So it looks like memory speed does matter when CPU bound

[SteveGrabowski0]

In this Digital Foundry benchmark they showed some pretty big gains on an i3-6100 by using 2666 MHz DDR4 vs the standard 2133 MHz DDR4. So the Z series boards and good RAM look like they can be worth the money, I mean an 8 fps gain in Witcher 3 is nothing to sneeze at.

 

[Nena Trinity]

I love this channel it got so much cool stuff! 

[NeatSquidYT]

holy fucking shit guys 8fps better spend at least 30 quid more on RAM

[JuztBe]

@wolfgang
Could you OC your RAM for science?! And see if it makes any difference in CPU bound titles
______________________________
 

[Guest]

who's idea was it to say

Lets test low speed RAM on a high end CPU and High speed RAM on a low end CPU...

Seriously.

You'd put that money into A. A CPU, B. A GPU C. Storage not RAM for christ sake. Now wait for Skylake-E to come out and test it on a dual core with hyperthreading and a hexa core or octa core with RAM speeds

[SteveGrabowski0]

holy fucking shit guys 8fps better spend at least 30 quid more on RAM

 

Look at the Ryse results where it's 50 fps. Similar on DDR3 too, as shown in the video below.

 

[Me1z]

holy fucking shit guys 8fps better spend at least 30 quid more on RAM

 

That's actually not half bad. If you consider the performance premium you'd be paying to get a higher tier of GPU or CPU, 30 quid ($45 USD) is a pretty good deal. The price difference between the i5 and the similarly performing i3, thanks to overclocked RAM, is about 50 quid, so I'd take the faster RAM.

[SteveGrabowski0]

who's idea was it to say

Lets test low speed RAM on a high end CPU and High speed RAM on a low end CPU...

Seriously.

You'd put that money into A. A CPU, B. A GPU C. Storage not RAM for christ sake. Now wait for Skylake-E to come out and test it on a dual core with hyperthreading and a hexa core or octa core with RAM speeds

 

It was Digital Foundry's idea, and it was a good one since it seems to overthrow gamer dogma that RAM speed doesn't matter. It shows it does when CPU bound, and for people who want to keep their CPUs for years it shows it might end up being worth it to go with a Z series board and faster RAM. I found it a pretty eye opening video since there haven't been any others I know of that show how RAM speed affects gaming performance when CPU bound.

[Guest]

It was Digital Foundry's idea, and it was a good one since it seems to overthrow gamer dogma that RAM speed doesn't matter. It shows it does when CPU bound, and for people who want to keep their CPUs for years it shows it might end up being worth it to go with a Z series board and faster RAM. I found it a pretty eye opening video since there haven't been any others I know of that show how RAM speed affects performance when CPU bound.

and what is the CAS on the RAM aswell because that seems to be lacking

and the brand 

for all we know this could be dominator platinum 

[SteveGrabowski0]

and what is the CAS on the RAM aswell because that seems to be lacking

and the brand 

for all we know this could be dominator platinum 

 

They didn't say. I don't get your point about putting money into storage, when that has never been shown to give any fps gain while memory has been shown to have anywhere from good to night and day gains. Can you show me a benchmark where a 950 Pro or Intel 750 gets you more fps than a Seagate 7200 RPM drive?

[Sauron]

Or you could go for a b chipset and an i5.

[SteveGrabowski0]

Digital Foundry really is the best in the business. They were the ones who came and shat all over the Pentium G3258 with FCAT results back when the other sites were saying hurr durr 4.2 GHz G3258 is like an i7. They exposed how lousy AMD's DirectX 11 performance is when using low end CPUs. And now they have shown memory actually can matter quite a bit even in games if you're running into CPU limitations. It's awesome that they do a lot of testing in nonstandard configurations to isolate different variables.

[Guest]

They didn't say. I don't get your point about putting money into storage, when that has never been shown to give any fps gain while memory has been shown to have anywhere from good to night and day gains. Can you show me a benchmark where a 950 Pro or Intel 750 gets you more fps than a Seagate 7200 RPM drive?

no,

thats an unfair example, I can get you a load time

but if you insist how RAM speed and CAS don't make a difference 

chart here should have no point correct? Oh heres the load times between a 730, 850 and seagates SSHD, since the 950 is still fairly new we wont see it on benchmarks for a while  

[SteveGrabowski0]

but if you insist how RAM speed and CAS don't make a difference 

 

Where did I insist that? I never said CAS didn't matter. I said I didn't know what CAS speed RAM they were using. I think it's pretty cool they showed that RAM speed can matter in CPU bound situations, which goes completely against what most gamers believe. I wish more sites were willing to put standard accepted knowledge to the test to see whether it holds up.

[SteveGrabowski0]

holy fucking shit guys 8fps better spend at least 30 quid more on RAM

 

Why is 8 fps something stupid to care about? That's the difference between a GTX 960 and a GTX 780 Ti in this benchmark.

 

 

EDIT: Oops, the video was done at high, not ultra. So a more apples to apples comparison would be that it's in line with the difference between a GTX 960 and a GTX 780 at that resolution and preset:

 

 

I don't think many people would sneer at paying 30 pounds more for a 780 than a 960.

[SteveGrabowski0]

and what is the CAS on the RAM aswell because that seems to be lacking

and the brand 

for all we know this could be dominator platinum 

 

OK, I saw in the associated article it's Corsair Vengeance LPX low profile DDR4 run at full speed 2666 MHz and then scaled back to 2133 MHz. On newegg all the DDR4 2666 Vengeance is either CAS 15 or 16.

[Necrodead]

instead of the i5 to i7 comparison they should have done a lower speed ram comparison on one of those processors.

[App4that]

How about a GPU you would actually place with those CPU's and Ram configurations? And are we talking about drawing from a SSD or HDD? I used to go to these guys all the time but I'm seeing some conflicting tests using rational GPU's.

[Fate]

holy fucking shit guys 8fps better spend at least 30 quid more on RAM

Are you being sarcastic? £30 for 8fps/10-15% performance gain is an amazing value. You get only 6% more performance going from a 390 to a 390x, and that's a $70 difference. A 970 to 980 is only a 12% difference on average.

[SteveGrabowski0]

How about a GPU you would actually place with those CPU's and Ram configurations? And are we talking about drawing from a SSD or HDD? I used to go to these guys all the time but I'm seeing some conflicting tests using rational GPU's.

 

It wouldn't make sense to do that. The whole point of using an overkill GPU at a low resolution like 1080p is to minimize the effect of the GPU to isolate the memory for the test. Otherwise you're just testing the GPU, like in the ridiculous Skylake video done here where they tested the i5-6600k vs the i5-2500k at 4k with a single GPU, where the cpu wouldn't really matter. If they just built an i3 + GTX 960 + DDR4 2133 system and ran it against an i7 + 980 Ti + DDR4 2666 system it would completely mask the difference memory speed makes because GPU and CPU would dominate the results. It's the same reason why GPUs are almost always tested with heavily overclocked 5930k, 5960x, 4960x, etc.

[App4that]

It wouldn't make sense to do that. The whole point of using an overkill GPU at a low resolution like 1080p is to minimize the effect of the GPU to isolate the memory for the test. Otherwise you're just testing the GPU, like in the ridiculous Skylake video done here where they tested the i5-6600k vs the i5-2500k at 4k with a single GPU, where the cpu wouldn't really matter. If they just built an i3 + GTX 960 + DDR4 2133 system and ran it against an i7 + 980 Ti + DDR4 2666 system it would completely mask the difference memory speed makes because GPU and CPU would dominate the results. It's the same reason why GPUs are almost always tested with heavily overclocked 5930k, 5960x, 4960x, etc.

That makes sense. In that case I hope they take pity on us last gen'ers and do one on the DDR3 lol. Your explanation probably explains the conflict I saw in the results from other tubers.

[SteveGrabowski0]

That makes sense. In that case I hope they take pity on us last gen'ers and do one on the DDR3 lol. Your explanation probably explains the conflict I saw in the results from other tubers.

 

They did one showing DDR3 1600 vs DDR3 2133 on the i3-4130 also as part of their i3-6100 review. It's not as big of a difference but still pretty significant.

 

[Nena Trinity]

After seeing these videos I was happy I got 1866MHz RAM for FX-8370 & 2133MHz RAM for Athlon X4 860K! :3

 

They did one showing DDR3 1600 vs DDR3 2133 on the i3-4130 also as part of their i3-6100 review. It's not as big of a difference but still pretty significant.

 

To bad they did not test FX-6300 with 2133MHz!

[SteveGrabowski0]

After seeing these videos I was happy I got 1866MHz RAM for FX-8370 & 2133MHz RAM for Athlon X4 860K! :3

 

To bad they did not test FX-6300 with 2133MHz!

 

Yeah, I was floored to see the difference memory makes. I hope they'll get a better AMD board so they can run the same kind of tests on the FX-6300. In this one they said they couldn't get the RAM up to 2133 MHz for their FX-6300 test.

[App4that]

They did one showing DDR3 1600 vs DDR3 2133 on the i3-4130 also as part of their i3-6100 review. It's not as big of a difference but still pretty significant.

 

 

Mother fucker :lol: , "memory speed doesn't matter in gaming" they said, so I bought 1600MHz Latency 9 Oh well, split milk and all. Thanks for the link.

 

 

EDIT, watch that 6300 AMD go though, that's surprising.