Hardware Fermi Testing Update

Therein lies a driver optimisation issue: load balancing and resource optimisation. Its not shunting tesselation to a dedicated unit, but handling it in the shaders. Thats an area ripe for thorough tweaking.

 

bit-tech

+1

 

I'm sure the nvidia drivers will mature and the cards will get faster, but Ati have had their card out for 6 months and their probably working on bringing out a new one.

 

id prefer a stock review than an overclocked review to the poster a while back

i dont want to thrash my cpu at 4.2ghz to get hte best performing card

 

Probably, but if its on 28nm or 32nm they're in for quite the wait. 40nm might be around till the very end of this year :/

 

:s second paragraph

 

One more issue with overclocked test rigs is that they are a lot less reliable than stock clocked ones - when you're testing graphics cards, 1 or 2fps matters, and you need to know that when you do your tests, those differences are because of the GPU's on test, not the rig being a bit flaky.

 

+1 again.
Great work Bit-Tech team. Not many sites would do this.

 

+1 You said it all

ZING! He shoots, he scores, with this BT is in the lead =D

 

Still... at this time... the Fermi cards dont seem worth it. Im still going to hold out a bit longer until some new drivers come out or at least they do a revision to the chip.

 

So then as just a gamer i read all these reviews just out of interest.

But if my understanding is correct then there all as good as each other for me?

None of them will provide anything better over each other?

I mean my monitors refresh rate is 60hz, so a card that can go 100fps is just pointless because i'm not going to see a difference. I assume that's right. Correct me if i'm wrong please.

 

I trust the results you're printing, but.....you think the 480 is low rent in comparison? I'd have thought the very people who can afford a 480 are the ones that can stretch to a core i7 system!

Anyway, come on guys, how's it go with Arma II on 10k draw distance & everything on max....I'm sure JG already knows....

 

But it also means we'll have to test everything again PITA factor aside - it took a week to test it first time!

Also what Sifter said - we ran 3.6GHz OC machines last year actually and they constantly fell over in the end after months and months of hammering them.

 

Oh god, flashbacks! Broken DDR3 sticks everywhere! Staring at BIOS screens for hours on end. I won't go back!

 

It may be uninformed, but my opinion is that you should only overclock test rigs if the CPU is significantly limiting the hardware being benchmarked.

If you have an always overclocked system, then where do you stop? Liquid Nitrogen cooling for all hardware reviews?

I prefer it when hardware is tested among it's equals, so the highest end GPU should be paired with the best commercially available CPU. Mid range GPU? Pair it with a mid range CPU. So on and so forth.

Once you start benchmarking hardware that's significantly out of place, the reviews become useless as they do not represent common real world scenarios and thus become relevant to nobody.

Only problem with this methodology is that you can argue all day about what is currently mid-range, high-end etc. But then this is the internet, so there has to be something to argue about.

 

The review was tested with a core i7 965 - it just wasn't overclocked. I personally think that using an overclocked test rig isn't terribly responsible, as it would inflate results past what would normally be considered a real-world environment. It wouldn't be by much, mind you, but I appreciate Bit-Tech's real-world approach to testing. It's the reason I go to Bit-Tech for my reviews.

It's really nice to be able to check out the hardware, make a purchase, and then see my own experience correlate with what I expected. It's easy enough at this point to estimate the impact of my own system on the end results (given at what point in my own upgrade path I'm currently at). It would be much harder to factor in the overclocking potential of an unfamiliar system and the impact of OCs that get flaky as well

 

I think there is merit in the overclocked CPU for testing argument. Not for all cards obviously, but for cards like the 5970 and GTX480 (basically any high end, $600+ cards etc) it is surely worth it. I mean, if someone is willing to spend so much on a single component, do you really think they won't be willing to spend £40 on some serious air cooling, assuming they haven't gone for water?

I think benching everything at stock is clearly the best choice, but I think for high end VGAs, especially multi-GPU ones, it would be worth running some overclocked CPU benches as well. That way the hardware doesn't get constantly thrashed, and you add yet another level of detail to your reviews that you don't see anywhere else.

 

Bit-Tech the real No spin zone, fair and balanced reporting. wanna see the 2 gb eyefinity cards vs. 480 and where is the HD 4890 crossfire review you never ever did.

 

Thank you for the update. Always great to get the nitty-gritty details and clarification.

That is why you are on my daily reading list

 

The problem with overclocked systems is reliabilty if you use your pc all day every day for testing or high end design cad work
they willl buckle then collapse after 6 months 180c of heat going into your case does dmg. Condensation alone would become a problem for most. There's no water cooling block for 480 yet. And i dout when gaming at 2500 resolution if the CPU is the problem. Some test showed the dif between a 3ghz quad and a 4ghz quad was 10-15 fps at most on low resolutions and 2-4 at high

If your using the 480 for folding rig you never overclock. Folding is like doing prime95 + furmark on loop for days upon days

a i7960 is hardly standard CPU for most people 920 or 930 is alot more realistic. 45c at 3.6ghz or 80c at 4.2ghz doesn't take a maths kid to tell you what will break first Thats the temps my own i920 Averages under load on water cooling