Why does the DDR memory speed effect FPS ?

[chaparral]

I know I posted this in the GPU section but am guessing you GPU guys would know more about FPS speeds then any other part the forum..


Now back to the thread title question
How does the PC DDR memory speed cause such a big effect on FPS as surely once the game is loaded into ram it shouldn't make any difference to the FPS

But going by these gaming benchmarks your PC memory speed can effect the FPS you get by over 20% eek: 41 minimum FPS to 61minimum FPS just from changing ram speeds )

More Benchmarks here http://www.techspot.com/article/1171-ddr4-4000-mhz-performance/page3.html

 

[shankly1985]

Guessing the faster the RAM transferring data the less waiting the GPU needs to do resulting in higher performance.

 

[chaparral]

Guessing the faster the RAM transferring data the less waiting the GPU needs to do resulting in higher performance.

That a good reason....

I didn't think of that ..

 

[Aretak]

Digital Foundry also investigated this and found that DDR4 scaled well up to ~3000MHz in the games they tested (neither of the above were among them).

One thing I found pretty interesting is that The Witcher 3 actually scaled better with a memory overclock than a CPU overclock.

 

[RavenXXX2]

980ti SLI them benchmarks, single card there is hardly any difference.

 

[Rroff]

DDR4 seems to have a little more effect on framerates than DDR3 on previous platforms but you still have to go to high end multi GPU systems and compare the fastest to the slowest RAM to see significant gains.

 

[chaparral]

DDR4 seems to have a little more effect on framerates than DDR3 on previous platforms but you still have to go to high end multi GPU systems and compare the fastest to the slowest RAM to see significant gains.

But a 38% FPS increase from just changing DDR4 2133 to DDR4-4000 is giving more FPS then upgrading from a 980ti to a 1080...

 

[varkanoid]

Doesnt make any difference on a single card. As said above SLI/Crossfire is where it benefits and those benchmarks were done on SLI.
Plenty of tests and benchmarks were about when DDR4 was introduced with the 170 boards.
Been said many times on here only real reason to get faster DDR is for benchmarking. In games means nowt.

On a high-end gaming system equipped with two GeForce GTX 980 Ti cards.........

 

[Rroff]

But a 38% FPS increase from just changing DDR4 2133 to DDR4-4000 is giving more FPS then upgrading from a 980ti to a 1080...

I'd get a 30% fps increase or something if I changed from say DDR3-800 to my current DDR-2400 on my setup as well doesn't really work like that.

 

[drunkenmaster]

980ti SLI them benchmarks, single card there is hardly any difference.

Yup, sli is cpu limited, so supplying cpu a little faster helps. Single card was so gpu limited that increasing data rate for the cpu had almost no effect.

It also didn't mention(that I spotted from a quick glance) latency settings. if they used the same memory with high latency and only changed the clock speed then you are effectively hurting the cpu performance by reducing clock speed. IF if this were ddr4 4000 with cas 20 and they ran ddr4 2400 also at cas 20 rather than cas 14-15, then it's a relatively silly test as ddr4 2400 cas 14 would gain back some speed.

You'd still expect sli to be more cpu limited and gain both from cpu overclocks and memory overclocks by a much larger amount than single card, but using slower memory settings at lower speeds may artificially inflate the difference they are showing.

 

[TonyTurbo78]

what would have been nice on those benchmarks was to show some DDR3 speeds as well. I've seen plenty of graphs and comparison videos previously saying the difference between DDR3 1600 compared to some DDR4 3000 was a few frames and not worth it. But these provided graphs above certainly hint at something completely different.

 

[Aretak]

Doesnt make any difference on a single card.

Digital Foundry's testing suggests otherwise. Their video shows a notable performance boost going from 2133MHz to 3200MHz with a single overclocked Titan X.

 

[Roughneck]

But a 38% FPS increase from just changing DDR4 2133 to DDR4-4000 is giving more FPS then upgrading from a 980ti to a 1080...

I guarantee that the 38% is specific to Fallout 4, you wont see that in other games, it could just be the way fallout 4 is optimised, Im guessing its not the norm

 

[LegendaryDread]

I take it this only applies to Skylake..?
Perhaps due to the GPUs running at 8x/8x?

I'd love to give it a try myself on X99, but that's an expensive experiment

 

[cadmunkey]

Glad you posted this chaparral , going to do some more research into this myself. Those results are surprising!

 

[chaparral]

I take it this only applies to Skylake..?
Perhaps due to the GPUs running at 8x/8x?

I'd love to give it a try myself on X99, but that's an expensive experiment

No It should apply to all CPU & memory setups but the amount you gain or lose would most likely be different..

 

[nashathedog]

I remember a Corsair promotion for its 2400mhz Vengeance pro ram when Bf4 released where it showed there 2400 ram providing a good 5+ frames in Bf4 over 1600mhz ram.

I imagine it's also part of why the Skylake i5 & i7 has such a good fps advantage over the Devil's canyon i5 & i7 in some games.

 

[Minstadave]

I suspect they've engineered situations that aren't GPU bound to try and show this. I'm not familiar with the Fallout 4 engine but the ARMA 3 engine is a total dog, it's completely CPU bound and mostly runs on a single thread - so anything that will boost CPU performance will help.

In most modern engines on single cards they are much more GPU bound and RAM speed won't do a great deal.

 

[psychas]

not sure if done right you have to include cas rating i would believe to .

as im with 2200 at cas 9 faster then 6700k with 2400 or 2600 at cas 14 or 16 , or even more can't remember need to check benchmarks

 

[Mei]

20% performance gain from memory doesnt seem real in my brain
surprising!

was it sponsored by kingston or corsair or something...not that im jaded.....