Crossfire beats SLI in Doom 3?!

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

nRollo

Banned
Jan 11, 2002
10,460
0
0
Originally posted by: Dkcode
rollo where did u buy those reference gts from? did they come with a box?

I got the cards from nVidia, the only packaging was the shipping box. I have a friend who works there and can get me good deals every now and again, occasional tidbits of info.
 

nRollo

Banned
Jan 11, 2002
10,460
0
0
Originally posted by: nitromullet
Those framerates make no sense at all if they were using the standard Doom3 demo:
http://www.hardforum.com/showthread.php?t=881152
I'm getting 67fps at 16X12 4X8X on 6800GT SLI, they're getting 50fps at 12X10 4X8X? Has to be they had the vsynch on, used a different demo, or have downclocked their FX CPU. It makes no sense.

http://www.tweaktown.com/document.php?dType=article&dId=773

Here is the same test in an English article. Yes, they used a different test then the standard D3 benchmark.

Ah. I imagine their benches are correct. Too bad they couldn't put up the only one that mattered, 16X12 4X8X. :(

 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
They didnt use timedemos in doom3, which means it was running SMP aware.

The dual core CPU has a strong advantage over single core in non-timedemo benchmarks.

They also only ran retarded low settings.
 
Mar 11, 2004
23,444
5,852
146
Originally posted by: Rollo
Originally posted by: nitromullet
Those framerates make no sense at all if they were using the standard Doom3 demo:
http://www.hardforum.com/showthread.php?t=881152
I'm getting 67fps at 16X12 4X8X on 6800GT SLI, they're getting 50fps at 12X10 4X8X? Has to be they had the vsynch on, used a different demo, or have downclocked their FX CPU. It makes no sense.

http://www.tweaktown.com/document.php?dType=article&dId=773

Here is the same test in an English article. Yes, they used a different test then the standard D3 benchmark.

Ah. I imagine their benches are correct. Too bad they couldn't put up the only one that mattered, 16X12 4X8X. :(

Yeah, thats the one that I look at in benchmarks anymore. I hope they add 1920x1080 or 19x12 in the future.

Their framerates still seem awfully low for being lower resolution. I remember playing Doom 3 at 16x12 with 2xAA 8xAF and it being totally playable in almost all cases on my 6800 @ 400/1100.
 

PrayForDeath

Diamond Member
Apr 12, 2004
3,478
1
76
Originally posted by: Yossairian
Originally posted by: Lonyo

ATi aren't based in the USA.
Unless Canada just because part of the USA?

*nod*

Northern Montana, the 51st state.

Originally posted by: Ronin
Originally posted by: PrayForDeath
Originally posted by: Lonyo
Originally posted by: PrayForDeath
Link

I find this hard to believe. And I was wondering, why do the europeans always get to test the hardware first? Especially that the companies are based in USA.

ATi aren't based in the USA.
Unless Canada just because part of the USA?

Okey, North America.

ATi has offices in Redmond.

For god's sake what's the big deal? Anyway, I was talking about where the company is based, not where it has offices.
 

Pr0d1gy

Diamond Member
Jan 30, 2005
7,774
0
76
Originally posted by: Acanthus
They didnt use timedemos in doom3, which means it was running SMP aware.

The dual core CPU has a strong advantage over single core in non-timedemo benchmarks.

They also only ran retarded low settings.

I thought software had to be coded for dual core to take advantage of the hardware?
 

trinibwoy

Senior member
Apr 29, 2005
317
3
81
Originally posted by: Pr0d1gy
Originally posted by: Acanthus
They didnt use timedemos in doom3, which means it was running SMP aware.

The dual core CPU has a strong advantage over single core in non-timedemo benchmarks.

They also only ran retarded low settings.

I thought software had to be coded for dual core to take advantage of the hardware?


Correct. Doom3 is not multi-threaded and shows no gains from dual-core CPU's.
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: trinibwoy
Originally posted by: Pr0d1gy
Originally posted by: Acanthus
They didnt use timedemos in doom3, which means it was running SMP aware.

The dual core CPU has a strong advantage over single core in non-timedemo benchmarks.

They also only ran retarded low settings.

I thought software had to be coded for dual core to take advantage of the hardware?


Correct. Doom3 is not multi-threaded and shows no gains from dual-core CPU's.

Check again, Doom 3 IS SMP AWARE.
 

Drayvn

Golden Member
Jun 23, 2004
1,008
0
0
I suppose the east gets it first because of the illegal selling of hardware is more rampant than in the west?
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: trinibwoy
Originally posted by: Acanthus
Check again, Doom 3 IS SMP AWARE.

That is a pretty empty post. Check what again? Where are your links with hard info? The Doom3 engine may be SMP aware but Doom3 certainly isn't multi-threaded - http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=7

SMP aware means that the software is written for SMP, also known as symmetrical multi processing. SMP is mutli-threading...

When you are running timedemos (like anand always does) the SMP threads are not executed beacuse the AI and physics are not used, this skews scores in the favor of single core CPUs.
 

imported_Noob

Senior member
Dec 4, 2004
812
0
0
Originally posted by: PrayForDeath
And I was wondering, why do the europeans always get to test the hardware first? Especially that the companies are based in USA.

Probably the same reasons why North America's cell phones are 2 years behind Europe's. The delivery time is shorter for the products to get from Aisa to Europe then from Aisa overseas to North America. Don't forget the manufacturing labour is outsourced, along with tech support lines.
 

trinibwoy

Senior member
Apr 29, 2005
317
3
81
Originally posted by: Acanthus
Originally posted by: trinibwoy
Originally posted by: Acanthus
Check again, Doom 3 IS SMP AWARE.

That is a pretty empty post. Check what again? Where are your links with hard info? The Doom3 engine may be SMP aware but Doom3 certainly isn't multi-threaded - http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=7

SMP aware means that the software is written for SMP, also known as symmetrical multi processing. SMP is mutli-threading...

When you are running timedemos (like anand always does) the SMP threads are not executed beacuse the AI and physics are not used, this skews scores in the favor of single core CPUs.

I'm well aware of the definition of SMP. What I'm waiting on is some kind of proof from you that Doom3 is multi-threaded and sees gains on dual-core CPU's. All you've given me so far is your 'word' and I can find no evidence to back it up so I'm hoping you can find some for me....
 

xxxAgentxxx

Member
May 28, 2005
32
0
0
From what I've read, the doom 3 engine isn't multi threaded, nor is any other engine out. First engine that is supposedly supposed to support the multiple threading is the unreal 3 engine.
 

Regs

Lifer
Aug 9, 2002
16,666
21
81
I wouldn't say 3-5 fps beats it. Especially not 1024 x 768! We need more powerful CPU's for dual GPU tests.
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: trinibwoy
Originally posted by: Acanthus
Originally posted by: trinibwoy
Originally posted by: Acanthus
Check again, Doom 3 IS SMP AWARE.

That is a pretty empty post. Check what again? Where are your links with hard info? The Doom3 engine may be SMP aware but Doom3 certainly isn't multi-threaded - http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2410&p=7

SMP aware means that the software is written for SMP, also known as symmetrical multi processing. SMP is mutli-threading...

When you are running timedemos (like anand always does) the SMP threads are not executed beacuse the AI and physics are not used, this skews scores in the favor of single core CPUs.

I'm well aware of the definition of SMP. What I'm waiting on is some kind of proof from you that Doom3 is multi-threaded and sees gains on dual-core CPU's. All you've given me so far is your 'word' and I can find no evidence to back it up so I'm hoping you can find some for me....

Is this enough? Screenshot after 15 seconds of playing doom 3 - CPU usage.
 

trinibwoy

Senior member
Apr 29, 2005
317
3
81

No you're still wrong. All that says is that Doom3's instructions are alternating between the different cores - not that they are being processed in parallel. If you were to sum those graphs you would get a relatively flat line around 100%. You obviously don't understand how this thing works.

http://www.techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=5

Techreport says about Doom3:
This should give you some indication of the amount of threading in the application. In some cases with single-threaded apps like the games below, the task will oscillate back and forth between one CPU and the next, but total utilization generally won't go above 50% for a dual-core or 25% for a quad-core (or quad-front-end, in the case of the XE 840 with Hyper-Threading) system.



 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: trinibwoy

No you're still wrong. All that says is that Doom3's instructions are alternating between the different cores - not that they are being processed in parallel. If you were to sum those graphs you would get a relatively flat line around 100%. You obviously don't understand how this thing works.

http://www.techreport.com/reviews/2005q2/athlon64-x2/index.x?pg=5

Techreport says about Doom3:
This should give you some indication of the amount of threading in the application. In some cases with single-threaded apps like the games below, the task will oscillate back and forth between one CPU and the next, but total utilization generally won't go above 50% for a dual-core or 25% for a quad-core (or quad-front-end, in the case of the XE 840 with Hyper-Threading) system.

Thats windows 2000 HT support. There is a bug in the sceduler. There was also very little going on in the game as i just loaded it and did nothing.
 

canadageek

Senior member
Dec 28, 2004
619
0
0
Originally posted by: Ronin
Originally posted by: PrayForDeath
Originally posted by: Lonyo
Originally posted by: PrayForDeath
Link

I find this hard to believe. And I was wondering, why do the europeans always get to test the hardware first? Especially that the companies are based in USA.

ATi aren't based in the USA.
Unless Canada just because part of the USA?

Okey, North America.

ATi has offices in Redmond.

yeah, and ford has factories in canada. is ford canadian? no.
 

trinibwoy

Senior member
Apr 29, 2005
317
3
81
Originally posted by: Acanthus
Thats windows 2000 HT support. There is a bug in the sceduler. There was also very little going on in the game as i just loaded it and did nothing.

Sorry to say this but, you my friend, are full of shit. Is it an ego thing? Just admit you're wrong. If Doom3 was multi-threaded we would have a billion threads with evidence to support your claims. Instead all we have are a bunch of reviews that totally deny what you are saying.
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
The scene test consisted of saving at a point of looking at a scene in a large room with a moving machine in the middle.

The Fight test consisted of a non-timedemo approximated gameplay and moving around in an large area trying to get the lowest fps possible.

Neither of these are very scientific, but approximate results are...

1600x1200 4xAA 8xAniso, Ultra Quality, Everything all the way up, Vsync disabled.

Scene:

HT on : 47fps
HT off : 44fps

Fight:
HT on: 29-60fps (doom 3 is capped at 60fps)
HT off: 27-60fps
 

trinibwoy

Senior member
Apr 29, 2005
317
3
81
Originally posted by: Acanthus
Neither of these are very scientific, but approximate results are...

If it is true, why can't you find a third party to support your claims? It should be very easy to do if it's true, don't you think?

 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Can we please get some more HT or dual CPU people to confirm this...

The doom 3 engine is multithreaded, Carmack says its multithreaded... The results show its multithreaded.

ENCODING VIDEO uses 50% 50% on windows 2000, that must be "switching threads" too right? Even though theres a 19% performance increase.
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: trinibwoy
Originally posted by: Acanthus
Neither of these are very scientific, but approximate results are...

If it is true, why can't you find a third party to support your claims? It should be very easy to do if it's true, don't you think?

I would like someone to independantly verify my results from this forum so captain 100 post can stop harassing me like he knows what hes talking about.

But this might do as well Thread