• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

6800NU SLI benches

Rollo

Gawd
Joined
Jul 11, 2004
Messages
972
For those curious about the little seen 6800NU SLIs performance, I've done a little benchmarking. More to come.

System:
A64 3800+/Asus A8N SLI/ 1GB Geil PC3500/2 Leadtek 6800 256MB/SB Audigy 2
Drivers: 71.80, High Quality

Doom3

12X10 4X8X = 57fps

12X10 4X8X=77fps

16X12 2X8X=61fps

16X12 4X8X=42fps

Half Life 2

12X9 4X8X, all "High"

at_c17_12=58fps

at_canals_ 08=88fps

at_coast_05=91fps

at_prison_05=90fps

16X12 4X8X, all "High"

at_c17_12=53fps

at_canals_ 08=76fps

at_coast_05=85fps

at_prison_05=70fps

3Mark2003

14275

Far Cry Maximum Details

10X7 4X8X
Regulator=57fps
Research=91fps
Training=63fps
Volcano=91fps

12X10 4X8X
Regulator=44fps
Research=66fps
Training=50fps
Volcano=65fps

16X12 4X8X
Regulator=34fps
Research=50fps
Training=39fps
Volcano=49fps



3DMark 2005

6761

Painkiller
16X12 4X8X =67fps
19X14 0X0X= 66fps

Half Life2/High Res
19X14 0X0X
at_c17_12=56fps

at_canals_ 08=77fps

at_coast_12=76fps
 
Not bad.... I would have thought that the 03 scores would have been a bit higher, but nice score none the less!
 
Nice, thanks!

(What, no single-card benchies for comparisons? :p J/K, good work!)
 
mavalpha said:
Nice, thanks!

(What, no single-card benchies for comparisons? :p J/K, good work!)

You're welcome, and maybe on the single cards. :)

Far Cry benches added.
 
Chaballaman said:
can u unlock extra pipes with rivatuner on pci express ones?

I haven't tried, but according to the Rivatuner forums the "extra pipes" don't exist on PCIE 6800s.

PCIE 6800s are second generation, PCIE native nV41 chips, not bios/laser crippled nV40 chips.
 
You have a lot of room left for tweaking, but those aren't bad scores. Not quite a match for my dual 6600GT's but close ;)
 
ThomasE66 said:
You have a lot of room left for tweaking, but those aren't bad scores. Not quite a match for my dual 6600GT's but close ;)

Are you sure you don't mean your dual 6800 GTs?

Reason I ask is I also have a dual 6600GT set, and my benches and Firing Squad show the 6800NUs faster?
 
I've gotten 14,900 in 3dmark03, and about 6700 in 3dmark05 with 6600GT's running 550/1.1. I would have thought the 6800nu's would do better. 71.81 WHQL drivers.
 
kirbyrj said:
I've gotten 14,900 in 3dmark03, and about 6700 in 3dmark05 with 6600GT's running 550/1.1. I would have thought the 6800nu's would do better. 71.81 WHQL drivers.

It's all good, running synthetic benchmarks at 10X7 0X0X is not "gaming".

Like I said, I have a 6600GT SLI set as well and like it a lot. The one thing you can't get around is the 128MB, 16X12 4X8X is not an option with that rig, and sometimes 12X10 4X8X pushes it.

6800NUs have their place. ;)
 
Rollo said:
It's all good, running synthetic benchmarks at 10X7 0X0X is not "gaming".

Like I said, I have a 6600GT SLI set as well and like it a lot. The one thing you can't get around is the 128MB, 16X12 4X8X is not an option with that rig, and sometimes 12X10 4X8X pushes it.

6800NUs have their place. ;)

Good point. Can you overclock those 6800nu's very far?
 
Hate_Bot said:
You should unlock the pipes. Then, youll have 6800GT SLI

Can't unlock PCI-E 6800nu pipes. NV41 is 12 pipe native.
 
kirbyrj said:
Good point. Can you overclock those 6800nu's very far?

Don't know, and am afraid to try.

I am of the old school "you break it, you bought it" mentality, and can't afford to lose what I just spent on them.

I did OC the RAM to 700MHz and bench Doom 3 around 10% higher. :)
 
I would think the nv41 core would overclock nicely...stock on those is still 300Mhz right? or 325? and how high have you gotten it?
 
SlasherDNA said:
6800NU has RC 700MHz default.. of what OC you talking about?

Stock memory clock on PCIE 6800NUs is 600MHz. The only place it is listed higher is on newegg for the MSI model, and that is in error.

PCIE 6800NU 325MHz core clock, 12 pipelines, 5 vertex processors. 256MB of 600MHz DDR1.

AGP 6800NU 325 MHz core clock, 12 pipelines, 5 vertex processors. 128MB of 700MHz DDR1.

The RAM that comes on the PCIE 6800NUs is rated for 700MHz, but comes clocked at 600MHz.
 
^eMpTy^ said:
I would think the nv41 core would overclock nicely...stock on those is still 300Mhz right? or 325? and how high have you gotten it?

Like I said, I am an honest person. I can't afford to replace the 6800NUs if I damage them OCing them, so I'm not going to OC them. The core is at 325MHz though, and the one review of them out there on Firing Squad shows they OC well:

http://www.firingsquad.com/hardware/msi_nx6800/page12.asp

They perform well enough at stock I'm not tempted to risk losing them by OCing.
 
Rollo said:
Stock memory clock on PCIE 6800NUs is 600MHz. The only place it is listed higher is on newegg for the MSI model, and that is in error.
Ooh, I see. Well that I didn't know (=
 
SlasherDNA said:
Ooh, I see. Well that I didn't know (=

Don't feel bad, most people think the PCIE 6800s are the same as the AGP ones, and they are not.
 
Yea I see. Though it seems funny, why is the PCI-E version downclocked.. who knows.
 
SlasherDNA said:
Yea I see. Though it seems funny, why is the PCI-E version downclocked.. who knows.

If I were to guess, I would say they do that to make the performance "equivalent" with the AGP card since the AGP only has 128mb. Kind of like how AMD plays around with the PR ratings by changing speeds/cache sizes.
 
kirbyrj said:
If I were to guess, I would say they do that to make the performance "equivalent" with the AGP card since the AGP only has 128mb. Kind of like how AMD plays around with the PR ratings by changing speeds/cache sizes.

They have done the same type of thing before, more memory, but slower. I believe it's the same on 6600s?
 
The 64MB Ti4200 actually had faster RAM than the 128MB version, so it was faster in a few benchmarks.

Even more interesting, the GDDR3-based FX5700Ultra had its RAM clocked 25MHz (50MHz DDR) higher than the GDDR2-based cards. This was done to offset latencies, though.
 
Rollo said:
Like I said, I am an honest person. I can't afford to replace the 6800NUs if I damage them OCing them, so I'm not going to OC them. The core is at 325MHz though, and the one review of them out there on Firing Squad shows they OC well:

http://www.firingsquad.com/hardware/msi_nx6800/page12.asp

They perform well enough at stock I'm not tempted to risk losing them by OCing.

There's no real risk of damaging the card unless you get all crazy changing the voltage around...just modifying the clock via coolbits or something you can't really hurt anything...

Somebody back me up on this...
 
http://www20.graphics.tomshardware.com/graphic/20040405/index.html
the third linked page said:
To counterbalance the negative impact of the new memory's higher latencies (and probably also to differentiate the new card from the previous model), NVIDIA decided to raise the memory frequency from 450MHz (900MHz DDR) to 475MHz (950MHz DDR). The core frequency of the NV36 GPU remains unchanged at 475MHz, however.
 
No I didn't. So when did they make a ggdr3 5700Ultra, not 'long' ago eh?

I only knew that exister DDRII, and the thing was hitting damn high temperatures..
 
SlasherDNA said:
No I didn't. So when did they make a ggdr3 5700Ultra, not 'long' ago eh?

I only knew that exister DDRII, and the thing was hitting damn high temperatures..

It was about the same time they introduced the 5700 PCX (the pci-e version)
 
Well, that article that I linked was written in April of '04. I don't remember if this was a preview or a review, though, but still...

And from the same article:
The only graphics cards known to use GDDR-2 memory were/ are the 256MB version of ATi's Radeon 9800 Pro, NVIDIA's GeForce FX5800 Ultra and the aforementioned GeForce FX5700 Ultra. In practice, the greatest problem with this memory type was the high heat dissipation associated with it. Thanks to the lower VDDQ voltage of only 1.8V, GDDR-3 promises to solve these temperature-related problems - at least for the slower chips.
 
BTW/OT: I find it interesting that OCZ & co are proud of their DDR600 modules, while cards like the GeForce6800 run with 700MHz speeds by design. This card uses DDRI, not a GDDR counterpart- the chips should be the same or similar, so the only reason I can see them having trouble reaching similar speeds in DIMMs is that the DIMMs have different/greater limitations, like greater distances between the memory controller and the chips- therefore, higher latencies and more noise.

Edit: I knew about the G.Skill, A-Data, and OCZ, I just didn't want to list all of them. OCZ's was the most recent announcement, so it came to mind.
 
mavalpha said:
BTW/OT: I find it interesting that OCZ & co are proud of their DDR600 modules, while cards like the GeForce6800 run with 700MHz speeds by design. This card uses DDRI, not a GDDR counterpart- the chips should be the same or similar, so the only reason I can see them having trouble reaching similar speeds in DIMMs is that the DIMMs have different/greater limitations, like greater distances between the memory controller and the chips- therefore, higher latencies and more noise.

G.Skill has DDR600 and has had it longer than OCZ.

I have no idea why they cant put high speed graphics quality chips on a DIMM unless it really is just the delay between the DIMM and the memory controller.

Some GeForce 6800's can hit 900MHz memory. DDR900. God thats what, 7.2 GB/S if used on a motherboard? On a graphics card, thats about 30GB/S bandwidth.
 
Maybe it's something we can count on from Asus or Fatal1ty's next motherboard...? Maybe they'll redesign it to get the DIMM slots unbelievably close to the northbridge(P4)/CPU(K8)? Might get some insane OCs on the RAM!

Not likely, but it'd be nice.
 
mavalpha said:
Maybe it's something we can count on from Asus or Fatal1ty's next motherboard...? Maybe they'll redesign it to get the DIMM slots unbelievably close to the northbridge(P4)/CPU(K8)? Might get some insane OCs on the RAM!

Not likely, but it'd be nice.

It might just be because the video card's PCB and components are engineered to be dedicated for pure bandwidth between the memory and the GPU, whereas the motherboard has to do many other things besides memory <-> cpu bandwidth.
 
^eMpTy^ said:
There's no real risk of damaging the card unless you get all crazy changing the voltage around...just modifying the clock via coolbits or something you can't really hurt anything...

Somebody back me up on this...

If they did, they'd be wrong. It's called electro-migration and it's the tendency of current to skip across traces on the cpu/gpu as heat rises. After it has skipped across enough, it forms a permanent path, damaging the cpu/gpu.

Heat rises with frequency (MHz) lifespan of the chip falls. So even if it runs seemingly "fine" at the OCd frequency for six months, the damage you're doing may prevent the chip from lasting a year.

This leaves the OCer in an unfortunate position:
Steal another card from the company with a fraudulent RMA, or buy a new card/cpu.

In the old days, when OCing first started, it was primarily a method to get passable "enthusiast" performance out of cheap parts for people who couldn't afford good parts.

These days people use it to get meaningless performance increases out of good parts. If you don't believe me about the "meaningless", check out the benchmark differences between a 6800U and a Gainward CoolFX 6800U at 450/1200 stock:
http://graphics.tomshardware.com/graphic/20040721/index.html

There is no performance gain there that could be noticed actually playing the game, but around the world people void their warranties and risk their expensive parts largely to say "My 3dmark went up a few 100 points". Meh.

BTW- those who are about to say "But it will throttle down at 127 degrees"- yep, it will but how do we know for sure damage isn't being done at 126?
 
Electromigration is not caused by heat, it is caused by over voltage. It's a side effect of quantum electron tunnelling. High heat does contribute to the process, but it is mainly a function of high current. Eventually it causes inhibition of the conductor paths.

Early Northwood core users who overvolted found this out the hard way.

As for overclocking being meaningless for performance I have to disagree. While it *can* be a meaningless endeavor, a good overclock on a mid range videocard can make the difference between 10x7 and 12x10, all other eye candy settings being equal.
 
Back
Top