HardOCP 6990+6970 CF vs 580 Tri Sli

[Dark Shroud]

The nf200 is basically a bridge chip that duplexes pci-e lanes, it allows greater bandwith whether its crossfire or sli or a raid card .

Just to be clear the nf200 does not make/create/add more bandwidth. It manages the exsisting bandwidth & lanes in specific ways. The down sides are greatly increased latency and when something goes wrong this chip usually makes the problem worse.

 

[notty22]

Just to be clear the nf200 does not make/create/add more bandwidth. It manages the exsisting bandwidth & lanes in specific ways. The down sides are greatly increased latency and when something goes wrong this chip usually makes the problem worse.

You mean like the internal bridge chip in the 6990 ?

The end result is the gpu's have access to more bandwith traffic.
Sorry angry fans don't suddenly know more than the engineers that choose to use the nf2000 in high end boards designed for crossfire and SLI. Its a proven design and why its used also used on boards like this :
ASUS Maximus IV Extreme for LGA 1155 platform, Intel P67 + NVIDIA NF200 bridge.
Gigabyte P67A-UD7 for LGA1155, Intel P67 + NVIDIA NF200 bridge.
ASUS P6T6 WS Revolution for LGA 1366 platform, Intel X58 + NVIDIA NF200 bridge.

 

[AtenRa]

16-8-8

The MSI Eclipse SLI has 3x PCI-e x16 ports
Two of them have 16 lanes and the third only have 4 lanes
So we have 16+16 + 4

If they used the MSI Eclipse Plus then it would be 16+8+8
I bet they used the Eclipse SLI board.

 

[Silverforce11]

Its a conspiracy I tell you...

All joking aside, I doubt a simple bridge chip could impact crossfire scaling. And plus it was only for one game, while the others were naturally within the margin of error.

If it happens in one game you don't think it could impact in other games? (There was less or little scaling for other games moving up to the massive clock speed difference in CPU) Why should it even interfere in any games at all if what its doing isn't inherently bad for AMD cards? Evidence is there. It's a poor chipset to run CF.

 

[notty22]

At the thread for this article at Hocp, a poster linked to the new
3GB EVGA model

With all these features and more, it is easy to see how the EVGA GeForce GTX 580 obliterates the competition.

 

[Awkward]

more like obliterates your wallet

 

[3DVagabond]

It's going to take more studying, but I can't imagine that nVidia designed the nf200 to work with 2gpu in one slot and 1gpu in the second, since their cards can't do that. I also doubt that AMD spent much time considering the nf200 chip while writing their drivers.

GPU performance is so reliant these days on perfect optimization that I think it's likely where the problem lies.

 

[Dribble]

Not true? ok, then direct us to your ban-inducing thread on the hardforum

Kyles basic mode of operation is "I am right, and if you don't agree with me I'll ban you", this theme runs through the whole site. I no longer post and visit a lot less after I got a ban from the video forums for politely suggesting some gamers like to play games at > 30fps and this is more important then even higher settings. Turns out you are not allowed to say things like that because "Kyle" has decided 30fps is enough. You are also not allowed to like gpu accelerated physics, 3D, any game that doesn't actively support eye infinity, etc...

Fortunately it's a big internet and [H] is just one of many little fish.

 

[Madcatatlas]

Are you sure it isnt you thats the fish? ¨
The internett is a ..big place with many fishers, among them [H], we are the fish.. deep i know.

 

[ActionParsnip]

Kyles basic mode of operation is "I am right, and if you don't agree with me I'll ban you", this theme runs through the whole site. I no longer post and visit a lot less after I got a ban from the video forums for politely suggesting some gamers like to play games at > 30fps and this is more important then even higher settings. Turns out you are not allowed to say things like that because "Kyle" has decided 30fps is enough. You are also not allowed to like gpu accelerated physics, 3D, any game that doesn't actively support eye infinity, etc...

Fortunately it's a big internet and [H] is just one of many little fish.

They are very firm but they are also fair from my experience. If you got a ban it's likely due to going off topic in a thread or flogging a dead equine that in previous pages on a thread was forbidden to be done so by the mods. For instance this review so someone banned who had been a member for more than 7 years. They take zero cr@p. It's a tough existence on the [H] but they do such informative articles and the editors really do care how they're articles are percieved if not sooo much how they're behaivour is percieved.

 

[Zap]

16-8-8

No. MSI Eclipse does 16/16/4.

 

[Dribble]

They are very firm but they are also fair from my experience. If you got a ban it's likely due to going off topic in a thread or flogging a dead equine that in previous pages on a thread was forbidden to be done so by the mods. For instance this review so someone banned who had been a member for more than 7 years. They take zero cr@p. It's a tough existence on the [H] but they do such informative articles and the editors really do care how they're articles are percieved if not sooo much how they're behaivour is percieved.

Then they are bad editors. If you want to run a *fair* forum then you have to be impartial, cold and logical when patrolling it. In [H] the only thing bigger then the monitor resolutions is the ego's of the people running it. Criticism goes down like a lead balloon, and often just having a different (and perfectly valid) opinion is taken as criticism.

I agree [H] have some good articles, and I still occasionally take a look, but am no longer a regular reader and I no longer participate. Then everyone's happy - I have plenty of other forums where I can have a discussion and sites where I can read up on my tech, and [H] can exist safe in the knowledge one less dissenting opinion has gone.

 

[happy medium]

http://www.hardocp.com/article/2011/05/03/nvidia_3way_sli_amd_trifire_redux/

Looks like Hardcop screwed up, the tri-sli system was 21.7% faster in their redux.

 

[Lonyo]

They "screwed up" by using a different CPU.

 

[Zap]

http://www.hardocp.com/article/2011/05/03/nvidia_3way_sli_amd_trifire_redux/

Looks like Hardcop screwed up, the tri-sli system was 21.7% faster in their redux.

Yes, it was already discussed earlier in this thread, but I guess it isn't as big a deal as when the Tri-SLI was slower.

Was it because Tri-SLI was bandwidth starved (third card PCIe x4) on the original platform?

Was it because Tri-SLI was CPU limited (new platform with faster CPU)?

Maybe they'll revisit it a third time and lower the clocks of their new platform to roughly match the CPU performance of their old platform? This way they can find out if it was a bandwidth issue.

 

[RussianSensation]

http://www.hardocp.com/article/2011/05/03/nvidia_3way_sli_amd_trifire_redux/

Looks like Hardcop screwed up, the tri-sli system was 21.7% faster in their redux.

Good to see Hardocp listening to their readers.

"Our testing done here today were just apples-to-apples performance, where VRAM limits were not hit. In real-world scenarios, faster CPU speed or not, we simply aren't going to be able to play at the same settings between these configurations. Every highest playable setting we found playable in our original evaluation, still holds true for this one."

So we can conclude that if not faced with the VRAM limitation, the 3x GTX580 setup is faster overall. Whether or not it's worth another $400-500 is up to the buyer. Both setups are great assuming you have a fast enough CPU to push them, but NV setup is much more sensitive to CPU speed. Therefore, if you have a Core i7 @ 3.6ghz or lower, the AMD setup wins hands down. If you have a SB overclocked setup, the NV's setup is ultimately faster (unless you crank AA very high) but costs $400-500 more. Sounds like both options will find their buyers.

Is Kyle going to do 2x HD6990s vs. 3x GTX580s then? Both setups cost around $1400-1500. That seems like the next logical comparison to me.

 

[Skurge]

Good to see Hardocp listening to their readers.

"Our testing done here today were just apples-to-apples performance, where VRAM limits were not hit. In real-world scenarios, faster CPU speed or not, we simply aren't going to be able to play at the same settings between these configurations. Every highest playable setting we found playable in our original evaluation, still holds true for this one."

So we can conclude that if not faced with the VRAM limitation, the 3x GTX580 setup is faster overall. Whether or not it's worth another $400-500 is up to the buyer. Both setups are great assuming you have a fast enough CPU to push them, but NV setup is much more sensitive to CPU speed. Therefore, if you have a Core i7 @ 3.6ghz or lower, the AMD setup wins hands down. If you have a SB overclocked setup, the NV's setup is ultimately faster (unless you crank AA very high) but costs $400-500 more. Sounds like both options will find their buyers.

Is Kyle going to do 2x HD6990s vs. 3x GTX580s then? Both setups cost around $1400-1500. That seems like the next logical comparison to me.

Yep, Its surprising really how much CPU power those cards need even those settings.

 

[AdamK47]

No. MSI Eclipse does 16/16/4.

Why the hell would they use that board for a 3-way SLI / Crossfire review to begin with?

 

[notty22]

I think in some games, the performance hit would not be a review game changer, but it seems F1 , does not like x4 slots.
There must be alot of 2-way communication, requiring much bandwidth.
This article was to test pcie lane amount effect.
With a single graphic card.
F1 showed some poor fps for X4,
http://www.tomshardware.com/reviews/pci-express-scaling-p67-chipset-gaming-performance,2887-6.html

 

[Zap]

Why the hell would they use that board for a 3-way SLI / Crossfire review to begin with?

I don't know. Maybe they intentionally did it to cripple the third Nvidia card? Maybe they didn't realize it and thought it was 16/8/8? I do know that HardOCP seems to favor MSI, so they probably just pulled out their trusty ol' board.

 

[Dark Shroud]

I don't know. Maybe they intentionally did it to cripple the third Nvidia card?

No [H] wouldn't do that.

Maybe they didn't realize it and thought it was 16/8/8? I do know that HardOCP seems to favor MSI, so they probably just pulled out their trusty ol' board.

These seem to be more likly reasons.

 

[happy medium]

I don't know. Maybe they intentionally did it to cripple the third Nvidia card? Maybe they didn't realize it and thought it was 16/8/8? I do know that HardOCP seems to favor MSI, so they probably just pulled out their trusty ol' board.

After some thought I came up with ...

This review site seems to pit Amd vs Nvidia with controversial reviews every so often.

Remember the gtx460 sli vs 5970 spectical?
http://www.hardocp.com/article/2010/09/16/nv_gtx_460_1gb_sli_vs_ati_hd_5870_cfx_redux/

That review was also backed up with a redux.

This is another posssible senerio......

Do a review that will generate a massive amount of page hits through controversy. It also boast many page hits in there own forum.
Have AMD and Nvidia fans using it for a base of arguement and linking to there review.

When the controversy dies down some , do a redux of the review to generate even more page hits.

It also has people thinking/saying, "hey they care about there readers" or "at least they came out and corrected their error".

So even though they screwed up the review, they get double the page hits, and actually have people thinking, "they did it for them" and add to their review site credibility.

 

[cusideabelincoln]

After some thought I came up with ...

This review site seems to pit Amd vs Nvidia with controversial reviews every so often.

Remember the gtx460 sli vs 5970 spectical?
http://www.hardocp.com/article/2010/...870_cfx_redux/

That review was also backed up with a redux.

This is another posssible senerio......

Do a review that will generate a massive amount of page hits through controversy. It also boast many page hits in there own forum.
Have AMD and Nvidia fans using it for a base of arguement and linking to there review.

When the controversy dies down some , do a redux of the review to generate even more page hits.

It also has people thinking/saying, "hey they care about there readers" or "at least they came out and corrected their error".

So even though they screwed up the review, they get double the page hits, and actually have people thinking, "they did it for them" and add to their review site credibility.

No.

 

[Lonyo]

After some thought I came up with ...

This review site seems to pit Amd vs Nvidia with controversial reviews every so often.

Remember the gtx460 sli vs 5970 spectical?
http://www.hardocp.com/article/2010/09/16/nv_gtx_460_1gb_sli_vs_ati_hd_5870_cfx_redux/

That review was also backed up with a redux.

This is another posssible senerio......

Do a review that will generate a massive amount of page hits through controversy. It also boast many page hits in there own forum.
Have AMD and Nvidia fans using it for a base of arguement and linking to there review.

When the controversy dies down some , do a redux of the review to generate even more page hits.

It also has people thinking/saying, "hey they care about there readers" or "at least they came out and corrected their error".

So even though they screwed up the review, they get double the page hits, and actually have people thinking, "they did it for them" and add to their review site credibility.

If there was ever a post which should get edited, this might be it.

What "error" was there?
The controversy here wasn't even about the issue which the redux covered, it was generated by people who couldn't accept NV could lose and looked for every excuse to discredit the article EXCEPT what ended up making a difference.
"NV is behind because they only have 1.5GB RAM" (not true).
"NV is behind because of the PCIe layout of the motherboard" (excuse used even after the redux but the other way round).
"NV is behind because of the games tested"
"NV is behind because their triple monitor implementation sucks".

That's user generated controversy about issues which can be easily explained/are a non-issue/are just idiotic anyway.

Then when they test with what made a difference, a faster CPU, you get people saying "oh now this motherboard is NV favoured because it has an NF200 chip on it".

Then you get posts like yours "all this is deliberately done unfairly to generate controversy".

People will find fault in ANYTHING that they do, and try and make excuses about this or that, or say that something is wrong or unfair.
Hocp doesn't even need to try and generate controversy, all they need to do is make an article and people will find SOMETHING to make a conspiracy theory about.

 

[Sable]

after some thought i came up with ...

This review site seems to pit amd vs nvidia with controversial reviews every so often.

Remember the gtx460 sli vs 5970 spectical?
http://www.hardocp.com/article/2010/09/16/nv_gtx_460_1gb_sli_vs_ati_hd_5870_cfx_redux/

that review was also backed up with a redux.

This is another posssible senerio......

Do a review that will generate a massive amount of page hits through controversy. It also boast many page hits in there own forum.
Have amd and nvidia fans using it for a base of arguement and linking to there review.

When the controversy dies down some , do a redux of the review to generate even more page hits.

It also has people thinking/saying, "hey they care about there readers" or "at least they came out and corrected their error".

So even though they screwed up the review, they get double the page hits, and actually have people thinking, "they did it for them" and add to their review site credibility.

hahahahahahahhahaaha