HOLY SHIZNIT!!!! V5-6000 - "performance explained"

[RoboTECH]

Very very interesting.

I wrote the author of the article a little email asking about texture compression.

Here is his response, cut and pasted:

" "Compression was set to 0 in the q3aconfig.cfg file. Unfortunately, we
didn't have time to run benchmarks with compression enabled"

 

[Wingznut]

Ok, that's pretty irresponsible to run benchmarks such as those, without releasing that tiny-little detail.

Is it just me, or is videocard reporting getting worse and worse?

 

[RoboTECH]

dude only had about 2.5 hours to do his testing, apparently

this is pretty major information, on the order of 10-15 fps

it also really explains why FSAA and hi-res performance were so "lackluster"

 

[Muerto]

Yeah but how long does it take to enable texture compression? Is it an hour long ordeal or a few key strokes? I'm assuming the later. I think video card revewing IS getting worse on a lot of sites.







Not this one though.

 

[CplHicks]

Correct me if I'm wrong, but I don't believe the guy was a "professional" hardware critic. Another interesting fact about the benchmarks is that they disabled dynamic lighting, which could theoretically erase the gain from TC. Just some food for thought...

 

[PG]

But remember that he ran the benchmarks with dynamic lighting turned off. Maybe these 2 issues cancel each other out. What do you guys think?

 

[lsd]

<< Another interesting fact about the benchmarks is that they disabled dynamic lighting, which could theoretically erase the gain from TC >>


Well.. more like cut it in half. Anyway the performance would have still have been dismal for a card that should have been a power house.

 

[DefRef]

I caught that he turned off the dynamic lights right away. It doesn't matter if he likes them or not, the purpose of benchmarks is to run a UNIFORM set of tests to measure performance. He should've turned off the visible weapon and cranked down the curve complexity while he was at it.

 

[Wingznut]

Am I mistaken, or...

Arent the 1600x1200x32 benchmarks the same as the GTS Ultra (referencing AnandTech's review)?

Considering this is without TC and with immature drivers... I don't see the V5-6000 being a disappointment.

 

[TGCid]

There are simply too much wannabe hardware review sites out there. Some people don't even know jack.

 

[RoboTECH]

well, there are a few things of note here:

1) TC was disabled

the 5500 (mine) loses a good 20 fps @ 1024x768x32 max quality (no $hit)with TC disabled

2) The drivers were based off of 3dfx's original 5500 drivers an did NOT include the speed gains seen in each of the last 3 releases

3)Trilinear filtering was enabled AND WORKED - Quantum 3d tweeked this apparently

4)The depth precision setting wasn't tweeked

overall, after talking to the guy, the 6000 isn't &quot;so bad&quot; afterall

dynamic lighting results in a 4 or 5 fps &quot;hit&quot; on my 5500, FWIW. TC is FAR greater than dynamic lighting, as far as FPS is concerned.

 

[Soccerman]

yeah guys, trilinear filtering is a big thing on these cards. AFAIK, pushing trilinear filtering on these cards means yet ANOTHER pass for each pixel pipe on that pixel, meaning that in Quake 3, many of the pass counts amount to 3, thus reducing the speed of each chip by a factor of 3.

that is, if the VSA-100 works like I think it does..

as for TC, yeah a pretty good boost would be achieved with that, not to mention that the above trilinear filtering takes up more memory bandwidth (I think, I'm not sure), so whatever can be saved, will and does help.

anywho, they didn't mention the clock speed either.. the V5 6K was supposed to be released at 183mhz, but we don't know for sure if that's what this one was at..

as for quality of review, this one was ok, considering the low amount of time they had to do this. I wish they didn't change any options though, then we could compare them to ones from Firingsquad, and Anandtech better.

though I think you are correct, many sites do not have quality reviews. Anandtech is (fortunately) one of the better ones, however I think they need to do more investigation type work, to prove to readers why the card is good or bad in certain areas. Also, I think that Anandtech needs to post many more system specs and options that are enabled, or disabled (BIOS tweaks, fresh install of windows or no, DMA enabled? options in display properties panel etc).

 

[Wingznut]

And not crippling one card by using old drivers and the less common API.

(Heh... Soccerman, I think you hit a nerve. )

 

[RoboTECH]

yeah, i'd like to see some Evolva scores in there