If processor speed isnt the answer, what does a good gaming pc need?

[Thrawn]

My highly technical assesment of this question is that a perfect gaming pc would have stuff.

That is all.

 

[Zakharov]

If you follow the article, you will notice that in the tests done with Doom 3 and Far Cry (very demanding) the performance gain is absolutely negligible, while in the test made with Unreal Tournament 2004 (less demanding) there's a wide performance gap between the various CPUs tested.

You are correct in your conclusion here. However, these tests are using conditions which artificially cause the graphics card to be the bottleneck.

Firstly the graphics card used is a single 6800GT, when the current top end setup would include a 7800GTX SLI pair. Secondly, the settings used are 1280x1024 with what appears to be no AA or AF applied. Thirdly, the demos used here will have been configured to stress the cpu (physics, AI etc.) rather than test the fill rate of the gpu.

What this article does do though, is show that the perceived cpu bottleneck is much less of a factor than I had thought. Unfortunately these review sites do not push the entire system to its limits. If you read an article which reviews graphics cards, then you will see the results skewed in the opposite direction.

Regardless of this discussion, your earlier statement holds very true:

Considering those results, the best gaming pc is a balanced thing. Good CPU + good graphic card, or high-end CPU + high-end graphic card, depending on your budget.

I should add that after achieving a balance, any extra room in the budget should be spent on the cpu as it is much easier to upgrade the graphics card at a later date if necessary.

 

[dannyevilcat]

The dual core stress test:
http://www.tomshardware.com/cpu/20050603/index.html

Basically, the whole point of the stress test was to find out which system is more stable. So they pitted an Intel and an AMD against each other... and then the Intel system started rebooting, experiencing downtime. Tom's went through IIRC 3 or 4 motherboards in about as many days before finally blaming it on the heatsink and fan that Intel supplied them with.
Everytime the Intel system went down, Tom's came up with an excuse for it, and reset the test, to make the test "fair", even though it was defeating the whole purpose of the test in the first place. I'm pretty sure this is glossed over or unmentioned in their stress test final report, but it is captured for posterity in forums (including their own, which I'll try to find later if you're still interested).

The whole time, the AMD system just hummed along, solid as a rock. Tom's minimized the Intel failures, and downplayed the stability of the AMD, but not counting the results until they had finally managed to keep the Intel system running.

In other, less spectacular failures, the site has been widely derided for publishing biased information.
Of course, not everything he* does is bad, or wrong. I just don't trust his testing methods.

EDIT: by "he", I mean the site and it's reviewers. From what I've heard, Tom himself cashed out a few years ago.

 

[tR1cKy]

You are correct in your conclusion here. However, these tests are using conditions which artificially cause the graphics card to be the bottleneck.
...

Uh. I didn't consider that thing. I'm not much into graphic cards tests, but this statement seems logical. Tom's tests were intended to test the CPU and not the graphic cards. With a test tuned to test the graphic card limits, the results should look different. Thanx for having me consider an aspect of the thing that i didn't give the necessary attention.

Basically, the whole point of the stress test was to find out which system is more stable. So they pitted an Intel and an AMD against each other... and then the Intel system started rebooting, experiencing downtime. Tom's went through IIRC 3 or 4 motherboards in about as many days before finally blaming it on the heatsink and fan that Intel supplied them with.
Everytime the Intel system went down, Tom's came up with an excuse for it, and reset the test, to make the test "fair", even though it was defeating the whole purpose of the test in the first place. I'm pretty sure this is glossed over or unmentioned in their stress test final report, but it is captured for posterity in forums (including their own, which I'll try to find later if you're still interested).

The whole time, the AMD system just hummed along, solid as a rock. Tom's minimized the Intel failures, and downplayed the stability of the AMD, but not counting the results until they had finally managed to keep the Intel system running.

In other, less spectacular failures, the site has been widely derided for publishing biased information.
Of course, not everything he* does is bad, or wrong. I just don't trust his testing methods.

I have re-read the stress test, taking the necessary time. Well, my impression is different.

For what is reported there, the whole point of the test was how the 2 platforms behave in various types of utilization - in respect to stability, of course! An unstable box is simply unusable. So stability is more a prerequisite rather than the goal of the test.

Yes, the intel box had problems. Various mobos had to be swapped, the cooler wasn't sufficient (Intel's blame! The processor is extremely power-hungry, so a mega cooler is necessary to cool down the thing), and there were also RAM issues. At the end, the pentium turned out to be rock stable if coupled with a mobo with an intel chipset.

So the stability issue was solved at the end. And this, in my opinion, is the most important aspect. Why? There's one important thing to be taken into account, and Tom's failed to give it the necessary importance: the intel cpu was still a production sample! It's been common in the past years to stumble upon problems with early samples, and those issue are in the most cases solved when the final product hit the shelves.

Obviously, those stability problems are something to worry about. I think we need to see a stability test of the final product to give a final judgement on the intel cpu, and to the validity of the stress test as well.

Hovever, the article is not free from incongruences and flaws. First, it's badly written and badly organized. Second, it gives conflicting impression to what platform is to be considered the best in a business environment. It is said in the first pages that the intel cpu is to be preferred due to the best support. But later on it is said that, given the increased power demand, in a fairly sized corporation with 100+ puters, the increased cost of the power bill would be something to be taken in serious considerations. So, at the end, what's the best platform for a business environment?

About the accusation of being biased, i'm a bit suspicious. In the past years, Tom's didn't refrain to lambast intel products whenever they were faulty or they lagged behind the competition. You should remember the harsh words used against the first P4, the infamous willamette core, that at 1.4 ghz was smoked by an athlon at 1.0 ghz. Tom's was also the one to expose the inherent flaws of the old P3 1.133, and intel was forced to recall the product.

The same thing, of course, happened to amd whenever their product was inferior or had some troubles. Everytime Tom's puts one of the 2 parties into better lights that the other, the corrispondent fanboy community cries loud and accuses tom's to be biased. It's an old story that always repeat itself.

 

[dannyevilcat]

Here are some threads from their own forums. Note the disdain for the site.

http://www.community.tomshardware.c...cpu&Number=569995&page=60&view=collapsed&sb=5
http://www.community.tomshardware.c...cpu&Number=569349&page=61&view=collapsed&sb=5
http://www.community.tomshardware.c...cpu&Number=571066&page=52&view=collapsed&sb=5

Besides the usual crap found in any forum, and the length, it's actually quite informative.

EDIT: I re-read through these posts just now. I swear my IQ dropped. Anyway, we can agree to disagree. if you trust Tom's, I'm not about to stop you.