• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

SLI 770 phantom 4gb

Soldato
Joined
11 May 2014
Posts
5,472
Location
Edinburgh
Hi Guys,

So, I currently have one gainward 7700 phantom 4gb....and I want another. However I have a few questions.

This is my mobo:
Asus M5A99X EVO R2.0

It has 1 x16 slot, and 1 x8 slot, and according to the description when you SLI it will go down to 2 x8 slots.

So if I SLI two 770's, on x8 will it give much gain over one 770 on x16?

Regards
Colin
 
Awesome :) Im still fairly new to this so just wanted to double check everything before I commit.

Thanks guys :)
 
its less than 0.1% performance difference from x8 and x16 nothing to worry about, Im running x8/x8 and both at pci-e 2.0 due to the mobo doesn't like running at pci-e 3.0 but x8/x8/2.0 is more than enough bandwidth so im not fussed.


Go SLI and enjoy it :D just watch out for the extra heat generated with summer coming up
 
Last edited:
That's insane, I'm stunned that the difference between x8 and x16 was so negligible.

A friend benchmarked his 6990 crossfire setup:
- old mobo only allowed one card at x16 and the other at x8
- new mobo allowed both on x16 slots

He literally saw a calculated 50% increase in benchmark performance! So why do you think these 770 and 6990 setups were so wildly different?
 
Last edited:
That's insane, I'm stunned that the difference between x8 and x16 was so negligible.

A friend benchmarked his 6990 crossfire setup:
- old mobo only allowed one card at x16 and the other at x8
- new mobo allowed both on x16 slots

He literally saw a calculated 50% increase in benchmark performance! So why do you think these 770 and 6990 setups were so wildly different?

Possibly it was a really old mobo and he was also going from PCIe 1 to PCIe 2? Meaning, equivalent to PCIe 3, he was going from 3.0 x2 to 3.0 x8.

Possibly because he was running 6990, which means each card has a PCIe bridge on it then two GPUs, so he's talking about 4 way crossfire.

Possibly he changed his CPU when he changed his mobo.

Possibly he updated his drivers, which he hadn't done in ages, reinstalled windows (strongly recommended with new mobo) getting rid of malware/bloat, or something like that.

Possibly he had it misconfigured previously.

Possibly he is lying :P

See below tests with a 7970:

http://www.anandtech.com/show/5458/the-radeon-hd-7970-reprise-pcie-bandwidth-overclocking-and-msaa
 
Possibly it was a really old mobo and he was also going from PCIe 1 to PCIe 2? Meaning, equivalent to PCIe 3, he was going from 3.0 x2 to 3.0 x8.

Both PCIe 2.1, he has the early adopter disease so his "old" mobo was only a few months old.


Possibly because he was running 6990, which means each card has a PCIe bridge on it then two GPUs, so he's talking about 4 way crossfire.

Wouldn't this also be the case with his old mobo too?

Possibly he changed his CPU when he changed his mobo.

Same CPU.

Possibly he updated his drivers, which he hadn't done in ages, reinstalled windows (strongly recommended with new mobo) getting rid of malware/bloat, or something like that.

Very true actually. He would tend to play around with driver versions since some played havoc with his crossfire setup.

Possibly he is lying :P

Hehe! I know what you mean, seen that so many times, but I was there the day he was installing the new mobo so we ran a before and after bench to see if he would get any performance boost.

Possibly he had it misconfigured previously.

I'd be surprised, he's an early adopter and runs a tight ship but that doesn't mean it's impossible.


So are you saying that doubling the PCIe bus width is not supposed to give a large performance increase? Surly that's not right.

When we saw a 50% increase it seemed mathematically incorrect at the time; one card running on full bus width (x16) and the other card running on half bus width (x8), so you have three parts of a cake (hehe) and when you place both cards on x16 you have four parts of a cake... which equates to a 25% increase in total bus width. So we were expecting around a 15 - 20% performance increase but were unsure how PCIe bus width relates to total power because sending and receiving more data doesn't automatically mean a system can process the extra data at the same speed.

Wow! This is a fantastic example. Actual proof that x8 to x16 makes almost zero difference to a single card setup... but what about an SLi/xFire setup since you have twice the power being hindered by the amount of data being able to transfer.
 
Last edited:
Anyone know? I'm curious if PCI-e bus speed would make a difference in an SLI/xFire setup since the power is there to process the additional data.
 
As long as its running a minimum of x8/x8 you'll be fine - SLI isn't supported below that.

HyperSLI gets around that little problem chap ;)

I ran my 670s in SLI @ X16 X16, X16 X8 and X16 X4.

Basically I was having heat issues, so I moved the bottom card into the very bottom slot which was X4. I had to buy a new SLI bridge and hack it into action with HyperSLI but it did work 100%.

In Firestrike my score was higher :confused: however, when I ran certain games I lost up to 15% of the performance, especially in Crysis 3.

Lesson learned I 'spose. Don't buy open air cards to SLI with.
 
Back
Top Bottom