Realistic Benchmarking

I am also in the market for a new PC - although the video card is actually the component which I won’t be replacing. It’s not as good as your 5900 - I only have a Radeon 9600XT - but I expect to squeeze another 12 months out of it. Instead, I need a CPU/Mobo/RAM upgrade. It’d be nice to have DDR RAM for a change, but because I will also need a new power supply for all this, I’m looking at AU$700. Not cheap. I have the cash, but it puts a considerable dent in my car savings.

The thing with review sites, which I’ve seen, is that they do indeed use fast and unreasonably priced processors to show to true performance gap between different video cards. As far as I can tell, this is what people want. To an extent, it’s also what I want. That’s not to say review sites don’t sometimes use lesser processors. Nor is it to say I don’t appreciate it when they do. Toms Hardware used to do it on their big VGA charts, and it was interesting. Also, many sites include CPU scaling articles, which show how much the performance of a card changes as you use a less/more powerful CPU. They were always helpful to me when looking at my purchasing decisions.

That said, I remember HardOCP tweaked their reviewing system some time ago to compare video cards at their “highest playable settings”, so cards would sometimes be compared whilst on different detail settings. After some public dismay, they started to include the old “Apples to Apples” comparisons that everyone seems to want. They seem to want to know what the card is capable of, not how it will run for them.

So it seems quite evident that people prefer to see exactly what performance they are getting for their money, component- for-component, rather than how each piece of hardware affects their overall system performance. I don’t think we can say the benchmarks are misleading - they’re not. The only people who are misleading the consumers seem to be themselves.

As Cyrris said, those reviews arent to show how many fps you’ll get with the card, but to show what FPS rates this card is capable of. to do this they have to make sure that the video card is the bottleneck, not the CPU or RAM.

Plus, I just realised, if you want top of the line, get the fastest AMD 64 CPU, 4 GB of PC3200 (or more) ram, and TWO PCI-X GF6800s, in SLI mode.

I am not saying we should do away with the pure power of the card tests. But also include something that is a bit more realistic. Most of the card battles turn out to be so close that only someone that is obsessive or anal would truely notice (or care) about the 1.2 difference in FPS between an nVidia or ATI card.

I am thinking about doing a more massive upgrade making either a 6800 Ultra or x800 XT the probable card. But when I look at the next slot down, say the x800 pro, I could save myself a great deal of money but I have no idea what to expect of the card with my current system. The x800 pro would slide right into my PC and work with everything I already have while a XT or 6800 Ultra would require an entire motherboard and CPU change.

So its hard to compare “bang for the buck” when I can’t grasp the exact performance I will get out of the x800 pro since it is always tested on systems that are out of my reach.

If you are considering buying a card like the X800 Pro, and with it NOT upgrade your system, you arent going for max preformance. You cant expect it, and exactly BECAUSE you dont upgrade your mobo and CPU.

Personally, going from your 5900 to an x800 pro is a fucking waste of money, but hey, its your money.

You kidding? The x800 pro is double the performance of my card. The new generation of video cards is one of the biggest jumps in performance ever. Upgrading from a x800 pro to a x800 XT is a waste yes.

Bang for the buck its more than worth it. My last card was a GeForce 2 64 meg DDR. I upgraded from that to the GF 5900 which cost $4-500 at the time and it was probably a 75% increase. That’s 3 generations of cards for less than the increase of just one generation. I’d say its worth it either way I go.

Uhh, going from a GeForce2 (even the best one) to a GeForceFX 5900 would be 2-3 times the framerates, depending on the game. If you didn’t get that much of an increase, either your processor was crap, or you were running with the details on low with the GF2, and very high with the 5900.

It’s true that the leap from last generation’s video cards to the current ones are, as nVIdia said when it’s launched the 6800, “the largest single performance leap in graphics card history“, but the truth is that there are no games that your 5900 can’t handle just fine, so I’m inclined to agree with JohnDoe on this one. If I were you I’d wait until the next ones come out. ATI has it’s R480 due out in a couple more months.

Amen brother!

Only a couple months? Why so soon? Any links by the way? It seems like the “Just wait for the next batch” excuse could go on forever though. You have to buy something sometime.

I wasn’t thinking straight about frame rates and GF2 vs 5900. Technically the framerates didn’t increase nearly at all actually. But that is because these days its not about getting 100 fps but rather getting around 60 with the most detail and extra options (FSAA, AF) turned on. Kind of a balancing act to see how much visual splendor you can get at a smooth and stable framerate. I forget that while I ran UT2003 back on the GF2 at about 50fps and now run it at about 60fps, back then was 1024×768 now its 1280×1024 with 4x fsaa & AF.

Also I must restate I am a consumer whore. I am not looking for a card to simply run games well. I want a card that makes games scream for mercy. And that will always be the top of the line card for whatever generation we are in. Its always $4-500 price bracket as well. This generation is a proven performer and huge increase. Plus I can still sell my GF 5900 at a decent price right now, by next generation I wouldn’t be able to give it to a hobo.

The way I see it, upgrading your video card every year really isn’t that terrible of an investment. I could resell my 5900 right now for about $175-200. That’s about a 50% return on my money (I bought it at $399). So when I go and buy the next top of the line video card for another $400 I am really only spending 200 new dollars. This is like buying a new card every 2 years and spending the $400. But instead every single year you have a top of the line video card. If you saved just $5 a week you would have more than that $200 each year to upgrade. Makes sense to me, plus I want to see HL2 in all its blazing glory.

I suppose that HL2 is the only reason I would recommend steering clear of your 5900. Despite your card being high end, in HL2 it performs on par with my mid-range Radeon. nVIdia’s FX series really was a flop.

I don’t know who would buy your 5900 though. For $175-200US you can get a 6600GT of a X700XT, both of which perform better in, well, all situations. Nevertheless, if you’re a cunsomer whore, there’s nothing I can do to sway your decision.

If your upgrading from the 5900, i would wait for 512MB cards to hit the market. When you think about all the new tech comming out it seems crazy to buy now. Im spending shit loads of cash on a new system in January next year, and this is what its going to be:

Athlon 64 4000+
nForce 4 chipset board with PCI-e, DDR2 and SATA2 (SATA2 is a mabey, more likely this tech wont be available till mid next year)
>2GB DDR2 (hopefully latencies drop like a bitch, as the 64bit CPUs with the integrated RAM controlers really really love low latency)
6800 GT (or Ultra) PCI-e with 512MB GDDR3 ram

everything else is superfulus but when segates 400GB SATA drives come out i would love to run 3 or more of them in RAID5.

Chef - correct me if I’m wrong (and if you can, I’d love a link to the place you found out), but I don’t believe that Athlon chipsets (including the nForce 4) have, or will have in the near future, DDR-2 support. While Intel requireds DDR-2 speeds for the way their processors work (1066MHz FSB speeds and all), AMD took a different approach, so RAM faster than 400MHz really isn’t beneficial except for slight overclocking.

So while your nForce4 will have PCI-e, it won’t have DDR2, I’m quite sure. If it did, you’d get no real performance gain, and you’d be wasting a lot of money because DDR2 is very expensive.

Aye, while the nForce 4 chipset is revealed the actual mobo manufacturers havent released any solid specs for their new ranges of mobos.

Upon further reading on the integreted RAM controller in the AMD cpus it seems that extreme low latency DDR is the best possible and DDR2 would indeed either not work or be completly useless.

Thanks for pointing me in the direction of truth