Comparison of GTX960 and GTX1080

[smartroad]

Okay, this isn't that scientific I'll admit but a quick comparison of instantaneous FPS from my old GTX960 and my new GTX1080 As Elite doesn't have any benchmarking options (at least that I am aware of) I tried to get similar shots for each screen cap (like I said, not scientific!)

http://imgur.com/a/51Li3

[Sevin Church]

Did you run the fps counter through the GeForce experience utility to get the fps count?

Screenshots look great. I'm currently running a 980ti.

[Cmdr Kharma]

Looks like in game Ctrl-F ....

[Sevin Church]

Indeed, but not what I was asking.

[smartroad]

I jut used Elite's own FPS counter. I only done a quick screen grab aI was too excited to try out a longer test lol

I did do a couple of 3D Mark tests, on the fire strike utlta the 1080 was about 140% faster. Although my old i5 2550k cpu could be bringing it down a bit.

[thebs]

I jut used Elite's own FPS counter. I only done a quick screen grab aI was too excited to try out a longer test lol

I did do a couple of 3D Mark tests, on the fire strike utlta the 1080 was about 140% faster. Although my old i5 2550k cpu could be bringing it down a bit.

Unless you're comparing that i5 to an i7, an i5-2xxx won't be much of a slouch to even an i5-6xxx product.

The GTX 960 2GiB is really a step down from even a GTX 970 4GiB, so it doesn't surprise me one bit that you're seeing well under half the performance on the GTX 960 2GiB compared to a GTX 1080 with several times over the VRAM.

As others have noted, 4-5GiB of buffer+texture usage is typical with Elite on Ultra, which explains why we're seeing the GTX 980Ti 6GiB "holding it's own" against the GTX 1070 8GiB, even at 4K, while the other GTX 900 models with 4GiB or less drop off at 1440p and even 1080p.

The GTX 1080 is the new performance king, while the GTX 1070 is the coolest customer that matches or slightly exceeds the GTX 980Ti, especially if it ever gets to its $379 list price. The GTX 980Ti's saving grace is that 6GiB, and it has many more units than the much higher clocked, but newer 16nm FET process GTX 1070.



Disclaimer: Please excuse any grammar, pathetic typos, or satanic versus known as "auto-correct" as this was posted via Tapatalk from my budget Honor 5X

[smartroad]

@thebs

My 960 had a 4gb framebuffer, spent a bit more because I knew Elite was quite demanding on the buffer

Currently overclocking my 1080 at a pretty steady 2012MHz and at 62C well chuffed

[JustSomeGuy]

I'm now between RX 480 and GTX 1070. Both are lightyears ahead of my current GTX 660 2gb. Amd would be half the price but I wonder how it performs in real life on 1440p. I would like to see reviews of that Sapphire custom version. Gtx 1070 would be fast for sure on 1440p and better bang for buck compared to GTX 1080.
I have i5-3570K @ 4.4GHz (default 3.4GHz) so I guess it'll do well.
Right now I am leaning towards the green camp.

[thebs]

@thebs

My 960 had a 4gb framebuffer, spent a bit more because I knew Elite was quite demanding on the buffer

Currently overclocking my 1080 at a pretty steady 2012MHz and at 62C well chuffed

Oh, cool, sorry I missed that. Better explains why you're still at 40-50% in the benchmarks.

Disclaimer: Please excuse any grammar, pathetic typos, or satanic versus known as "auto-correct" as this was posted via Tapatalk from my budget Honor 5X

[thebs]

I'm now between RX 480 and GTX 1070. Both are lightyears ahead of my current GTX 660 2gb. Amd would be half the price but I wonder how it performs in real life on 1440p. I would like to see reviews of that Sapphire custom version. Gtx 1070 would be fast for sure on 1440p and better bang for buck compared to GTX 1080.
I have i5-3570K @ 4.4GHz (default 3.4GHz) so I guess it'll do well.
Right now I am leaning towards the green camp.

We still use AMD in a lot of applications, from their superior FPU precision (when one is actually using double precision floating point and not those "lossy" SIMD instructions --AMD had always bested Intel here**), to pushing around bits through its superior interconnect (Intel options that are equivalent are just too costly in comparison). Intel has also screwed around at the low-end with Atom for the last decade, and utterly lost to ARM (ever time there's a newer Atom, there's already a newer ARM that destroys it), and has only recently brought the i-Core to the System-on-a-Chip (SoC) to finally start competing with AMD when x86 is required.

But Intel has the performance consumer market. I'm hoping AMD's new architecture brings Intel's pricing down. But AMD just cannot compete with Intel's fabrication lead, which went from 12 months to 36 months over the last decade.

** Yes, even the AMD K5/K6 destroyed the Pentium's FPU in performance and precision, let alone loads were 3x faster in the K5-90+ and K6 ALU (both were actually NexGen's RISC86). The problem was that because the Pentium ALU was so poorly designed (something Digital helped Intel solve with the Pentium Pro, which is the ALU in the P2-P4 as well), it was faster to load 2x 32-bit integers through the Pentium's pipelined FPU than it's multi-issue, but defective, ALU. This led most people to believe the Pentium had a superior FPU, when it did not ... ALU integer code "optimized for Pentium" was designed to bypass it's buggy ALU load instruction, so anything else running that code would suffer if the FPU load wasn't pipelined. E.g., ALU to ALU, 32-bit integers in Doom code loaded 200% faster (3x as fast) on AMD than Intel, and still even 50% faster (1.5x as fast) on AMD's ALU versus via Intel FPU, but ID didn't build a seperate binary for AMD, since Pentium has 93% of the market, so AMD products until the K7 suffered a major performance hit in such "Pentium optimized" code.

In the Linux world, this is why we optimized for i486, then i686 (PPro-P4), at least until in-line order only Atom (which takes a 20% performance hit with GCC i686 optimized objects, while only a 2% hit with Atom optimized on P4, while AMD K7 was actually faster), and skipped i586 optimizations. A few distros did i586 to their own peril, although they did win benchmarks on the specific Pentium P54/P55C units.

Disclaimer: Please excuse any grammar, pathetic typos, or satanic versus known as "auto-correct" as this was posted via Tapatalk from my budget Honor 5X