Oblivion benchmarks by Anandtech

[Vile]

http://www.anandtech.com/video/showdoc.aspx?i=2746

=D

 

[pauldh]

Just got done reading it; thx for the link. They did a pretty good job explaining their testing methods, although I didn't see if they mentioned running multiple tests. They rather mentioned the 5% fluxuation line seeming to mean they ran just one test. Firingsquad ran each one a whopping 10 times and averaged the scores. Seems to be the way to bench oblivion. Anyway, it's welcome to read more of an explanation as some sites post the numbers without clueing us in as to what they ran to achieve such numbers.

Interesting how when you go less than max settings as in Firing squads high end review, the picture changes a little. For one thing, the X1900XT pulls a big lead on the X1800XT, which FS did not show. Of course that could just be the actual manual run through difference too.

It was good to have our suspisions confirmed that the X800's and X850's do really shine in oblivion. X800XT and X850XT > 7800GT in the high settings outdoors test, while the X800XL even beats the 7800GT in the medium settings test. One things for sure, despite not having HDR capabilities, the X800XT and X850XT are serious cards for oblivion.

 

[Vile]

You said it yourself. I really like AnandTech reviews. It's nice to know the settings (Int. Shadows, Ext. Shadows, Shadow Filtering, etc) they used for the High quality and Medium quality. Also showing the min frame rate and the average on the same chart is nice...

It kinda pisses me off to know the 7900GT can't stay above 25fps in most outdoors tho

 

[mattyink]

Well to me that test isnt a good indication.
First of all they enable self and grass shadows, nobody actually plays with them on because self shadow just looks awefull and grass shadow isnt really noticiable. With both those off you get a good increase in performance.

They also dont even enable AA, well i do understand sine only ATI cards with the chuck patch can do both but still. I dont call those settings they use high quality.

 

[mattyink]

You said it yourself. I really like AnandTech reviews. It's nice to know the settings (Int. Shadows, Ext. Shadows, Shadow Filtering, etc) they used for the High quality and Medium quality. Also showing the min frame rate and the average on the same chart is nice...

It kinda pisses me off to know the 7900GT can't stay above 25fps in most outdoors tho

Beth said it that even currenty card couldnt play the game maxed out.
Most people are just pissed cuz they were stupid enough to buy a PC for this game before the game was even out.

Its like Doom3 at the time even a top of the line PC had trouble with it.
People cant stand to not put all slider to max wich is pretty stupid if you ask me.
Also a lot of people focus way too much of Frame rates, my suggestion to them is turn off the FPS counter. When it drops bellow 30 people freak out but if they didnt even saw the FPS counter they wouldnt even notice the frame rate drops most of the time.

 

[TheGreatGrapeApe]

The fact that they didn't enable HDR+AA just shows it's not a review about the game and the hardware but pitting company A against company B, and we already know which camp Anandtech is in.

I saw this review when it was first posted and said, got alot of info there but far too much of it is contradictory for buying a good card (GF6800 outperforms GF6800GS far too often IMO, and the same card slightly clocked differently performs in reverse of the clocking?)

If they wanted to do a thorough review then all performance should have been investigated, their omissions just confirms that the more things change the more they stay the same.

At least they've finally moved to showing min fps, guess they've finally learned from Xbit, FiringSquad and even Tom's of late.

Still say FS's review is the best and most descriptive, especially after the AMD CPU review.

 

[Flakes]

thye have to run the tests multiple times to get accurate numbers in the outside areas...this is cause the grass is created from a random number gnereator so every run through has a different grass layout/pattern and therefore the FPS is different every run through hence they do the test several times to get the average.

 

[silentcoercion]

The discussion over there brought up a lot of things people should consider, as well.

As for no AA, they did it because the primary effect would be to lower frame rates across the board. Their goal was to show what was playable and what wasn't, and enabling AA would have very few setups, if any, registering as playable.

 

[TheGreatGrapeApe]

thye have to run the tests multiple times to get accurate numbers in the outside areas...

Yeah I don't really care, I didn't ask the question.

I expect good reviewers to run multiple tests anyways since 1 run can be affected by system peculiarities. For a personal benchie 1 run is fine, for websites that this is what they do, 2-3 runs would be minimal accpetable IMO regardless of such factors as the Oblivion grass.

 

[TheGreatGrapeApe]

As for no AA, they did it because the primary effect would be to lower frame rates across the board.

Well F'in Duh!

Do you know a time when AA increased framerates across the board?

They basically did a half-a$$ed review to get it up quicker IMO, and they did it long after the other sites so they had time. Not showing AA let alone HDR+AA simply shows that it's not a hardware investigation, just something for their own personal test. You figure out who that benifits, the consumer looking for info, or someone else's interests?

Their goal was to show what was playable and what wasn't, and enabling AA would have very few setups, if any, registering as playable.

Ahh, so they're deciding what's playable, despite the fact that someone may have a 1024x768 LCD and not have the 12x10/16x12 option, but does have the AA option, well thanks for saving us from too much information, never would've been able to filter it like I do for so many other sites.

I already have [H] and BitTech for the 'best playable' reviews, I don't need another one.

 

[sojrner]

nobody actually plays with them on because self shadow just looks awefull and grass shadow isnt really noticiable.

hmm... I play with grass shadow on. I like it...

I agree w/ your point though about no AA and the test not being good.

also agree w/ grape: the FS test so far is the best test for this game.

I notice alot of holes in the engine quality. Distance textures look horrible and the game does not use AF. I have to run a few mods (named on tweakguides.com) to get the distance to look palatable, which is retarded for a release level game, I also have to force aniso in the drivers... also bad on a game at this level. seems like many "glitches" are there b/c perhaps they spent more time on the console version? Why should I have to "manually" tweak a "next-gen" game in its .ini file to get it to look servicable at distance?

@silent:
"primary effect" is to improve image quality, which does lower frames, but that is not its purpose. my system plays w/ AA just fine, in fact w/ everything at max (self shadow off 'cause they DO look like dump) it runs great w/ AA set at 6x. (i think it is 6, whatever is next above 4...) at 1280x1024, and easily at 2x @ 1600.

 

[pauldh]

You said it yourself. I really like AnandTech reviews. It's nice to know the settings (Int. Shadows, Ext. Shadows, Shadow Filtering, etc) they used for the High quality and Medium quality. Also showing the min frame rate and the average on the same chart is nice...

It kinda pisses me off to know the 7900GT can't stay above 25fps in most outdoors tho

I like th FS review better. It was nice for them to spellout the settings they used, by FS included their .ini even, so that's more info if you dig into it. I liked to see the pic of where they tested, but sure wish sites would put themselves on the line and include a small low res video of their runthough. Also I like when they give a summary of their results rather than just the charts.

But the big problem I have is they didn't spell out how many runs they did. That coupled with the fact their results vary enough from what FS got. I'm leaning on the belief that they did only one run; which is totally unacceptable. Why, because of their wording about how to deal with close results as well as not mentioning the number of runs. If you poor the time into doing 5 or 10 runs for an average; you flaunt it to the readers. And also as Grape pointed out, no way can a GF6800 beat out a GF6800GS if they averaged many runs. The 6800GS is a much higher clocked 6800; how could it lose unless it was a fluke single run?

 

[Flakes]

thye have to run the tests multiple times to get accurate numbers in the outside areas...

Yeah I don't really care, I didn't ask the question.

I expect good reviewers to run multiple tests anyways since 1 run can be affected by system peculiarities. For a personal benchie 1 run is fine, for websites that this is what they do, 2-3 runs would be minimal accpetable IMO regardless of such factors as the Oblivion grass.

i tend to just hit the nearest reply button....

 

[pauldh]

Great points Grape. I missed the 6800>6800GS results. No way would that happen in GPU limited settings if they did multiple runs. IMO Oblivion needs 5 runs anyway. I've had consistent results in some areas where 5 runs all average the exact same number. But then in other map areas, each run can vary up to 3-4 fps.

What caught my eye was the X1800XT losing to the X1900XT by way more than in FS's review, which puts it down under the 7900GTX and not much ahead of the 7900GT. Who to believe, the site that provided their ini file and did 10 runs to average, or the site that neglected to mention whether they did one run or multiple runs? And IMO in no way is FS ATI biased; just look at their benchmark suite that is full games that favor NV. whereas Anand ...

Actually, I too would like to get way more info than may seem needed. Shoot, even doing 6 Oblivion map areas (2 in each area) 5 times each and giving an average.


Hey since you mentioned FS CPU review. I agree with some of the peoples comments to the article. I swapped out an A64 4000+ and put in a 3000+. I was using the 6800U at the settings I play at. The only area I tested was in a castlle with multiple NPC's. Anyway, the 30 second run I did coming off a save point yielded consistent results. The 4000+ San Deigo average 43 fps in all 5 runs while the 3000+ Venice average 38 fps in 4 runs and 36 fps in one run where it seemed to studder a little. Anyway, then FS's review came out the next day and they showed no difference at playable settings. Anyway, I think as was said in the comments, the cpu does come into play in towns with multiple NPC's, even at normal best-playable detials.

 

[TheGreatGrapeApe]

Who to believe, the site that provided their ini file and did 10 runs to average, or the site that neglected to mention whether they did one run or multiple runs?

Actually FS gives you a link to their ini file and Anand shows you their tweaks, so that part is somewhat consistent, but I like the fact that FS one you could cut and past into your ini folder and reproduce at their settings instead fo double checking each one. IMO changing grass size is a huge tweak for cards like mine.

Hey since you mentioned FS CPU review. I agree with some of the peoples comments to the article.

HeHe $2 if you can pick my 'Anon' comments. :twisted:

Anyway, I think as was said in the comments, the cpu does come into play in towns with multiple NPC's, even at normal best-playable detials.

And I would definitely have found it more interesting if they kept the minimum FPS throughout the tests, since while the average may be the same, perhaps on the dual core or faster CPUs there are less 1second dips. Just something I'd be more interested in confirming/denying.

 

[sojrner]

And I would definitely have found it more interesting if they kept the minimum FPS throughout the tests, since while the average may be the same, perhaps on the dual core or faster CPUs there are less 1second dips. Just something I'd be more interested in confirming/denying.

very good point... I am sure you are right there.

 

[Heyyou27]

The fact that they didn't enable HDR+AA just shows it's not a review about the game and the hardware but pitting company A against company B, and we already know which camp Anandtech is in.

Oh shut up ATI fanboy! :lol:

Yeah, they should have showed with antialiasing. I myself don't play Oblivion, but if I did, ATI would be the way to go.

 

[jodo]

Yeah those scores don't really reflect my gaming experience either.

So far MOST reviews show the 6800GT beating the 6800GS by a few points but since its the closest card to mine I use that to judge. I also am only running a 2.0Ghz 64 3200+ so im nowhere near the FX-60 in terms of CPU performance. I get at least 10FPS higher than Anand did max/min on my setup---so I'd love for them to tell me how that is considering I use the same settings they do..yes even self-shadowing/distances.

 

[pauldh]

HeHe $2 if you can pick my 'Anon' comments

Ha, that should be easy; I'll get right on that.

I didn't respond in that article (yet), but am usually good for 1 or 2 Anon's myself in most of their articles. 8) I like the fact that that Brandon takes the time to reply to most of the serious/well thought out statements/questions.




Edit: Hmm, on quick glance, didn't see the comment I expected: "Why in the world waste your time on testing AGP, anyone but the dumbest of Ludites have already made the switch to PCI-e" :lol:

 

[TheGreatGrapeApe]

Ha, that should be easy; I'll get right on that.

Just FYI 1 old 1 new.

I didn't respond in that article (yet), but am usually good for 1 or 2 Anon's myself in most of their articles. 8) I like the fact that that Brandon takes the time to reply to most of the serious/well thought out statements/questions.

Yeah they're pretty good for that.
It's a good idea and usually not abused.... too much.

 

[pauldh]

check my edit.

thx for the hint.

 

[TheGreatGrapeApe]

LOL!

If anything you'd see the opposite.

I also seem to have forgottent to add my question about the min fps in the first one. I wanted to post another comment and got distracted, so I never did ask that which still makes me wonder if multi-core/more power smooths out the momentary drops.

Time to go and re-ask.

 

[pauldh]

If anything you'd see the opposite

Hmm, really; I may have a guess then.

I myself would love to see some of the currently available AGP cards tested on a barton clocked at 2500+ & 3200+ speeds, and for comparison sake an A64 4000+ just to see how those rigs do. SHoot, make the A64 system a Asrock Uli 939 dual sata2 so we can throw in a X1900XT or even X1800XT for fun too.

 

[TheGreatGrapeApe]

Actually the last two on the list were mine, regarding elderscrolls.com and the one after it.


My first one was a few days back.

 

[pauldh]

Ok, I can believe that. Any more hints on the old one?

I agree that minimum fps while doing cpu scaling at normal playable settings would be very useful. It's weird but I have seen a few choppy moments on the 3000+ that I never noticed on the 4000+. And Fraps was used during each. Fraps hit, NPC hit, combination of both or simple unrelated...I am not sure. But the 3000+ does seem to occasionally take a noticable hit I did not see on the 4000+. But it still plays very well on the venice at stock 1.8GHz.


Hey, on a comical note: if he has been pulling late to all nighters playing/benching, then he is going to have to read that first paragraph in the last comment 5 times or more. I can picture him with his hands against his forhead, shaking his head back an forth followed by smacking the palm of his hand against his head, all while thinking "I need more sleep...addtional computational power moemntary drops in fps....argh I need to start over reading that again!"

I of course only had to read it twice.