Computers & Electronics

Faster RAM makes a difference in games

  • Last Updated:
  • Mar 21st, 2017 1:03 am
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes

Faster RAM makes a difference in games

I just wanted to create this thread to share some benchmarks I ran with you guys. I benchmarked GTA5 with my RAM at 1600mhz, then again at 2400mhz with the CPU cache also overclocked. Here are my results:

Before:
CPU 4.2ghz
CPU cache - stock
DDR3 1600
min 33
max 57
avg 45.4

After:
CPU 4.2ghz
CPU cache - 4ghz
DDR3 2400
min 38
max 66
avg 50.4

That's a 10% gain. There seems to be this conventional wisdom that faster RAM makes no difference in games. I need to run more tests, but I really think I'm on to something here. There's also the notion that overclocking the CPU cache has no effect. I have a feeling that overclocking them both together made a difference.

I ran the test on my i5 4690k CPU with an overclocked GTX 1070 at 4k resolution with all the settings in the game maxed out, aside from the reflection AA which I left at 2x. I also left motion blur turned off.

Anyhow I just thought that was interesting. Food for thought for any of you guys still running Haswell systems if you have not tried overclocking your RAM yet. If you're going to overclock it, though, please be careful. I ran memtest overnight to make sure it was stable.
I am a Zoomer customer.
18 replies
Member
Jan 10, 2017
404 posts
149 upvotes
GTA
I was aware that higher speeds will help ram. It mostly has to do with resources such as textures. Textures sit in ram till your graphics card needs them which need to be copied into gpu ram. Same for code and the cpu. When you notice that the game is taking a long time to load a map (cause by hard drive loading into ram), those are the resources that will be touched by cpu and gpu. So having faster RAM helps out for big chunks of resources.
Information Security Analyst
---
Fido $15/3GB + voip.ms $1/M + 3cx PBX voip push client $0 + Google Server vps $0 = $16/Month voip solution with no battery drain.
Thread
Deal Addict
Nov 22, 2015
2441 posts
1321 upvotes
SickBeast wrote:
Mar 19th, 2017 2:14 pm
I just wanted to create this thread to share some benchmarks I ran with you guys. I benchmarked GTA5 with my RAM at 1600mhz, then again at 2400mhz with the CPU cache also overclocked. Here are my results:

Before:
CPU 4.2ghz
CPU cache - stock
DDR3 1600
min 33
max 57
avg 45.4

After:
CPU 4.2ghz
CPU cache - 4ghz
DDR3 2400
min 38
max 66
avg 50.4

That's a 10% gain. There seems to be this conventional wisdom that faster RAM makes no difference in games. I need to run more tests, but I really think I'm on to something here. There's also the notion that overclocking the CPU cache has no effect. I have a feeling that overclocking them both together made a difference.

I ran the test on my i5 4690k CPU with an overclocked GTX 1070 at 4k resolution with all the settings in the game maxed out, aside from the reflection AA which I left at 2x. I also left motion blur turned off.

Anyhow I just thought that was interesting. Food for thought for any of you guys still running Haswell systems if you have not tried overclocking your RAM yet. If you're going to overclock it, though, please be careful. I ran memtest overnight to make sure it was stable.
I think you might be confused about RAM speed making no difference in games... it certainly does make a difference. However, the price to performance doesn't scale well. I've heard that DD4-2400 is the current sweet spot in terms of value, and going with a faster 3200 or 4000 isn't going to make a perceptible difference.
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes
superfresh89 wrote:
Mar 19th, 2017 2:45 pm
I think you might be confused about RAM speed making no difference in games... it certainly does make a difference. However, the price to performance doesn't scale well. I've heard that DD4-2400 is the current sweet spot in terms of value, and going with a faster 3200 or 4000 isn't going to make a perceptible difference.
You're right, there is definitely a sweet spot. When I bought my DDR3 2133 memory it was only $70 for a 16gb kit on Newegg.ca. It was so cheap that I bought 32gb worth. Overkill, I know, but for that price it was almost an investment. I could probably get $120 for each of those kits today. It's nice that it overclocks so easily to 2400 speeds.

Memory prices have gone up. I would have to look at the DDR4 prices to see what I would do exactly.

Also, just an interesting note for you guys, overclocking my CPU to 4.6ghz only gained me 1fps in that same test. So the memory and perhaps the cache were the bottleneck before. It's faster at 4.2ghz with the faster memory and cache compared to 4.6ghz with slower memory. I will have to test some more to confirm this.
I am a Zoomer customer.
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes
I just had a look and DDR4 3000 may be the current sweet spot if you're looking for the best bang for your buck with DDR4 memory. For example, this kit at Newegg is just as cheap as the DDR4 2400 kits:

https://www.newegg.ca/Product/Product.a ... -_-Product

As time moves on the faster memory will just continue to become cheaper.
I am a Zoomer customer.
Deal Fanatic
User avatar
Nov 21, 2002
7156 posts
978 upvotes
Winnipeg
Ultimately $ determines the value of the 10% performance gain
Deal Fanatic
User avatar
May 9, 2006
9895 posts
918 upvotes
SickBeast wrote:
Mar 19th, 2017 3:31 pm
I just had a look and DDR4 3000 may be the current sweet spot if you're looking for the best bang for your buck with DDR4 memory. For example, this kit at Newegg is just as cheap as the DDR4 2400 kits:

https://www.newegg.ca/Product/Product.a ... -_-Product

As time moves on the faster memory will just continue to become cheaper.
Well except right now... That RAM was under $90, 6 months ago.
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes
joeyjoejoe wrote:
Mar 19th, 2017 5:20 pm
Well except right now... That RAM was under $90, 6 months ago.
True. I was speaking more in a general sense. RAM prices do fluctuate a lot, more than any other system component actually. Over the long term, prices always fall, and speeds go up, though. They're going to reach a ceiling in terms of DDR4 speed, and then economies of scale will take over and the faster speeds will become more mainstream. Then we will be on to DDR5. :)
I am a Zoomer customer.
Deal Addict
User avatar
Oct 9, 2010
1691 posts
332 upvotes
Windsor
Did you mess with timings? I haven't played with "fast" RAM since DDR2 days, and back then (generally), you would increase memory speeds, but have to loosen timings, and you ended up with VERY small gains, basically regardless of what you spent.

I'm curious if you could have tightened your timings @ 1600 to match the relatively looser timings @ 2400.
One who is offended by truth, has no place among those who seek wisdom.
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes
ChubChub wrote:
Mar 19th, 2017 11:13 pm
Did you mess with timings? I haven't played with "fast" RAM since DDR2 days, and back then (generally), you would increase memory speeds, but have to loosen timings, and you ended up with VERY small gains, basically regardless of what you spent.

I'm curious if you could have tightened your timings @ 1600 to match the relatively looser timings @ 2400.
I can run the memory at 1866 at cas 9 at 1.5v. It would probably give me similar performance but I would have less bandwidth for apps like winrar.
I am a Zoomer customer.
Deal Addict
User avatar
Feb 23, 2007
1209 posts
230 upvotes
Toronto
I don't have the numbers to back me up, but the last time I compared differences in ram speed (2666 vs 2133) had negligible speed difference in games. I'm now interested in doing some tests again. I honestly don't expect much gain. I don't think CPU overclock has much impact for me either (again no actual numbers ATM).

I'm running DDR4 quad channel 48 GB @2666.
CPU: Core i7-5820K @ 4.2GHz GPU: GTX980 2xSLI (MSI Gaming 4G, Asus Strix) RAM: G.Skill Ripjaws 4x4GB DDR4-3000 Case: CM Storm Trooper
SSD: Samsung 850 EVO 250GB x2 Raid 0 Mobo: ASRock X99 Extreme4 Cooler: Noctua NH-D15 Dual Fan
PSU: SeaSonic M12II 750W 80+ Bronze Monitor: Dell P2715Q 27.0" UHD
Deal Addict
User avatar
Feb 23, 2007
1209 posts
230 upvotes
Toronto
As promised, ran some tests

Monster Hunter Online Benchmark - 3840x2160 no AA - 2x980 SLI
DDR4 2166 - Score 11499 - AVG FPS 52.3 - MIN FPS 21.4
DDR4 2666 - Score 11517 (+0.16%) - AVG FPS 52.4 (+0.19%) - MIN FPS 22.1 (+3.3%)

Firestrike Ultra - 2x980 SLI
DDR4 2166 - Score 5672 - Graphics Score 5731 - Physics Score 16258 - Combined 2763
DDR4 2666 - Score 5598 (-1.3%) Graphics Score 5612 (-2.1%) - Physics Score 16413 (+0.95%) - Combiend 2791 (+1.01%)

Ghost Recon Wildlands - 2560x1440 - 980 Single GPU
DDR4 2166 - FPS 32.38 - MIN 22.56 - MAX 39.76
DDR4 2666 - FPS 32.78 (+1.24%) - MIN 24.34 (+7.89%) - MAX 38.88 (%2.21)

For my system, it's negligible. I have a tendency to attribute the difference of performance to margin of error.

Processor: 5820k @ 4.2 GHz
GPU: GTX 980 x2
RAM: GSkill Ripjaws 4 4x4GB + 8x4GB DDR4 @ 2667 (15-15-15-35 CR2)
Last edited by LKane1 on Mar 20th, 2017 5:24 pm, edited 1 time in total.
CPU: Core i7-5820K @ 4.2GHz GPU: GTX980 2xSLI (MSI Gaming 4G, Asus Strix) RAM: G.Skill Ripjaws 4x4GB DDR4-3000 Case: CM Storm Trooper
SSD: Samsung 850 EVO 250GB x2 Raid 0 Mobo: ASRock X99 Extreme4 Cooler: Noctua NH-D15 Dual Fan
PSU: SeaSonic M12II 750W 80+ Bronze Monitor: Dell P2715Q 27.0" UHD
[OP]
Deal Guru
User avatar
Mar 14, 2009
10572 posts
3850 upvotes
LKane1 wrote:
Mar 20th, 2017 5:24 pm
As promised, ran some tests

Monster Hunter Online Benchmark - 3840x2160 no AA - 2x980 SLI
DDR4 2166 - Score 11499 - AVG FPS 52.3 - MIN FPS 21.4
DDR4 2666 - Score 11517 (+0.16%) - AVG FPS 52.4 (+0.19%) - MIN FPS 22.1 (+3.3%)

Firestrike Ultra - 2x980 SLI
DDR4 2166 - Score 5672 - Graphics Score 5731 - Physics Score 16258 - Combined 2763
DDR4 2666 - Score 5598 (-1.3%) Graphics Score 5612 (-2.1%) - Physics Score 16413 (+0.95%) - Combiend 2791 (+1.01%)

Ghost Recon Wildlands - 2560x1440 - 980 Single GPU
DDR4 2166 - FPS 32.38 - MIN 22.56 - MAX 39.76
DDR4 2666 - FPS 32.78 (+1.24%) - MIN 24.34 (+7.89%) - MAX 38.88 (%2.21)

For my system, it's negligible. I have a tendency to attribute the difference of performance to margin of error.

Processor: 5820k @ 4.2 GHz
GPU: GTX 980 x2
RAM: GSkill Ripjaws 4 4x4GB + 8x4GB @ 2667 (15-15-15-35 CR2)
Interesting. Perhaps the memory creates a bottleneck but only up to a certain speed. Itested 1600 vs. 2400. Perhaps 2400 is the sweet spot, or even 2133.
I am a Zoomer customer.
Deal Addict
User avatar
Feb 23, 2007
1209 posts
230 upvotes
Toronto
SickBeast wrote:
Mar 20th, 2017 5:35 pm
Interesting. Perhaps the memory creates a bottleneck but only up to a certain speed. Itested 1600 vs. 2400. Perhaps 2400 is the sweet spot, or even 2133.
Have you tried benchmarking the differences in memory ONLY? And not memory AND CPU cache?
CPU: Core i7-5820K @ 4.2GHz GPU: GTX980 2xSLI (MSI Gaming 4G, Asus Strix) RAM: G.Skill Ripjaws 4x4GB DDR4-3000 Case: CM Storm Trooper
SSD: Samsung 850 EVO 250GB x2 Raid 0 Mobo: ASRock X99 Extreme4 Cooler: Noctua NH-D15 Dual Fan
PSU: SeaSonic M12II 750W 80+ Bronze Monitor: Dell P2715Q 27.0" UHD

Top