron based pc:s. There are number of free trial licenses for CAD-programs, but unfortunately programs like MATHLAB requires license, but I think it is so essential in scientific world that it shouldn't matter. Also FEM (finite element method) calculations needs significant number crushing capabilities, some free softwares exists atleast in linux world, but some commercial software should be included. FEM is the standard solver in mechanics and physics so it is in interest of many. Besides, I havent seen anyone doing performance reviews on scientific and industrial cases, so it would be great merit for TOM's.

We need to get civ 5 in there for a game. As can be seen form numerous forums the game kills systems during turns.

Not totally on topic but seems a good place to put in a request!.

It would be good to see a rolling (e.g. repeat lab every 6months say) hi-spec engineering/scientific parallel cluster chart (best performance with CPU boxes and best price per bang).

A good solid benchmark would be using OpenFoam CFD 1.7.1 freeware. A 50 million cell, steady state, turbulent, incompressible model would make a good solid test case metric to benchmark on multi CPU's.

It would be nice to see the equivalent of alt-tabbing from a particularly demanding process (i.e. a game) to desktop. While to some extent that's more a RAM, hdd and process-specific issue, its still something that people do fairly often and also tends to vary widely with the system. How to make an effective benchmark out of it, I don't know, but its an interesting point of comparison nonetheless.
It would also be nice to see more benchmarks in general that deal with running several different applications at once, like a web browser, a music player, copying a large file, encoding a video, running a virus scan or similar operation, and playing a game all at once. Sure, so many things at once is a fairly extreme case, but it should provide an interesting comparison of processors taxed nearly to their limits in a somewhat realistic environment (at least an environment that is more familiar than the many synthetic benchmark tests).

Linux benchmarks. And i love to see some benchmarks running inside a VM (i.e. compressing 10gb inside a Ubuntu guest in VirtualBox).

Lightroom!

I think it would be a good idea to consider hyper-threading on and off separately, where appropriate.

It would be useful to know how much better (or worse!) a processor with (e.g.) 4 physical cores performs with 8 logical cores, as opposed to 5 logical cores.

This would be of particular interest to overclockers, as the cpu overclock limit is reduced when HT is on (obviously).

I suspect that applications that are not heavily multi-threaded will respond negatively to hyper-threading, so it would be useful for users to be able to weigh up the different options (i.e. HT on or off) when looking at your benchmarks.

^ typo: I meant "4 logical cores", not 5! Whoops.

There's something that I've been wanting to see in the mass-benchmark comparisons for a long time:

There are so many processors included in the comparison that the charts can be a bit overwheming. I think that there should be an alternative way to view this information that is more meaningful. Here is my suggestion:

Exploit the fact that there are many processors that are effectively the same thing (e.g. i7-920 and i7-940), i.e. the same architecture, clocked at different frequencies. So, why not display a graph with cpu frequency on the x-axis, score on the y-axis, and have one line/curve for each architecture, instead of for each cpu. Imagine it - this would reduce the number of lines to a minimum, and would make it obvious to the user which architectures are better for each test. Furthermore, it would make it more obvious whether or not cpu frequency has a significant effect on that particular benchmark.

This could be expanded by overclocking (or underclocking) the processors within some reasonable range (e.g. 2Ghz-4Ghz, or whatever the max of that particular architecture is) in order to extrapolate the results, in an overclocker-friendly fashion, so that people will be informed of the expected benefits (if any) of running their cpu at a particular frequency. This would result in each of the stock cpus being a point on a continuous line on the graph.

In the past, when researching which cpu to get, I have personally implemented this at home in a spreadsheet (using your charts as reference), and the results are fantastic, and very informative. Please consider implementing this as a additional feature, as I think people would find it very useful. To my knowledge, no-one else on the web does this at current.

OK, maybe this doesn't really fall into the category, but there is a hardware test I would love to read about in Tom's Hardware.
It would actually be a comparison between "typical" pc configurations from the last, let's say, 20 years. But the point is not finding how faster an i5 is to a Pentium, but answer the question whether modern systems are faster in daily use, than older ones used to be.
And to be more specific, being a Pc uses for 14 years now, I am not so sure my PC or my laptop is actually faster than my first PC (Pentium 133Mhz) was.
Why? Well, in my Pentium, Doom and Doom 2 worked great, but in my Athlon XP, Doom 3 was almost unplayable, unless lowering it's graphics quality. What is more, I really think Windows 95 loaded faster on my old Pentium, than Windows Vista do on my Core 2 Duo. Not to mention that internet browsing doesn't look that faster on a DSL line than on my PSTN, ten or so years ago.
So, to sum things up, yeah, everybody knows that a modern multi-core system with 4GB of RAM is a lot faster than a 486 with 4MB of RAM, but the software we use also gets heavier and cpus and gpus try to cope up with it, maybe more than they used to.
My idea is to assemble "typical" systems of the past, like 386, 486, Pentiums and "modern" systems, like Core2 series, i series, load them with the corresponding software (Windows, games, other software) and benchmark them! Did Windows uses to load faster in the past? Did Doom I and Duke Nukem used to have a better framerate, at full graphics, than Crysis or Halo do today, at medium detail? Those are some questions I would like to see answered!

Thanks for your time,
Kostas Karakostas

I'd like to see Malwarebytes Antimalware in the suite of test products... oh and the internet security comparison as well...