HD 4870 review thread.

Is there any explanation why ATI cards seem to be able to handle high resolutions with AA well without needing over 512MB of RAM? Are ATI drivers/cards just more efficient with memory? It's just seems odd that Nvidia needs over 512MB of Video RAM.

 

lopri :
"Say, in a hypothetical situation:

- Card A (512MB): constantly (quickly) moves or reuses 300MB of buffer efficiently thanks to extremely high bandwidth and smart design
- Card B (1GB): slow to switch frame buffers or constantly redraws same buffers again and again (brute force) "


As for AA performance read this page :
http://www.rage3d.com/reviews/video/atirv770/architecture/index.php?p=6



"The RBE were another primary focus for improvement, mainly in order to increase AA performance, another R600 weak-point. Gone is the dedicated alpha/fog unit-it's use was rather questionable in terms of benefits anyhow. Depth and stencil capacity is doubled, which means that the RV770 is actually capable of Quad-Z/Stencil rates"

"RBE improvements are centered around this area: the RV770 handles 4AA samples per clock and is thus capable of outputting 16 pixels per clock even with 4X AA, whereas the R600 could only manage 2 samples per clock, which resulted in it outputting only 8 pixels per clock with AA enabled. This means an overall doubling of fillrate with AA compared to the R600. As a bonus, non-AA fillrate is also doubled for FP64 (16 bit per component) color formats going from 8 pixels per clock to full 16 pixels per clock."


read the article to get a better understanding :!

 

If people are interested in AA scaling, TR used HL2:Ep2 (dunno about AF) and AT used Oblivion. Neither used FW 175.19.

 

ATi is better at memory management, currently.

 

You are right, Amit. AT's article was actually pretty fair; it is the editor's "comments" that were made that was irritating to some of us - comments that did not necessarilu agree with their own conclusions and benchmarks, as i remember.

Yep, IQ comparisons were formerly pretty irritating timewise. But you can make FRAPS videos as 'proof' - i use them as the basis for my own comparisons and i have used a digital camera also, when it is tough to see the differences that show up only when the scene is in motion. Some games are also much harder to capture - i have an awful time with BioShock, but Crysis is easy because of the tools built in.

i also prefer to take my screen shots at a lower resolution as it is just easier to see differences. Then i can retake them at whatever resolution looks best in the shot. Lots and lots of time involved, seems to be the main problem doing IQ reviews.

 

But what's the point or usefulness of an IQ comparison that highlights differences nobody would notice at normal resolutions during gameplay? Magnified low-resolution screenshot comparisons ftl.

 

Because during gameplay those differences actually stand out even more. Motion causes the jagged edges to appear to "swim". I certainly see it in Oblivion @ 1080P w/HDR & 4xMSAA. Not so bad when you're just standing there looking at something, but start moving (character or camera) and it's noticeable. Then again, I do play on a 42" screen...

 

I wasn't referring to artifacts that are more apparent in motion than in screenshots...hence the "during normal gameplay" reference in my post.

In very few cases I've found screenshot comparisons useful as so many other things are different - brightness, gamma, contrast - that it's impossible to isolate miniscule differences.

 

I'm not sure if I'm remembering this right, but if you use Fraps to grab your screenshots instead of the built-in HL2 feature it'll grab accurate images... or that could quite possibly be the other way around: if you're using Fraps, use the HL2 thing...

<offtopic> I used to really enjoy makinmg those AA comparison things. I'd spend hours and hours grabbing screenshots and preparing images in Photoshop then making the javascript apps work in a way so that it was easy to compare images ... check these out:
http://www.rage3d.com/reviews/video/nvidia8800/aa.php
http://www.rage3d.com/reviews/video/nvidia8800/aa2.php
</offtopic>

 

From what I've seen mentioned elsewhere, it seems to only affect GX2 / SLI cards due to missing SLI profiles.

 

Is the driver that TechReport used, "Catalyst 8.501.1-080612a-064906E", the same as the Catalyst HotFix drivers?

 

Catalyst 8.5 WHQL [8.493]
Catalyst 8.6 WHQL [8.500]
Catalyst Hotfix / Radeon™ HD 4800 Series [8.501.1]

 

Yes.

 

Yes, the drivers that were made available to the press were also publically posted as the Hotfix; we made an update to the drivers to include better Crossfire X profiling for Bioshock and The Witcher and the Hotfix was also updated with these profiles.

The Hotfix largely forms the basis of Cat 8.7 as far as the HD 4850/4870 is concerned.

 

i am using the Hotfix drivers [after installing what was on the ooB CD] and this is what my CCC reports:

However, i am getting some real oddities in my results and i will try 8.6 next, benchmark it, and then rewrite it again with the Hotfix and redo my benches over. 2900xt Crossfire should not get more in 3DMark06 than my 4870, i believe; so something may have gone wrong with my original driver install.

The hot fix is "beta 8.7", right?

i didn't go out to that party and i am about to head back to benchmarking-land. i will check back in tomorrow with more sure results. So far i am not terribly impressed with 4870; it is a step up over 8800GTX, however. Now i will go figure it out. i also purchased "Vantage" but have to re-download the Advanced version as it is much larger than the 1-try demo - another 30 hours to go over 56K

 

Really miss your reviews, Mark. Those IQ comparisons were the best on the net!

 

I know :! you were the ultimate GPU reviewer .. damn rage3d owners did crap all with the site. I really wished they had let you buy the site.

 

Apparently the "elsewhere" you are referring to is here. Definitely not comments from Nvidia sources.

However Benchmark_reviews.com did state:

Which indicates that there were performance hits on the 9800 GTX cards as well, but not the 55% decrease noted in some GX2 benchmarks. The good news is not many review sites used the 175.19's.

Pharma

 

Looks like Dave saved the day, thanks to him people who broke their cards by flashing the bios can fix them

http://www.techpowerup.com/articles/overclocking/vidcard/152

 

Have you read it lately?