8800GTX and 3870x2 - Shootout at the OK Corral

[mtl]

I agree 100%. Competition is key.

[pkirby11]

Look ATI has certainly had their lesser moments these last years but you gotta give them props. Look at most of all the review sites out there, they love this card for the most part given it does have some flaws. The fact that it has two GPU's is a moot point because the card cost as much or less than competing single GPU cards making it's performance to cost ratio very impressive.

I also like the whole fact that come March, I can pick another one up and do CrossfireX withour needing to change my motherboard. I wanted to get another 8800GT and do SLI but that would require me changing my motherboard along with buying the second 8800GT. Not to mention I get to keep my amazing Intel chipset board. What's not to like? I can buy two 3870x2 cards for less than what it would cost me to buy a new motherboard and 2 or 3 of NVIDIA's top selling card.

I look at this as a sign of things to come from AMD/ATI, which no one should see as bad. Competition is great, yes it took them forever just to best a 1 year old card but if they continue then hopefully they should be able to keep up and maybe even beat NVIDIA in the future which will in turn make NVIDIA get off their asses and make a new card. NVIDIA has known they have no competition and therefore we get the short end of the stick, it's been how long since we've seen a really ground breaking totally new product from NVIDIA? I doubt it's becuase they have nothing new, I'm guessing it's because they don't need to have anything new.

I'm hoping the wars heat up again, I'm excited and have decided to get a 3870x2 so I can go quad Crossfire. I will be picking a second up in March and I'm selling my 8800GT. But this doesn't mean I hate NVIDIA or want to see them loose, I only want to see them continue to improve so that one day I will go back to NVIDIA but only if they come up with better boards or allow SLI on Intel motherboards because I hate NVIDIA's Intel chipsets right now. :D

[Xion X2]

I call em like I see em. When ATI stops trying to put lipstick on a pig and comes out with a superior product I'll be happy to applaud. They have done great(9800 pro) in the past so I hope they do it again.
Really? Were you around when Nvidia came out with the inferior 7900GTX of last generation while the 1900XTX was spanking it in image quality, performance AND pricing?

I think we all know that AMD/ATI has the inferior product right now and do not need every single thing said about them, good or bad, turned into a negative. It is what it is; let it go.

The 2900XT I owned was neither as bad as people made it out to be nor as great as the fanboys made it out to be. Most things lie somewhere inbetween.

[noko]

Finally ATI after a year has something that performs slightly faster then a GFX :). If I was going to buy a high end card today then ATI would probably get my bucks but since my GFX still kicks good no need yet. I will be upgrading to an Intel Qual core since I am CPU limiting the GFX from the day I got it in Nov 2006. Crysis has been my only dissappointment for this cards performance. 8800 s turning out for me to be longer lasting in performance then my Radeon 9700 Pro. Christmass maybe a great time to buy another high end card, that is if I just don't go SLI ;).

[damicatz]

What nobody has brought up is that ATI cards STILL have broken shadows in at least 3 popular games(well quite a few people still play them) and that is BF2.BF2142 and Company of Heroes(and expansion). Game compatability goes to the GTX....

They have broken shadows because the game developers used a non-DirectX compliant method exclusive to NVIDIA hardware and patented by NVIDIA to render the shadows. (NVIDIA's implementation of PCF or Percentage Closer Filtering).

The fault rests on the developers of the game and not ATI. That's what happens when you join TWIMTBS (The Way It's Meant To Be Shilled) and don't bother to test your game on anything but NVIDIA hardware.

PCF is not part of the Direct3D specs, either for 9 or 10. So it has to be accessed through non-DirectX compliant and proprietary methods exclusive to each vendor's hardware. ATI did add their own implementation of PCF to the R600 GPU but it's not compatible with NVIDIA's implementation (which, again, is patented). A better solution would have been to use Fetch4 for shadow filtering which is part of the DirectX 10 spec, is accessible in DirectX 9 and is supported by both ATI and NVIDIA.

Also, the fact that the 3870X2 has two GPUs is irrelevant because it's transparent. The end-user is not made away of the fact that there are two GPUs nor do the drivers operate in Crossfire mode. Everything is done completetly behind the scenes and the 3870X2 acts the same as a single GPU card.

[Xion X2]

Also, the fact that the 3870X2 has two GPUs is irrelevant because it's transparent. The end-user is not made away of the fact that there are two GPUs nor do the drivers operate in Crossfire mode. Everything is done completetly behind the scenes and the 3870X2 acts the same as a single GPU card.
Is this entirely true? I thought the X2 functioned somewhat the same as Crossfire.

I haven't really been keeping up with the scene lately, but I could've sworn I've seen/heard of benchmarks varying wildly in some games as compared to others--as if Crossfire wasn't scaling correctly, or something. If I'm wrong, please let me know. It would improve my opinion of the X2 if it did in fact function as a single GPU card.

[noko]

This is really good news if indeed it is transparent and the operating system only sees one GPU and it works that way. ATI maybe ahead of the game for this for future R700 series where they just keep stacking the GPU's to increase the performance with more memory bandwidth. One GPU card for the budget card to multiple gpus on the high end cards. I really don't know but that sure would reduce a hell alot of costs only having to manuafacture one type of GPU for all of your cards. Massed producing one core vice sorta mass producing the high end cores may indeed bring prices way back down. I hope Nvidia is planning on something similar in the future.

[pkirby11]

Is this entirely true? I thought the X2 functioned somewhat the same as Crossfire.

I haven't really been keeping up with the scene lately, but I could've sworn I've seen/heard of benchmarks varying wildly in some games as compared to others--as if Crossfire wasn't scaling correctly, or something. If I'm wrong, please let me know. It would improve my opinion of the X2 if it did in fact function as a single GPU card.

From what I've read it's transparent in the sense that there is no settings to select or turn on. It is seen as a single card according the the ATI Catalyst drivers, there are no crossfire options. It tries to force AFR rendering for any game.

However it is still crossfire in the sense that if a game has issues with the driver than you get one GPU performance until ATI corrects the issues with drivers. Again though from what I've heard ATI is doing a knock out job with their driver support for this card along with the crossfire aspect too. So if they can keep it up not sure if this will be an issue but yes it's still a form of crossfire.

[Blacklash]

In games where Crossfire causes a performance hit vs a single card you will get that same behavior with the HD 3870 X2.

I compared it to HD 3850 Crossfire; Crysis on the DX10 path had the same flashing textures in the distance, TheWitcher stuttered badly, TRL with next gen content with AA gave horrid FPS vs single card, like 17 FPS with on board Crossfire on vs 40+FPS disabled. I could not break 20FPS in NWN2 with on board Crossfire working.

In cases where on board Crossfire causes stutter or a performance hit vs a single card you may disable it by turning of CAT AI which also kills all driver fixes for various titles.

I was going to get a 2560x LCD and use the X2 to push it. When I realized I might often get left with slightly less than single HD 3870 performance I changed my mind. Single HD 3870 performance is good for a 16x res and full HD 3870 X2 performance is not needed. An overclocked 8800GT or GTS will work that res fine all day and all night.

[H]ard was right about the X2 and sites pimping it via good timedemo results were wrong. I am just glad I lost less than 25usd trying one.


From what I've read it's transparent in the sense that there is no settings to select or turn on. It is seen as a single card according the the ATI Catalyst drivers, there are no crossfire options. It tries to force AFR rendering for any game.

However it is still crossfire in the sense that if a game has issues with the driver than you get one GPU performance until ATI corrects the issues with drivers. Again though from what I've heard ATI is doing a knock out job with their driver support for this card along with the crossfire aspect too. So if they can keep it up not sure if this will be an issue but yes it's still a form of crossfire.

[fivefeet8]

The fault rests on the developers of the game and not ATI. That's what happens when you join TWIMTBS (The Way It's Meant To Be Shilled) and don't bother to test your game on anything but NVIDIA hardware.


It's up to the IHV to provide the hardware to the developer and answer questions on issues as they arrise during development. I've seen more than one developer talk about the issues they've had with ATi and their DevRel support. One even went so far as to say they couldn't get any ATi hardware to test with until after the game went final. Nvidia provides a whole building filled with every Nvidia hardware(old, new, future) for the developers to use to test their code with during the development cycle.

If anything, proprietary functions being used by developers for the benefit of Nvidia hardware speaks more about the lack of DevRel by ATi. Time and time again, good features by ATi have been left by the wayside simply because of poor DevRel.

[pkirby11]

In games where Crossfire causes a performance hit vs a single card you will get that same behavior with the HD 3870 X2.

I compared it to HD 3850 Crossfire; Crysis on the DX10 path had the same flashing textures in the distance, TheWitcher stuttered badly, TRL with next gen content with AA gave horrid FPS vs single card, like 17 FPS with on board Crossfire on vs 40+FPS disabled. I could not break 20FPS in NWN2 with on board Crossfire working.

In cases where on board Crossfire causes stutter or a performance hit vs a single card you may disable it by turning of CAT AI which also kills all driver fixes for various titles.

I was going to get a 2560x LCD and use the X2 to push it. When I realized I might often get left with slightly less than single HD 3870 performance I changed my mind. Single HD 3870 performance is good for a 16x res and full HD 3870 X2 performance is not needed. An overclocked 8800GT or GTS will work that res fine all day and all night.

[H]ard was right about the X2 and sites pimping it via good timedemo results were wrong. I am just glad I lost less than 25usd trying one.

Funny but that's the second time I've read HardOCP and realized they claim to do "real world benchmarking" unlike other sites. That's funny, because most site I go to Anandtech, Toms Hardware, Firing Squad they all seem to back each other in their tests but some how HardOCP does the "real world" test and suddenly the card sucks and negates three reputable sites. What ever, if you've tried one than maybe you know what you are talking about but you've posted no proof other than poorly worded statement.

Anyways, I'm sure the card isn't perfect by any means but it's definetaly a step in the right direction for ATI. I for one tend to read multiple review sites along with forums and other input and most has been quite good. Especially in the department of drivers, ATI seems to be getting props for thier 3870x2 drivers. "Real World" benchmarks, that's funny. :)

[mtl]

http://enthusiast.hardocp.com/article.html?art=MTQ2MSw2LCxoZW50aHVzaWFzdA==

Hardocp is the best review site going! ATI fanboys don't like it at the moment because it shows ATI's real world game play shortcomings. Also everyone knows the founder of Tom's went to work for ATI.

[Shocky]

Funny but that's the second time I've read HardOCP and realized they claim to do "real world benchmarking" unlike other sites. That's funny, because most site I go to Anandtech, Toms Hardware, Firing Squad they all seem to back each other in their tests but some how HardOCP does the "real world" test and suddenly the card sucks and negates three reputable sites. What ever, if you've tried one than maybe you know what you are talking about but you've posted no proof other than poorly worded statement.

Anyways, I'm sure the card isn't perfect by any means but it's definetaly a step in the right direction for ATI. I for one tend to read multiple review sites along with forums and other input and most has been quite good. Especially in the department of drivers, ATI seems to be getting props for thier 3870x2 drivers. "Real World" benchmarks, that's funny. :)

Well arguably the most respected site there anandtech actually benchmarked cut scenes and flybys, they have nothing to do with the performance of the game or gameplay, for me anandtech have lost allot of credibility.

As for the other sites benchmarking methods, well I have no idea as allot of them don't share that information so it’s hard to know how reliable they really are, but make no mistake it's not just [H] getting conflicting results.

http://www.pcper.com/article.php?type=expert&aid=512
http://plaza.fi/muropaketti/artikkelit/naytonohjaimet/amd-radeon-hd-3870-x2-r680,2
http://www.bit-tech.net/hardware/2008/02/01/bfgtech_geforce_8800_gts_oc_512mb/1

[fivefeet8]

Is this entirely true? I thought the X2 functioned somewhat the same as Crossfire.

I haven't really been keeping up with the scene lately, but I could've sworn I've seen/heard of benchmarks varying wildly in some games as compared to others--as if Crossfire wasn't scaling correctly, or something. If I'm wrong, please let me know. It would improve my opinion of the X2 if it did in fact function as a single GPU card.

The card functions like a single GPU to the end user because there are no extra connections or extra driver controls for it. But it's basically an internal Crossfire PCIe1.1 bridge between both GPU's. It's going to behave similarly to traditional crossfire cards in terms of performance and rendering issues.

The transparency to the end user may have come at a cost to flexibility though as there is no direct way to disable the crossfire mode in situations where it causes issues. Hopefully, ATi will provide a driver with that option in the future.

[pkirby11]

Well arguably the most respected site there anandtech actually benchmarked cut scenes and flybys, they have nothing to do with the performance of the game or gameplay, for me anandtech have lost allot of credibility.

As for the other sites benchmarking methods, well I have no idea as allot of them don't share that information so it’s hard to know how reliable they really are, but make no mistake it's not just [H] getting conflicting results.

http://www.pcper.com/article.php?type=expert&aid=512
http://plaza.fi/muropaketti/artikkelit/naytonohjaimet/amd-radeon-hd-3870-x2-r680,2
http://www.bit-tech.net/hardware/2008/02/01/bfgtech_geforce_8800_gts_oc_512mb/1

Curious but how are "In Game" cut scenes not CGI intro's not going to give you the same results? I have the Witcher and I have FRAPs, I turned it on with my 8800GT and guess what the performance was about the same if not a little lower in the "In Game" cutscene. It uses the same engine and it still requires 3D Rendering, not sure about COD4 seeing as I don't have that game.

Look you can say what you want, I'm not saying this is the greatest card in the world but I am giving ATI props for making a kick ass product with great driver support so far. Also if you read the end, HardOCP didn't use the latest drivers given to them by ATI, they only used it on Crysis because that's all they were told it would improve. Funny however that every review site I mentioned did test with the latest drivers. It's funny how some sites are eager to test with steriod induced drivers from one company but not from the other. I'm sure those newer drivers had some improvements in Crossfire rendering than just Crysis.

Well anyways, my original statement stands true, for me the 3870x2 is a better buy. It's peformance is as good or better than a single Ultra, is cheaper and won't require me buying a new motherboard I don't need or want. So yes this time in my books ATI has a winner and soon I venture NVIDIA will have a winner with the 9800GX2 but I still can't SLI them in my Intel board. :D

[fivefeet8]

Curious but how are "In Game" cut scenes not CGI intro's not going to give you the same results?

Depends on the ingame cutscene being benchmarked. It's much easier to optimize for specific rendering if you know ahead of time what's going to be rendered. Most ingame benchmarks render the same thing every run, while when you actually play the same level, enemies and objects are randomized.

A good example of an ingame benchmark that doesn't render the same objects is the Lost Planet ingame demo benchmark. Every run, the demo randomly generates the monsters and they don't do the same things every run either. This is exactly what happens when you play through the level as well and as such, it's ingame benchmark result is very close to it's actual gameplay.

Arguments can be made either way about how an ingame benchmark should be, but it does highlight the differences that can happen between ingame cutscenes demo benchmarks and actual gameplay.

"Crysis was probably the game that gave us the most headaches during our testing, as AMD released a new driver just before launch that claimed to drastically improve performance in the title—I mean, by almost 60 percent. However, no matter what we did, the game didn’t perform much better than it did with the older driver in our demo (around 1.5 fps better to be specific).

We decided giving the Crysis GPU benchmark a go to see whether that was where the performance advances were and, low and behold, there was an almost 60 percent performance increase between the two drivers. We even went so far as to manually play through the same section of the game that we’d benchmarked, just to check that it wasn’t something on our end that was the problem. The result was pretty much exactly the same and the difference between the two drivers was roughly two frames per second in favour of the newer driver.
" - bit-tech (http://www.bit-tech.net/hardware/2008/01/29/amd_ati_radeon_hd_3870_x2/7)

EB's reviews the 3870x2 vs 8800gt SLI (http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=531&Itemid=27&limit=1&limitstart=8)

[Blacklash]

Would anyone that owns a 3870 X2 and TRL shoot the pool room in Croft manor with FRAPS? I no longer have mine.

With AA enabled and next gen content on I got 23FPS with on board Crossfire off and 17FPS with it working. That's not a typo, off was higher than on.

It's easy to do, just enter the pool room and drop the camera down so you can see the HDR lighting up top. Below is a cap of the place and orientation I am talking about. The shot was taken with a 8800GT @ 700|1000 1700. It was next gen on, in game AA enabled and @ 1680x.

http://img149.imageshack.us/img149/4061/trl2008021117004500yf9.jpg

I've been using Vista x64 on all my computers. I was curious if other OS had the same results with the X2 and TRL. I also got very low FPS with the X2 on the first level near a lot of running water.

EDIT-

Hanners recently showed the same behavior in Need For Speed: Pro Street-
http://www.elitebastards.com/cms/index.php?option=com_content&task=view&id=531&Itemid=27&limit=1&limitstart=6

The same may be said for SupCom @ techPowerUp!-
http://www.techpowerup.com/reviews/HIS/HD_3870_X2/15.html

Gothic 3 also behaves the same way-
http://www.computerbase.de/artikel/hardware/grafikkarten/2008/test_ati_radeon_hd_3870_x2/12/#abschnitt_gothic_3

So this isn't isolated to TRL. There may be other titles I am not aware of.

[Xion X2]

It's easy to do, just enter the pool room and drop the camera down so you can see the HDR lighting up top.

Scenes with that type of lighting give R6xx fits for some reason. I remember mine slowing down like that.

[KasuCode]

dont have that game.

[Tho Jo Smale]

Gothic 3 also behaves the same way-
http://www.computerbase.de/artikel/hardware/grafikkarten/2008/test_ati_radeon_hd_3870_x2/12/#abschnitt_gothic_3

So this isn't isolated to TRL. There may be other titles I am not aware of.


I call bs, you see that the 8800GTS 512 wins all the benchies on that site. I mean come on, that guy must have a G92 fetish er something. No way does a GTS 512 beat out the Ultra and GTX at every single resolution.



JC

[tejas]

The 8800 GTX is a great card but the 3870X2 pretty much wipes the floor with it and in some cases the Ultra and gets competitive with the 8800GT SLi. Not too bad for a card with lower power draw on Idle and Load than the aforementioned Nvidia cards.

As for reviews, I swear by Anandtech, but also go to Toms Hardware and HardOCP in order to verify Anands results. Kyle Bennett (HardOCP) has some strange and unscientific methodology of testing GPU hardware tho...

Anyway Im not a fan of Nvidia after having TWO 780i mobos die on me within a week. Who needs SLi when you can have 2x RV670 cores on one board!!!Long live Intel and long live AMD! :)

[Blacklash]

The 8800 GTX is a great card but the 3870X2 pretty much wipes the floor with it


You're entitled to your opinion and my personal experience with both cards refutes that statement. I know what my GTX will do from title to title and with the X2 I could get better than HD 3870 Crossfire performance or less than a stock HD 3850. IMO its performance is not consistent enough to be called the new champ. If it were predictable and reliable, I would agree. Even in titles where it excels it often has poor min frames and odd behavior of erratically speeding up and slowing down seemingly without reason.

According to one fellow @ Rage using AA in some titles with the X2 will disable the second GPU. WiC on the DX10 path was provided as an example-

http://www.rage3d.com/board/showpost.php?p=1335352515&postcount=138

I've tried SLi|Crossfire and Multi-gpu a few times and I am not currently a fan of any of them.

EDIT: X-Bit's new review of the X2-

"The new card won six out of the nineteen tests, at least at high resolutions...

On the other hand, it was either equal to or slower than the single-chip ATI Radeon HD 3870 in seven out of the nineteen tests..."

http://www.xbitlabs.com/articles/video/display/ati-radeon-hd3870-x2_19.html#sect0

Just like I've been saying from my experience with the X2; inconsistent, unpredictable and not a new champ.

[K007]

These results will only make Nvidia work harder on there X2/Driver support....good win for all of us imo..

[Blacklash]

Scenes with that type of lighting give R6xx fits for some reason. I remember mine slowing down like that.

I remember you saying edge smoothing in Jericho hurt your card badly too. I had the same experience with an X2.

[Sazar]

His entire review seems on the contrary to everyone elses. However. He is a reviewer and I am not. So that is all I can say. I wish I could get one of these and test the livin bejeezes out of it myself with 2 dozen of my own games and see how I feel about it personally.

No one really raised any red flags. Just some cautionary stuff regarding "well this IS a dual GPU card and AMD will have to REALLY work hard to keep it working well into the future blah blah blah". Perhaps they are right? Perhaps that is where AMD is going and all will be well.

14 out of the 25 games tested were faster on the 3870x2. 4 were a tie so that's 18 were good to go. 6 were just below the GTX in performance and Supcom was broken.

For a launch at the same price that is pretty good and promising to say the least in terms of AMD investing the energy and support that they should.

C.

What? Come on, didn't you read how Hardocp is re-inventing the way reviews are done :wonder:

I mean, they've only been the ones who have put out some of the the most bogus and self-righteous reviews in history, but now they have changed their ways and since they say they are right, they are right, regardless of the caveats they introduce into the beginning of their reviews that no one bothers to read and make little to no mention of said caveats at the end.

The entire nv30 affair and their future trashing of synthetic benchmarks sullied their name and they haven't done anything to redeem themselves other than saying "we're better than everyone else".

Anyways, back to regular programming.