The difference between GTX-1070Ti SLI and GTX-1080FE SLI barely registers on benchmarks

[RealNeil]

I pulled the 1080FE cards out of my i9-7900X box and replaced them with a pair of Vega-64 cards.
Having the two 1080FE cards laying around, I decided to swap them for my two 1070Ti cards.

I ran before and after benches on the system to see how much the 1080s gave me over the 1070Ti cards.
There is barely any juice here.

 

[king of swag187]

This isn't uprising given that the 1070 ti is just under-binned 1080.

 

[trog100]

just for reference my pair of 1070 cards in sli scored 11800 in time spy and my 2080ti scores 14400...

i think time spy does scale over 90% though..

trog

 

[FordGT90Concept]

This isn't uprising given that the 1070 ti is just under-binned 1080.

Yup, 1070 ti isn't much slower than the 1080. SLI them, they will get similar results as well.

 

[RealNeil]

Yup, 1070 ti isn't much slower than the 1080. SLI them, they will get similar results as well.

My 1070Ti cards ~and~ my 1080FE cards are already running in SLI.

 

[Deleted member 67555]

Sounds about right...
My 1070ti is as good as the crappiest 1080's in most benchmarks...
I kinda wish I had another to SLI just to do it... I'm a little jelly

 

[eidairaman1]

I guess for me i just reinstall drivers as a precaution

 

[newtekie1]

My 1070Ti cards ~and~ my 1080FE cards are already running in SLI.

I think that was his point. The cards by themselves are within a few percent of eachother, so you're configuration in SLI will also be withing a few percent.

I mean, that was the whole thing behind nVidia not allowing pre-overclocked versions of the GTX 1070Ti, because it would match a GTX 1080 with just a very minor overclock.

 

[Durvelle27]

This is the reason I chose to skip the GTX 1070 Ti and go to the RTX 2070 when I was looking st new cards.

 

[newtekie1]

This is the reason I chose to skip the GTX 1070 Ti and go to the RTX 2070 when I was looking st new cards.

It kind of makes me wish I had grabbed the GTX 1070Ti Amazon had for $350 last week.

 

[cucker tarlson]

I guess for me i just reinstall drivers as a precaution

this will change nothing.
7% in graphics score is just about correct,given that 1080 is maybe up to 10% faster than 1070ti and he's running sli,where the faster cards you pair,the lesser the scaling.

In the other test you're clrealy bottlnecked at 500-700 fps.

All in all, the difference between 1070ti sli and 1080 sli in real world would be unnoticeable and this is pretty much a confirmation.

Frankly I thought of 1070Ti SLI as an upgrade for my 1080 at some point, since they were very attractively priced and in games that don't scale I'd see almost no performance decrease compared to a 1080. In the end I went with a less powerful but ridiculously quiet 1080Ti trio.

 

[newtekie1]

Frankly I thought of 1070Ti SLI as an upgrade for my 1080 at some point, since they were very attractively priced and in games that don't scale I'd see almost no performance decrease compared to a 1080. In the end I went with a less powerful but ridiculously quiet 1080Ti trio.

I started using SLI when I had two 7800GTX's, and I've used it every GPU generation since. But now, SLI has really fallen out of favor, and the scaling in a lot of games just wasn't where I think it should be anymore. Most new games really don't run all that faster with two cards than one. Even last generation, my main rig was a pair of GTX970's, and my "backup" gaming rig had a GTX980Ti. I had the 970's first, and the difference was so minimal I didn't even feel the need to bother with the hassle of tearing apart my main computer to swap out to the 980Ti. But now, I just wasn't happy with the GTX1070's I had in SLI. Synthetics were still great, but they just didn't scale that well in the new games anymore. So I went with the single GTX1080Ti, and it is so nice. I think nVidia, and game developers, are really putting multi-card optimization on the back burner and just treating it like an after-thought now.

 

[EarthDog]

We should know these cards are close in the first place. TPUs results have it ~4% apart (oc version to fe). Youll see less gap at low res in night hawk.

Anyway... seems about right. That really isnt an upgrade.

 

[RealNeil]

In the end I went with a less powerful but ridiculously quiet 1080Ti trio.

I have one of those in a box on the shelf. Waiting to find another one for SLI. It's an extremely powerful GPU.

Anyway... seems about right. That really isnt an upgrade.

Not trying to upgrade. I had the 1080FE cards in another system and removed them for a pair of Vega-64 cards.
I knew that they would be close to the 1070Ti cards, but a little better. So I swapped them out. (they were here, so why not)

 

[trog100]

I have one of those in a box on the shelf. Waiting to find another one for SLI. It's an extremely powerful GPU.



Not trying to upgrade. I had the 1080FE cards in another system and removed them for a pair of Vega-64 cards.
I knew that they would be close to the 1070Ti cards, but a little better. So I swapped them out. (they were here, so why not)

and were surprised at just how much little "better".. he he

the 1070ti cards are very good value at the moment.. which is something i keep saying..

trog

 

[newtekie1]

I have one of those in a box on the shelf. Waiting to find another one for SLI. It's an extremely powerful GPU.



Not trying to upgrade. I had the 1080FE cards in another system and removed them for a pair of Vega-64 cards.
I knew that they would be close to the 1070Ti cards, but a little better. So I swapped them out. (they were here, so why not)

The interesting thing, if the 1070Tis use aftermarket coolers, they might actually perform better in longer gaming sessions. The FE cooler can limit the 1080s. And when you overclock the 1070Tis, they should outperform the 1080s for sure.

 

[Vayra86]

1070ti would be the last thing I'd SLI out of Pascal's highend range, because the balance is already tight on VRAM bandwidth. No GDDR5X. When you push these at high res, the gap will increase, and it won't favor the 1070ti SLI.

This is one of those rare cases where SLI 1080FE is actually the better choice.

 

[newtekie1]

1070ti would be the last thing I'd SLI out of Pascal's highend range, because the balance is already tight on VRAM bandwidth. No GDDR5X. When you push these at high res, the gap will increase, and it won't favor the 1070ti SLI.

This is one of those rare cases where SLI 1080FE is actually the better choice.

The actual memory bandwidth difference between a 1080FE with GDDR5X and a 1070/Ti with GDDR5 is not that much.

 

[Hockster]

I'll toss this in for a comparison. The 8700K was clocked to 5.0 across all cores, the 2700X to 4.0 across all cores, and the brand new 9900K/2080Ti were stock clocks. For some reason I could never get the same OC on the GPU's in the Ryzen system.

 

[RealNeil]

I'll toss this in for a comparison. The 8700K was clocked to 5.0 across all cores, the 2700X to 4.0 across all cores, and the brand new 9900K/2080Ti were stock clocks. For some reason I could never get the same OC on the GPU's in the Ryzen system.

Lofty scores indeed!

 

[HD64G]

I'll toss this in for a comparison. The 8700K was clocked to 5.0 across all cores, the 2700X to 4.0 across all cores, and the brand new 9900K/2080Ti were stock clocks. For some reason I could never get the same OC on the GPU's in the Ryzen system.

I thought Ryzen 2700X (and all Ryzen 2X00X series) works better without any manual oc. Did you try it stock?

 

[Hockster]

I thought Ryzen 2700X (and all Ryzen 2X00X series) works better without any manual oc. Did you try it stock?

That was using Ryzen Master for the OC. Very few 2700X's get more than 4.0, and need a ton of voltage. At 4.0 it ran cooler than the 6 core 8700K, but 35% lower clocks probably help. Gotta say though I'm impressed with how cool this 9900K runs. Got it at 5.0 across all cores and it's nice, but not the point of this thread lol