Parhelia at tomshardware

[CuEdge]

Just noticed over at tomshardware that they are doing a round-up of Radeon 9700 PRO cards. The Parhelia is included in the benchmark as well as the ti4600 and ti4400.

The Parhelia is not doing very well, but they didn't write what driver version was used..

Click here

[Rimfaxe]

well, the P doesn't do good in speed compared to those...


THG says the review was made by Lars Weinand, the former of rivia station and he is one of the badest reviewers you get... i guess he uses old benchmarks on the first official drivers release
but on the other hand, so is nvidias cards reviews too.

[Cobos]

I did some quick research checking this review against the original Parhelia review and if the review settings of going from 85Hz (P review) to 60Hz (this one) doesn't change the performance then he did use a newer driver since the performance has improved. But why mention what refresh you are using if you do not run with vsync on and doing that removed the whole benchmark point doesn't it ??
Someone that can explain that ?

Cobos

[simpy2]

Cobos, even with vsync turned off, there can be an effect of the refresh rate on the speed of a graphics card. For each refresh, the card has to read the screen contents from the framebuffer and send them to the monitor. This leads to a 'baseline' amount of memory throughput. This amount is then not available to the 3D engine.

A quick calculation gives:
1024x768x32bit@60Hz ==> 188 MB/sec
1024x768x32bit@85Hz ==> 267 MB/sec

This isn't too bad, considering the total memory bandwidth of modern graphic cards, but it starts getting worse when you add features like triplehead and very-high-resolution displays. Also, because the refresh-reads CANNOT be allowed to be late, I can imagine they intrude on the other memory traffic in a disproportionate way (so the actual effect can be larger than the numbers above suggest). And finally, the available bandwidth is always smaller than the peak bandwidth.

For an interesting dicussion about the relation between peak bandwith and available bandwidth, see http://arstechnica.com/paedia/b/band...latency-1.html

[dshumake]

It was interesting to note in the article the following statement:

This test reflects the highest quality achievable in the real world. As expected, Radeon 9700 managed to consolidate its lead from the individual tests. The GeForce4 Ti cards fell off a cliff in this discipline; even at 1024x768, the games are almost unplayable with the NVIDIA chip. Please note that these results only show the average frame rate!

page link for text

I have never seen Tomshardware say this ever about the Parhelia

The relavent frame rates are:

GF4 Ti4400 12.49
GF4 Ti4600 13.99
Parhelia 11.68
9700 33.52
9700 Pro 39.03

The frame rates were measured in UT2003 with 4x FSAA +8X Aniso (Parhelia used16x FAA and 2X aniso) at 1600x1200 32Bit

[3dfx]

Yes but 16 X FAA of the Parhelia is on intended to be used at 1024 by 764 resolution as that is thr point of the 16 X

[R.Carter]

Originally posted by dshumake

(Quoting from the review)

The GeForce4 Ti cards fell off a cliff in this discipline; even at 1024x768, the games are almost unplayable with the NVIDIA chip. Please note that these results only show the average frame rate!

I'm guessing that the frame rate on the Nvidia card varied a whole lot more than the frame rate on the Parhelia. Choppy frame rates tend to be a lot more annoying than a smooth low frame rate.

[Yxalitis]

Originally posted by R.Carter


I'm guessing that the frame rate on the Nvidia card varied a whole lot more than the frame rate on the Parhelia. Choppy frame rates tend to be a lot more annoying than a smooth low frame rate.

Not so, the difference between the Parhelia and nvidia cards is the way they handle low frame rates. The Geforce cards stutter, they hold the current frame, then redraw the next frame when able, leading to jerky gameplay. The Parhelia goes into "Slo mo", which is why a Parhelia is more playable at lower frame rates then a Geforce. This is most noticible in UT2003, even with frame rates as low as 12, the Parhelia is still playable (NOT enjoyable, of course), whereas the Geforce is a jumpy slideshow.

[CaineTanathos]

When I run the benchmark I get almost the same score as the GF4 4600 (my parhelia is 5% overclocked)

Damn man , I mean, if people show these crappy results no wonder that people think that the Parhelia is crap...

Lars Weinand , dude, grow the hell up!

[Rimfaxe]

Originally posted by CaineTanathos
When I run the benchmark I get almost the same score as the GF4 4600 (my parhelia is 5% overclocked)

Damn man , I mean, if people show these crappy results no wonder that people think that the Parhelia is crap...

Lars Weinand , dude, grow the hell up!

Mmmm, tell me about it! He never learns...