ATI Radeon HD 3870 X2 @ [H]

Status
Not open for further replies.
McCartney said:
I'm just surprised that so many people came on and started taking shots at you Kyle... I guess people knew this was ATIs last chance to really have a uprising and maybe live another day. I guess when hell's coming we're all going to pray
Click to expand...


NO, I THINK people was wondering how kyle got one set of number and most other sites got totally different numbers
 
havand said:
Kyle, do we know if we'll be able to use more than 1 of these cards? will the crossfire on crossfire action play nice? (if this has been state, I apologize. I've read most, but not quite all of this thread)
Click to expand...

Simple answer is yes. Just a matter of waiting for drivers.
 
POPEGOLD said:
NO, I THINK people was wondering how kyle got one set of number and most other sites got totally different numbers
Click to expand...

Shouldn't they also realize that Kyle is the only one left that hasn't sold out his kind and actually talks with members of his board? Unlike Anand I see kyle being very active helping anyone that asks
 
heatlesssun said:
So if you already have an 8800 Ultra, would anyone trade it for a 3870x2?
Click to expand...

I wouldn't. The 8800 drivers aren't even close to finished yet. Just imagine when these Vista drivers get fully-tuned. Since XP is EOL this year (Isn't it?) I'd just wait and see.
 
heatlesssun said:
So if you already have an 8800 Ultra, would anyone trade it for a 3870x2?
Click to expand...

I wouldn't. Even in the scenarios where it might be faster than the Ultra, it isn't faster by enough of a margin to justify the cost in my opinion.
 
Another good review. I am glad I just picked up a 2900Pro to supplement me while this whole GPU war gets worked out. Looks like no great strides for ATI, but who knows what future drivers will provide.
 
Im 33 years old and been gaming forever and believe me , i learned my lesson by not buying into all the hype over the years thats why i stick with HardOcps methods of testing out new Video Cards, Mobos and Processors. HardOcp gives us real world performance and thats the bottem line.
 
McCartney said:
I wouldn't. The 8800 drivers aren't even close to finished yet. Just imagine when these Vista drivers get fully-tuned. Since XP is EOL this year (Isn't it?) I'd just wait and see.
Click to expand...
Yeah, June or July I think.
http://www.infoworld.com/article/08/01/28/04NF-save-xp-license_2.html

Doesn't mean I'm upgrading to vista anytime soon though. And when I do its going to be dual boot with vista on secondary

Plus, MS says that you can buy a vista license (ultimate or business) and use that key to "downgrade" to an XP license.

So even if they stop selling XP, there are still ways to get it.

Edit: I don't think its EOL then though, thats just when they stop selling it. From what I recall, they usually support older OS for at least 5 years after they officially stop selling it.
 
All the panty bunching aside, while I was hoping for better performance, I still plan on pickup up a pair. I've wanted to switch to an Intel platform, and not lose ground compared to my GTX's. While the graphics performance may be fairly close, it's certainly not the step back that regular ol' 3870's would have been. Sure, new drivers may bump the performance a little, but then again, don't they almost always?
 
Meh Vista runs good enough on this PC for me to not complain. Newer hardware typically suffers less of a hit than older hardware.
 
McCartney said:
Shouldn't they also realize that Kyle is the only one left that hasn't sold out his kind and actually talks with members of his board? Unlike Anand I see kyle being very active helping anyone that asks
Click to expand...


anand readers on his forum feel very strongly about anand...and not much about kyle.... anad i have seen anand on his forum alot...along with his editors

i think they both are great....
 
I haven't seen anand keep touch with his viewers at all. I"m East INdian too so believe me I'd root for Anand if I could because there are less of us in this market, but he's just a snob.
 
POPEGOLD said:
NO, I THINK people was wondering how kyle got one set of number and most other sites got totally different numbers
Click to expand...

And you should wonder. And I think we have done a good job of stating our case. It is not my fault if everyone else is dong it wrong. ;) I have lead the front on this for years now and the simple fact of the matter is that many sites will not commit to doing it this way is because of the huge investment in resources it takes.....what did Anandtech say this morning?

Unless the reviewer who ran the gameplay test has played the entire game and logged framerates of every scene, situation, weapon, and event and can run a test that hits all the most intense portions of the game (this is not possible in most cases) a gameplay test will be no better (and in most cases worse) than a timedemo that skips AI / sound / physics etc. that can take a better look at what happens when graphics are the most important element in rendering smooth framerates.
Click to expand...

Now while I do not agree with his thoughts, as he thinks timedemos fully represent gameplay....he asserts that it is not possible to test the way we do. Why? Because they do not want to put the time and money into it. It IS possible, we have done it for years now. Other sites still do "video card reviews." We analyze the gameplay experience supplied by the video card. There is a big difference in those philosophies.
 
McCartney said:
I haven't seen anand keep touch with his viewers at all. I"m East INdian too so believe me I'd root for Anand if I could because there are less of us in this market, but he's just a snob.
Click to expand...

i havent read ANAND's forum alot of late... so maybe u are right....maybe he does think he is the uber-shyt
 
Is ther any mention of ati including a crossfire indicator in the ccc yet? Something like nvidia have would be handy, ati have yet to include that for some reason.

Also quad gfx setups, bene reading some people are saying its pointless because you would be "fighting input latency". Any truth to this or is the latency not that noticable?
 
emailed him 4 times in like 3-4 months just asking about his career since, you know, our kind aren't huge and he just didn't respond. I don't care how busy he is, kyle is just as busy but he has time for us. That's the ultimate reward is when kyle feels he's like one of us. Anand I've never seen participate actively in anything regarding warranties etc.
 
Chris_B said:
Is ther any mention of ati including a crossfire indicator in the ccc yet? Something like nvidia have would be handy, ati have yet to include that for some reason.

Also quad gfx setups, bene reading some people are saying its pointless because you would be "fighting input latency". Any truth to this or is the latency not that noticable?
Click to expand...


Dude.....I have been BEGGING for this for literally over a year now, they tell me "yes" to my face, but we still never see it show up.

I think the big factor that will hold the cards back in quad setup is having 512MB of ram per GPU......Will be interesting to see what happens.
 
when it comes to connecting with his readers... no question KYLE pwns all sites

chris tom over at AMDZONE is pretty good to at reaching out... in a similar way

I dont visit TOMS forums because of the MODS... they are the worse on the net

mod there ... jackbarnes gave me a nasty warning for editing his edit...which i didnt do ofcourse...when i called him on it he emailed me a bunch of curse words and banned me....

this forum is nice n cozy..... and i think its bcause KYLE takes time to reach out...
 
hopefully those other reviews giving the X2 high marks will push Nvidia to move up their high end. I still believe that the [H]'s reviews are more accurate though
 
From article:

It is also worth mentioning that our cards performed flawlessly in multi-display configurations giving us the flawless support that should be expected with a product of this price.
Click to expand...

This is a good thing. Hopefully NV will start allowing this with SLI...

Thanks for the article.
 
Dan_D said:
I wouldn't. Even in the scenarios where it might be faster than the Ultra, it isn't faster by enough of a margin to justify the cost in my opinion.
Click to expand...

Thanks for the feedback. The 3870 x2 looks to be a good card for a decent price. But at the same time, this is not really a new level of performance, which is what many of us really want to see now out of a single card.

Its time for the next gen!
 
67shelby said:
What is the exact difference between capturing actual gameplay with FRAPs and then replaying it in order to generate benchmark results and creating a timedemo of actual gameplay and replaying it in order to generate benchmark results? On one hand you get numbers from FRAPs that are generated with the associated overhead and on the other you get results generated from the actual engine without overhead, right?

I understand the difference in using the games built-in benchmark (Crysis for example) and actual gameplay, but what is the difference between a FRAPs gameplay capture and a timedemo capture of actual gameplay in ET:QW as an example?

Because AT uses FRAPs for certain games (just like you), custom timedemos of actual gameplay for others, and then the built in benchmarks (assuming you call these canned demos).
Click to expand...

i dont know if a custom time demo is considered canned... Kyle will have to break down CANNED for us as per [H] methodology
 
67shelby said:
What is the exact difference between capturing actual gameplay with FRAPs and then replaying it in order to generate benchmark results and creating a timedemo of actual gameplay and replaying it in order to generate benchmark results? On one hand you get numbers from FRAPs that are generated with the associated overhead and on the other you get results generated from the actual engine without overhead, right?

I understand the difference in using the games built-in benchmark (Crysis for example) and actual gameplay, but what is the difference between a FRAPs gameplay capture and a timedemo capture of actual gameplay in ET:QW as an example?

Because AT uses FRAPs for certain games (just like you), custom timedemos of actual gameplay for others, and then the built in benchmarks (assuming you call these canned demos).
Click to expand...

Those games using AI that can do different things on each run probably wont be suitable for a timedemo as for example, you may end up shooting at thin air instead of the enemy whilst getting pasted by them from the sides.
It might be ok for some highly scripted games though.
 
67shelby said:
What is the exact difference between capturing actual gameplay with FRAPs and then replaying it in order to generate benchmark results and creating a timedemo of actual gameplay and replaying it in order to generate benchmark results?
Click to expand...
FRAPS is real-time, whereas timedemos are not real-time. A timedemo is a number of frames (say 3,840) that are rendered sequentially as fast as the system allows. Take the number of frames (3,840), the time it took to render those frames (say it took 90 seconds), and you have your 'frame rate' (42.6fps).

FRAPS doesn't record gameplay and replay it, it simply logs the actual frame rate at intervals, and these logs are used to build a graph.
 
HMM, World in conflict, i managed to do like 70 avg fps @ 3.4 ghz amd 2900 XT, that was the benchmark, the game was, higher fps, i dont usually experience benchmarks going better than the real game, im talking bout in game benchmarks, but what driver did you guys use, tomshardware mentioned something 108B and 123A or something.

its all about how the numbers are gathered though,could ya guys add 1280x1024 1440x1050 or something lower aswell, would help us low res alot, i still doesnt understand the point in 24" lcd @ 1920x1200 or something like that, hmm, on lan its like, ey go COD4, uhm, i cant play it, cause lcd = must have max res, if not, it looks just ugly aswell.(blurry)
my 19" 1280x1024 will last for some more months years who knows.

but bring world in conflict in aswell, the game require alot of texture memory over 320 mb, as we saw in world in conflict benchmarks when the 8800 gts 320 mb didnt perform.

anyhow, my 2900 XT had overclocking potential, and i dont know bout the 3870 nor the 3870 X2, but if they manage to be as lovely as the 2900 XT, i'd be really happy, im talking while having temperatures at 40 C with max voltage etc.

i took my 2900 XT from 743mhz core and 828 mhz memory to 965 mhz core and 1200 mhz memory.
the world record in 3dmark is done with 980//1200 and CF ofc.
previous 3dmark 06 world record.

i hope theese cards are equally impressive in overclocking =)
 
Another great review.

You know with every other review you do, you piss off the 33% of people who are NVIDIA fanboys and the next review you piss off the other 33% of people who are ATI fanboys.

The other 33% really like the reviews ! :)
 
The Fact that Anand's editor stated they dont use "Canned" benches and no cutscenes and then gets corrected by a forum user that Anand did use a Cutscene for COD4 benchmark is a bit woozie in itself. Its quite funny but Im taking the median from all these tests, Guru 3D stated that the card in a couple of gamse surpassed a 8800 GTX and could maybe in those same games a 8800 Ultra. To me this card has not surpassed the 8800 ultra from the benches i saw at guru3d, here and anand. I think that is almost safe to say.

So it took AMD 2 GPU's in one card to tie a 8800 GTX- at the same price- If I was going to buy one today- I would buy an AMD why? simply cuz its newer. But if i had the money i wuold buy an 8800 ultra simple...

Infact i think people would be much better off buying 8800 GT SLI instead. which is about 50-75 bucks more. And if your an extremist (like me:) ) buy 2 8800 GTX or 2 8800 Ultras.
Case solved.
 
67shelby said:
What is the exact difference between capturing actual gameplay with FRAPs and then replaying it in order to generate benchmark results and creating a timedemo of actual gameplay and replaying it in order to generate benchmark results? On one hand you get numbers from FRAPs that are generated with the associated overhead and on the other you get results generated from the actual engine without overhead, right?

I understand the difference in using the games built-in benchmark (Crysis for example) and actual gameplay, but what is the difference between a FRAPs gameplay capture and a timedemo capture of actual gameplay in ET:QW as an example?

Because AT uses FRAPs for certain games (just like you), custom timedemos of actual gameplay for others, and then the built in benchmarks (assuming you call these canned demos).
Click to expand...

I went to BB to buy some canned demo's & they told me to try the supermarket. :p
Went there & looked by the sardines & veggies but couldn't find them. :eek:
Maybe they were by the fruit cocktail :rolleyes:

Yes, most all those other reviews use the built in demo crap. I do think that the manufacturers take advantage of those. As proof by [H]'s crysis review. He said that the demo increased 60% with the newest drivers, while gameplay was just a few FPS.
So i'd say they were optimizing the drivers for the demo.

Cheers & keep up the great reviews [K]yle & [K]rew. :D
 
Marcdaddy said:
Im 33 years old and been gaming forever and believe me , i learned my lesson by not buying into all the hype over the years thats why i stick with HardOcps methods of testing out new Video Cards, Mobos and Processors. HardOcp gives us real world performance and thats the bottem line.
Click to expand...

They cant give "real world performance", they can only tell you how they think their sample cards perform, at best
 
I think [H] could do a better job of benchmarking by including the "canned" benchmarks alongside the "real world" benchmarks.

I generally don't like the [H] video card reviews for the simple reason that they don't show apples to apples in their reviews.... The "highest playable settings" thing is not always acurate.

I remember quite a few times where certain drivers coupled with certain games would play horrible at certain resolutions but great at higher resolutions. I've also seen quite a few games where turning off or on a certain "feature" would more than double the performance.

There is a LOT more to doing accurate reviews than "real world" benchmarks.

You also have to take into consideration how long Nvidia has had to get the bugs worked out of their drivers compared to AMD/ATI.

And no.. I am NOT an AMD/ATI fanboy as you can tell by my sig. I go with whatever I thing is the best price/performance ratio at the time and also by what I will need.

I really woudn't mind having one of those 3870X2 cards though..... they would be very nice for GPGPU programming.
 
gljvd said:
You don't make sense though.

I would agree with you had a way to get the same run through each time. However you don't.

I could understand using a time demo that you guys created to run through a section of the game. That way it would be the same each time. However since you don't your using two diffrent sets of data to come to a flawed conclusion.
Click to expand...

You are essentially saying there are no tests that will accurately demonstrate the performance of the cards.
If the AI in a game can make different decisions on each run, a timedemo wont cope unless it is scripted.
This defies the whole point as scripted actions wont use the AI engine which does affect performance.

In which case, whats the next best thing? Its [H]'s method, not canned benchmarks.
 
gljvd said:
You don't make sense though.

I would agree with you had a way to get the same run through each time. However you don't.

I could understand using a time demo that you guys created to run through a section of the game. That way it would be the same each time. However since you don't your using two diffrent sets of data to come to a flawed conclusion.
Click to expand...

You're assuming that "not identical each time" equals "not useful for comparison" and that is both simplistic and untrue. A certain amount of variance does not make the results meaningless. And even the supposedly "scientific" canned benches don't get the same results each time. They vary from one run to the next on the exact same hardware, replaying the exact same frames. Every site that uses the demo method states soemthing to the effect of "we ran this three times and averaged the results." So, by your standard, their scores are "meaningless" too.

Surely you have played a section of a game more than once yourself, where the AI did not behave exactly the same and neither did you, but you experienced the same performance issues each time. Then you had to go in and lower a setting somewhere to get acceptable performance in that game scenario. Then you played that part of the game again and it was okay.

If you have ever done that, congratulations--you've done a real-world gameplay evaluation, and you've demonstrated the value of [H]'s method in the process. If you haven't, then I would conclude that you don't play games much, in which case I wonder why you're here. But that's the point of [H]'s method--they do what you would do if you were playing a game on a new card and trying to figure out how high it could go. "Keeping score" (grinding out artificial statistics) is meaningless unless you view hardware like a sports fan would view his favorite teams. Which is why they call it "fanboy."
 
Nenu said:
You are essentially saying there are no tests that will accurately demonstrate the performance of the cards.
If the AI in a game can make different decisions on each run, a timedemo wont cope unless it is scripted.
This defies the whole point as scripted actions wont use the AI engine which does affect performance.

In which case, whats the next best thing? Its [H]'s method, not canned benchmarks.
Click to expand...

I disagree. Hardocp can never reproduce these scores. With canned benchmarks they can.

I agree that using canned bencmarks that come with the game is a bad idea , however using canned benchmarks of your own making where ati / nvidia can't optimise for it then there sholdn't be a problem.
 
Commander Suzdal said:
You're assuming that "not identical each time" equals "not useful for comparison" and that is both simplistic and untrue. A certain amount of variance does not make the results meaningless. And even the supposedly "scientific" canned benches don't get the same results each time. They vary from one run to the next on the exact same hardware, replaying the exact same frames. Every site that uses the demo method states soemthing to the effect of "we ran this three times and averaged the results." So, by your standard, their scores are "meaningless" too.

Surely you have played a section of a game more than once yourself, where the AI did not behave exactly the same and neither did you, but you experienced the same performance issues each time. Then you had to go in and lower a setting somewhere to get acceptable performance in that game scenario. Then you played that part of the game again and it was okay.

If you have ever done that, congratulations--you've done a real-world gameplay evaluation, and you've demonstrated the value of [H]'s method in the process. If you haven't, then I would conclude that you don't play games much, in which case I wonder why you're here. But that's the point of [H]'s method--they do what you would do if you were playing a game on a new card and trying to figure out how high it could go. "Keeping score" (grinding out artificial statistics) is meaningless unless you view hardware like a sports fan would view his favorite teams. Which is why they call it "fanboy."
Click to expand...


However in canned benchmarks the diffrences are slight.

With hey lets randomly play through this level test things can change drasticly. For instance what if both cards do badly with smoke and in one play through the ai fires off smoke nades or rockets while in the next run through they don't ? Or any number of diffrent combinations .
 
Status
Not open for further replies.
Back
Top