Crossfire beats SLi in Doom3!!!

Jbirney said:
They said something about this

"Den anvendte Tft-skzrm understbttede desuden desvzrre ikke 1600x1200"

Not sure what langauge this was in but if you look closer they were using an LCD and just maybe that LCD did not support 1600x1200 which could explain why those benches were not there. As you know not all LCDs support that res...

I will translate that for you
The danish frase:
"Den anvendte Tft-skærm understøttede desuden desværre ikke 1600x1200"
Means:
"The TFT-monitor used did unfortunatly not support 1600x1200"

So you where spot on..and I even missed that part...and I am a dane :eek:

Terra - Embarrased...
 
jon_k said:
yeah but his point still stands. The guys who wrote that crap are the biggest !!!!!!s I've ever seen. Everyone who buys a new card these days is either playing @ 1280x1024 (19in LCD) or 1600x1200 (CRT or 20in+ LCD). IMHO anything lower is worthless unless you're doing a CPU benchmark.

The monitor they used didnt support 1600x1200

jeez, just because it doesnt cater to your thinking, doesnt mean they are !!!!!!s.

I doubt ATI or whatever would let them take this home with them, they probably only had time for 1 quick benchie on whatever was near
 
No - I don't think so... but could be... who knows... but we /all/ know most of the real benefits of SLi usually only kick in at very high resolutions, and I bet the same goes for CF, so it is a bit puzzling that they left those out... anyways... still wating for CF to hit the streets (for reals) before I think too much more about it...
 
Hate_Bot said:
The monitor they used didnt support 1600x1200

jeez, just because it doesnt cater to your thinking, doesnt mean they are !!!!!!s.

I doubt ATI or whatever would let them take this home with them, they probably only had time for 1 quick benchie on whatever was near

oh shit... lol. Well, that's a good reason... I guess it /is/ one of those 20" 16:10 ratio monitors... Well, as I said before, I had bet CF is like SLi, in that, they both will shine at the higher resolutions, so even if their monitor did do 1600x1200, I woudn't think they were trying to hide anything by not running the test...
 
revenant said:
No - I don't think so... but could be... who knows... but we /all/ know most of the real benefits of SLi usually only kick in at very high resolutions, and I bet the same goes for CF, so it is a bit puzzling that they left those out... anyways... still wating for CF to hit the streets (for reals) before I think too much more about it...

Maybe I'm just paranoid...but it seems awefully strange to me to have soul access to ATi's Crossfire (which costs over $1500) and then not have a monitor to run 16x12...The whole point of SLi and CF is to run at stupid high resolutions...they HAD to know that...

And they're in fucking Taipei...they can walk to the corner market and buy a 1600x1200 monitor...

EDIT: And I do think it would have made a difference because you can already see SLi pulling away at the highest resolution they ran it at...
 
^eMpTy^ said:
Maybe I'm just paranoid...but it seems awefully strange to me to have soul access to ATi's Crossfire (which costs over $1500) and then not have a monitor to run 16x12...The whole point of SLi and CF is to run at stupid high resolutions...they HAD to know that...

And they're in fucking Taipei...they can walk to the corner market and buy a 1600x1200 monitor...

yeah - I was thinking that also... I had bet there were a few of them about at Computex. lol... I am guess they were just being lazy...

EDIT: SLi did have the edge @ the highest res highest IQ test that they did run, though... :)
 
^eMpTy^ said:
Maybe I'm just paranoid...but it seems awefully strange to me to have soul access to ATi's Crossfire (which costs over $1500) and then not have a monitor to run 16x12...The whole point of SLi and CF is to run at stupid high resolutions...they HAD to know that...

And they're in fucking Taipei...they can walk to the corner market and buy a 1600x1200 monitor...

EDIT: And I do think it would have made a difference because you can already see SLi pulling away at the highest resolution they ran it at...

to me it sounds like they were lucky enough just to get to run that one game, they probably didn't have much time to do it in either

if it were me i wouldn't be greedy and demand a better monitor, i'd take what i could get and run with it, something is better than nothing
 
CodeWaste said:
Thats interesting, i haven't played doom 3 im my rig (my roomate just didnt have use of his computer for about a day) so im mostly unfamiliar with all the tricks that ATi has been employing.

Curious though about the banding on the shotty, so when the AI is on IQ goes down? Didn't people shit a brick when nVidia was doing that a couple of years ago? Why is it suddenly ok for ATi to do it for Doom 3?


I wouldn't put too much salt into 1c3d0g bashing AI. If any of it were true I'm sure there would be a flood of threads started by Trancendenz and Empty "proving" just that.
 
Well I've noticed the shimmering with an X600 too although it doesn't show up on my 5700U.. I'm poor ATM.

if it were me i wouldn't be greedy and demand a better monitor, i'd take what i could get and run with it, something is better than nothing
My problem is that they're CPU limited for most of the benchmarks, all it shows is that both SLI and CrossFire are about equal at those low resolutions (although SLI looks like it's beginning to breaking away at 1280x1024 with AA and AF).
 
CrimandEvil said:
all it shows is that both SLI and CrossFire are about equal at those low resolutions (although SLI looks like it's beginning to breaking away at 1280x1024 with AA and AF).

yep, and you wouldn't have known that if they hadn't released those results, so at least you got something

i know, you want more, so do I :)
 
Brent_Justice said:
yep, and you wouldn't have known that if they hadn't released those results, so at least you got something

i know, you want more, so do I :)
Well there is something of an educated guess to that but yeah I want more. :)
 
Brent_Justice said:
yep, and you wouldn't have known that if they hadn't released those results, so at least you got something

i know, you want more, so do I :)

Yeah, you're right...it just seems strange is all I can say...really really strange...
 
^eMpTy^ said:
Yeah, you're right...it just seems strange is all I can say...really really strange...

How?

Th sight had an invitation to test out the system, they didnt have time to get a better monitor, and had to do with what they had
 
^eMpTy^ said:
Yeah, you're right...it just seems strange is all I can say...really really strange...


I really dont see anything strange at all. Have you ever been to one of those shows? Its crazy hetic from what I have heard. The had limited time and grabbed what the could. There is no hiding going on as they clearly stated why they had no benchies at that res... Now had the not stated that or hid the fact that the used and LCD then you maybe on to something...

Thanks Terra when I saw that LCD in the background I figured thats the case. My Sony HS94P only supports 1280x1024 so I am used to living in sub 1600 res..thats why it was easy to spot :)
 
Brent_Justice said:
to me it sounds like they were lucky enough just to get to run that one game, they probably didn't have much time to do it in either

if it were me i wouldn't be greedy and demand a better monitor, i'd take what i could get and run with it, something is better than nothing

They actualle say something about that in the last paragraph:

"Vi må desværre vente med at lancere flere benchmark resultater indtil NDA'en udløber senere på måneden, men vi har allerede en aftale i hus om test af et retail Crossfire system."

Wich translates to this:

"Sadly we have to wait until the NDA is lifted at the end of the month before we can post anymore benchmarks, but we have already gotten a deal about a test of a retail Crossfire system"

I just wonder how they can get such a deal..when they seem to already have broken the NDA? :S *L*

Terra - But then again..they are danes :D

EDIT:
They didn't seem to breake any NDA afterall:
"Vi vil også gerne rette en tak til DFI for at give os eksklusiv ret til at benchmarke platformen under Computex Taipei 2005."

Translated:

"we would also like to thank DFI for giving us the exclusive right to benchmark the platform during Computex Taipei 2005"

My bad...
 
Brent_Justice said:
yep, and you wouldn't have known that if they hadn't released those results, so at least you got something

i know, you want more, so do I :)

I imagine you probably already know :p . Even if they haven't shipped the test boards to HardOCP, I'd guess you probably have seen some benches at Computex but are under NDA.


Regardless, I can pretty much guess the results at high resolution. SLi has about 70-90% performance jump when you SLi a card depending on the game. Most likely ATI is gonna have a 65%-85% jump (I'm saying 5% less because of the fact that their isn't a high speed bridge like on SLi) which will make it about even with SLi becasue the x850 is slightly faster than the ultra.
 
Terra said:
"Sadly we have to wait until the NDA is lifted at the end of the month before we can post anymore benchmarks, but we have already gotten a deal about a test of a retail Crossfire system"

I just wonder how they can get such a deal..when they seem to already have broken the NDA? :S *L*

Terra - But then again..they are danes :D

hmm maybe ATI dosen't want the 1600x1200 results out just yet?
 
jon_k said:
Regardless, I can pretty much guess the results at high resolution. SLi has about 70-90% performance jump when you SLi a card depending on the game. Most likely ATI is gonna have a 65%-85% jump (I'm saying 5% less because of the fact that their isn't a high speed bridge like on SLi) which will make it about even with SLi becasue the x850 is slightly faster than the ultra.

Do you know what the data rates on their DVI connect is ver the SLI brigde? Just wondering...
 
jon_k said:
hmm maybe ATI dosen't want the 1600x1200 results out just yet?

Dunno...could be..
But I am glad benches are begining to show up...

Terra - I want more benches! :D
 
Big deal! 2-5 FPS! You can't even tell after you hit over 50 FPS anyway! :eek:

Its all a matter of what company you like better ATI or Nvidia.
 
Jbirney said:
Do you know what the data rates on their DVI connect is ver the SLI brigde? Just wondering...

actually, i specifically asked ATI this question, i'm still waiting on an answer for that

i too was curious what the bandwidth was with the DMS cable
 
What annoys me is why didn't ATi sell their motherboards with the "CrossFire hardware compositing engine" They should figure out a way in which the motherboard handles which card is the master card and go from there. They would then need not sell 2 different cards at different price points.

Both cards are connected in the back which will transfer large amounts of data and the chip on the motherbard will handle small packets of information telling the master card that frame "even" is ready.

Stop using those damm connectors and start using the motherboard itself. I don't wanna buy 2 different cards with 1 costing more than the other. Obviously the Crossfire motherboard is gonna cost more but we can deal with it .. the price has to be transferred from 1 to the other, but atleast the cards won't be out of stock because everyone has to buy the first one before the second one.
 
Jbirney said:
Maybe but thats usally is where you get the most bang from the 2nd video card... Plus we have seen other high res doom scores...see anandtech: http://www.anandtech.com/video/showdoc.aspx?i=2432&p=9

Strange..I noticed something:

HardInfo 1280x1024 4xAA/8xAF ATI Radeon x850XT PE Crossfire 52.2 FPS
Anandtech:1280x1024 4xAA/8xAF ATI Radeon X850 XT Crossfire 87.4 FPS

They both use a AMD Athlon64 FX-53, but HardInfo use 1024 MB RAM vs. Anandtechss 512..And still the result are not within margins of error?

Terra - Now I am really confused?
 
mohammedtaha said:
What annoys me is why didn't ATi sell their motherboards with the "CrossFire hardware compositing engine" They should figure out a way in which the motherboard handles which card is the master card and go from there. They would then need not sell 2 different cards at different price points.

Both cards are connected in the back which will transfer large amounts of data and the chip on the motherbard will handle small packets of information telling the master card that frame "even" is ready.

Stop using those damm connectors and start using the motherboard itself. I don't wanna buy 2 different cards with 1 costing more than the other. Obviously the Crossfire motherboard is gonna cost more but we can deal with it .. the price has to be transferred from 1 to the other, but atleast the cards won't be out of stock because everyone has to buy the first one before the second one.

the thing is, you need the chip on the video card otherwise the crossfire video card dosen't know to use that dongle connecter to communicate with the other card.
 
jon_k said:
the thing is, you need the chip on the video card otherwise the crossfire video card dosen't know to use that dongle connecter to communicate with the other card.

Yes it does, you tell it to.

How does it know that output goes out of the dvi connector at the end of the card ? Because that's the flow of the data.

When it's in Crossfire mode, it awaits information through the motherboard to release 2 which is connected too .. so ..card 1 release .. 2 says ready .. release .. 1 release ..2 ready. .. release... and so on .. or if both release together ( checker board) then both release together.

Not that hard. Maybe it's not that simple but I'm sure their team of crossfire experts can figure that one out.
 
Just a thought, but is there any reason a Crossfire setup not walk all over a SLI setup? SLI is what 6-7 mounths old? Since its been stocked on shelves? And crossfire is not even out yet? I would think newer is better....

I think ATI should be working on a new card and not just dubbling there current line.. I hope they are anyway...
 
mohammedtaha said:
Yes it does, you tell it to.

How does it know that output goes out of the dvi connector at the end of the card ? Because that's the flow of the data.

When it's in Crossfire mode, it awaits information through the motherboard to release 2 which is connected too .. so ..card 1 release .. 2 says ready .. release .. 1 release ..2 ready. .. release... and so on .. or if both release together ( checker board) then both release together.

Not that hard. Maybe it's not that simple but I'm sure their team of crossfire experts can figure that one out.

yes that idea is possible, but would be impracticle. Using the PCIe bus to allow the crossfire chip to communicate with the video card would take away a lot from the bandwidth. Putting it on the video card makes it much faster.
 
Not necessarily because both cards still have to communicate whether it's on the motherboard or an add on connector like SLI. They can hardwire connectors or do the same thing nvidia is doing. I just don't like the idea that you have to pay more for 1 card over the other. Especially if the master card is out of stock everywhere just like the X800 last year.
 
Terra said:
Strange..I noticed something:

HardInfo 1280x1024 4xAA/8xAF ATI Radeon x850XT PE Crossfire 52.2 FPS
Anandtech:1280x1024 4xAA/8xAF ATI Radeon X850 XT Crossfire 87.4 FPS

They both use a AMD Athlon64 FX-53, but HardInfo use 1024 MB RAM vs. Anandtechss 512..And still the result are not within margins of error?

Terra - Now I am really confused?

First of all, thanks for the clear translations Terra.

Maybe those were done using different time demo's or is there only the prebuilt in method into Doom 3? I dont know, hated the game, so I returned it about two days after buying it.
 
Skrying said:
First of all, thanks for the clear translations Terra.

Your welcome, we are only 5.2 million danes, and I don't expect other than danes to understand danish and since there is much speculation, it helps to understand what they are saying..and what they are not saying.
Just giving clean info to the [H]ard ;)

Maybe those were done using different time demo's or is there only the prebuilt in method into Doom 3? I dont know, hated the game, so I returned it about two days after buying it.

I know there is a default benchmark build into Doom3, but you can also make your own.
It just seems like bad benchmarking to me if they use individual benchmarks, leaving us users to second guees how the new rigs will preform compared to what we got, thats all...

On a side note I still have Doom3 and the expansion R.O.E installed, as I like the game..now I just wait for Creative to release drivers(EAX4.0) for my Audigy 2 Value so I can play it with EAX :(

Terra - I liked going oldschool..with state-of-the-art GFX :)
 
jon_k said:
yes that idea is possible, but would be impracticle. Using the PCIe bus to allow the crossfire chip to communicate with the video card would take away a lot from the bandwidth. Putting it on the video card makes it much faster.

Would it be possible for normal x800's to go dual using AFR? There's no blending needed there. Only drawback i can think of if it works is that you get staggering motionblur effects.
 
mohammedtaha said:
Not necessarily because both cards still have to communicate whether it's on the motherboard or an add on connector like SLI. They can hardwire connectors or do the same thing nvidia is doing. I just don't like the idea that you have to pay more for 1 card over the other. Especially if the master card is out of stock everywhere just like the X800 last year.

The nVidia SLi bridge is in essence a dedicated line that the two graphics cards can exchange SLi data over. With this solution the two cards can exchange data with no performance hit on the PCIe bus. nVidia can get away with doing this because the SLi instructions are already on the NV40 chips.

ATI *can't* simply add a bridge connector becasue the multi-GPU instructions aren't nativly on the chip. The NV40 has SLi on the chip, so all card manufacturers have to do is add a bridge connecter and that is it. Therefore, ATI had to create the crossfire chip which basically controls the communication between the two video cards. They could either put the chip on the video card or the motherboard. However, putting it on the motherboard would waste valueable PCIe bandwidth and slow the whole system down.
 
HRslammR said:
1024x768? spppsh. show me some benches of cards actually doing some work, let's say 1600x1200 at minimum, then i'll be impressed.

QFT

Those benchmarks are meaningless. LOL

Wow. 12X10 4X8X Doom 3 on >$1000 SLI setups?!?! Good work hans and Franz! :rolleyes:
 
I(illa Bee said:
Just a thought, but is there any reason a Crossfire setup not walk all over a SLI setup? SLI is what 6-7 mounths old? Since its been stocked on shelves? And crossfire is not even out yet? I would think newer is better....

I think ATI should be working on a new card and not just dubbling there current line.. I hope they are anyway...


Newer is certainly not always better.

Anyway, the tech is limited by the card. You cant get more gains than is limited by the card itself. ATi cards are known to be slower in Doom 3 than Nvidia's cards. Not as much now thanks to ATi's drivers, but they still are slightly behind in overall performance when it comes to Doom 3. So, Crossfire itself can not add any extra performance that is not there.

So, you'll most likely get the same pattern with all games. Ones that are faster on a single X850 will be faster in Crossfire, and those faster on a single 6800U will be faster in SLI.
 
Sly said:
Would it be possible for normal x800's to go dual using AFR? There's no blending needed there. Only drawback i can think of if it works is that you get staggering motionblur effects.

The problem is, you are going to want a dedicated piece of hardware to control the frame sycnhronization comming from the two cards. If you did it on the driver level it is going to take a big perfromance hit. Alienware attempted to do something like this and it resulted in a 30% performance increase I think. Needless to say, they discontinued it pretty quick.
 
Terra said:
Strange..I noticed something:

HardInfo 1280x1024 4xAA/8xAF ATI Radeon x850XT PE Crossfire 52.2 FPS
Anandtech:1280x1024 4xAA/8xAF ATI Radeon X850 XT Crossfire 87.4 FPS

They both use a AMD Athlon64 FX-53, but HardInfo use 1024 MB RAM vs. Anandtechss 512..And still the result are not within margins of error?

Terra - Now I am really confused?

Notice this as well:
http://www.hardforum.com/showthread.php?t=881152

I got 83fps at 12X10 4X8X and 67fps at 16X12 4X8X with 3800+/6800GTs. I think Hardinfo should change their name to Tardinfo if they're using the standard Doom3 benchmark.

Achtung!
 
jon_k said:
The nVidia SLi bridge is in essence a dedicated line that the two graphics cards can exchange SLi data over. With this solution the two cards can exchange data with no performance hit on the PCIe bus. nVidia can get away with doing this because the SLi instructions are already on the NV40 chips.

ATI *can't* simply add a bridge connector becasue the multi-GPU instructions aren't nativly on the chip. The NV40 has SLi on the chip, so all card manufacturers have to do is add a bridge connecter and that is it. Therefore, ATI had to create the crossfire chip which basically controls the communication between the two video cards. They could either put the chip on the video card or the motherboard. However, putting it on the motherboard would waste valueable PCIe bandwidth and slow the whole system down.


But ATI said that both the X800 and X850 cards are AFR compatible. This means that thought has been put into it already. I guess the problem essentially was they might have lied. Their cards aren't AFR ready, the new cards are and this is the only way it can be done.

We, however, can not think this way because there's always a different way of doing things. Not necessarily better, or worse, but different.
 
Back
Top