These are mostly pointless for marketing but would be fun to see if AMD or Intel see much benefit to such high speed low latency memory. These would provide a nice far edge case.
In many ways, memory tuning has replaced "overclocking" for AMD Ryzen CPUs (partially because Ryzen CPUs self-overclock according to thermals, amongst other factors, but also because memory tuning can result in similar FPS improvements) and can have a tangible benefit to various applications, but notably games. While memory tuning also helps Intel CPUs, it's not to the same degree as seeing possible ~10% FPS gains.
>ONLY the 4x32GB=256GB kits bring 4x2=10% improvement.
https://youtu.be/iHJ16hD4ysk?t=263 Not sure what you're smoking dude, same video I linked. Compare DDR4-2133 (JEDEC standard memory) (167 fps) and a DDR4-3200 CL14 kit at just XMP timings (182). 182/167 = 1.089 or 108.9% or 8.9% uplift, the memory timing can give another few percent. Hence the ~10% figure I gave. The performance gains aren't just from memory capacity, but scale very well from lower memory latency and higher memory speeds.
Look at the difference between DDR4-3800 [MANUAL] (manually tuned timings) and DDR4-3800 [CL16] (XMP timings only) for the GeForce RTX 2080 Ti section for Assassin's Creed Odyssey. 91 FPS avg (tuned) vs 80 FPS avg (xmp). Tuning the subtimings while keeping the same memory transfer speed gave a 91/80 = 1.1375 = 13.75% performance uplift over XMP. And the difference is even bigger over DDR4 3000 XMP, and would be that much of a larger uplift over default JEDEC-2133 timings.
So yes, on Ryzen platforms, it's not unusual to see ~10% gains in games just from better memory and/or better memory tuning. This is similar to seeing ~10% FPS gains on a 2500k by overclocking the CPU from 3.7GHz to ~4.3GHz. But in Ryzen's case today (since CPU clocks are largely determined by how big of a cooler you mount on it and how loud you run the fans), the same BIOS setting tweaking instead happens today in the memory section for similar ~10% gains.
You're still making literally zero sense. The other anandtech article just posts an advertisement/announcement of availability of other G.skill memory kits, in particular 32GB per single DIMM sets (reaching system totals of up to 256GB), whereas this advertisement/announcement of availability is for a 4x8GB=32GB total memory kit.
The people who are buying low-latency high-speed RAM aren't necessarily the same ones buying kits with as much capacity as possible.
Furthermore, reread the original post: "...but would be fun to see if AMD or Intel see much benefit to such high speed low latency memory. These would provide a nice far edge case." The poster is asking how big of a difference is there between using default speed memory vs the fastest available memory kit on the market, and if there is a differential between the two, is it worth it to buy "fast memory".
My response is essentially tl;dr: Yes, particularly for Ryzen systems these days, fast, low latency memory kits and manually tuning subtimings can provide ~10% gains or so, and I provide graphs and sources.
Your response to me is a link to another G.Skill anandtech article of a completely different memory kit, that wasn't reviewed at Anandtech, and you're supposedly saying only these different high capacity memory kits can provide that kind of 10% improvement, which is both wrong and an unsubstantiated claim on your part.
Dude, keep up, your reading comprehension is really garbage at following the conversation.
Really, really slow... I mean even literally quoting what you should look at didn't light up a candle in that darkness. Ok, let me spell it out for you: Anton Shilov's articles are garbage, they are full of embarrassing errors that never get corrected, and they deserve to be shamed with every opportunity.
Fitting name Joey Jo Jo. Now stop struggling so much to insult me or explain stuff to me. You're not really up to either task.
Now it's clear. I'm only taking the information/source you're presenting at face value, because that's exactly what I was doing in my reply, providing an honest source that didn't have any errors (typographical or otherwise). The problem is that you're presenting an irrelevant (and incorrect total memory capacity) article in response to an earnest question and earnest reply in the comments, and then you're surprised that people aren't picking up on what is (apparently) unexplained internet sarcasm.
Normally people expect sarcastic replies in response to sarcastic comments -- Neither my answer or FreckledTrout's question were sarcastic, which is why your "4x32GB=256GB" kits bring "4x2=10%" didn't make any sense since you're coming so far out of left field.
>Fitting name Joey Jo Jo. Now stop struggling so much to insult me or explain stuff to me. You're not really up to either task. Or in other words: >Oh no! Somebody's not picking up on my uninitiated internet sarcasm! You're hurting my feelings stop insulting me, and insult Anton Shilov instead!
Joey, your response was to the original posters question and was well done. Significant performance uplift, particularly on Ryzen platforms, is obtainable by getting higher speed memory and sub-timings. I've seen charts showing almost 25% uplift going from 2133 to 4000ish if the sub-timings weren't nerfed to make that happen.
Whether that justifies the price premium GSkill and others demand for these binned modules on the other hand is an entirely different question.
I completely agree with the author of the blog I want to advise everyone who is engaged in writing an excellent essay https://ca.payforessay.net/ and article writing service that will help you professionally write an article on any topic
lmao, retards linking to fucking youtube videos to prove a point. I ain't gonna click that shit. you are dumb, and overpriced overclocked memory is fucking dumb. enjoy ricing out your AMD until it burns your house down, I'll stick with intel and air coolers.
Don't forget the infinity fabric overclock (as shown by the good Antony at LTT). Reading and viewing video around the best ram performance with ryzen is achieved by having 4 ranks switching to 1T (from tom's hw) (dual rank with 1T duhh?) Lowest timings possible (common knowledge) Infinity fabric OC to get the highest ram frequency maintaining 1:1 ratio
So In my case (itx) I should buy a 2x16 kit capable to reaching 3800mhz Cas 14 1T. It doesn't exist and in my little knowledge, dual rank memory is not 2T by design?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
14 Comments
Back to Article
FreckledTrout - Monday, October 21, 2019 - link
These are mostly pointless for marketing but would be fun to see if AMD or Intel see much benefit to such high speed low latency memory. These would provide a nice far edge case.JoeyJoJo123 - Monday, October 21, 2019 - link
There are notable benefits to using low-latency memory at high speeds.https://www.youtube.com/watch?v=iHJ16hD4ysk
In many ways, memory tuning has replaced "overclocking" for AMD Ryzen CPUs (partially because Ryzen CPUs self-overclock according to thermals, amongst other factors, but also because memory tuning can result in similar FPS improvements) and can have a tangible benefit to various applications, but notably games. While memory tuning also helps Intel CPUs, it's not to the same degree as seeing possible ~10% FPS gains.
close - Monday, October 21, 2019 - link
Only the 4x32GB=256GB kits bring 4x2=10% improvement.https://www.anandtech.com/show/14964/gskill-launch...
JoeyJoJo123 - Monday, October 21, 2019 - link
>ONLY the 4x32GB=256GB kits bring 4x2=10% improvement.https://youtu.be/iHJ16hD4ysk?t=263
Not sure what you're smoking dude, same video I linked. Compare DDR4-2133 (JEDEC standard memory) (167 fps) and a DDR4-3200 CL14 kit at just XMP timings (182). 182/167 = 1.089 or 108.9% or 8.9% uplift, the memory timing can give another few percent. Hence the ~10% figure I gave. The performance gains aren't just from memory capacity, but scale very well from lower memory latency and higher memory speeds.
LTT not good enough? How about Hardware Unboxed?
https://youtu.be/iH3qq_mSxTM?t=531
Look at the difference between DDR4-3800 [MANUAL] (manually tuned timings) and DDR4-3800 [CL16] (XMP timings only) for the GeForce RTX 2080 Ti section for Assassin's Creed Odyssey. 91 FPS avg (tuned) vs 80 FPS avg (xmp). Tuning the subtimings while keeping the same memory transfer speed gave a 91/80 = 1.1375 = 13.75% performance uplift over XMP. And the difference is even bigger over DDR4 3000 XMP, and would be that much of a larger uplift over default JEDEC-2133 timings.
So yes, on Ryzen platforms, it's not unusual to see ~10% gains in games just from better memory and/or better memory tuning. This is similar to seeing ~10% FPS gains on a 2500k by overclocking the CPU from 3.7GHz to ~4.3GHz. But in Ryzen's case today (since CPU clocks are largely determined by how big of a cooler you mount on it and how loud you run the fans), the same BIOS setting tweaking instead happens today in the memory section for similar ~10% gains.
close - Monday, October 21, 2019 - link
I'll just give you a second and leave that link I posted above there... maybe it just sinks in slower :).Kit Config. Kit Capacity
1×32 GB 32 GB
2×32 GB 64 GB
3×32 GB 128 GB
4×32 GB 256 GB
JoeyJoJo123 - Tuesday, October 22, 2019 - link
You're still making literally zero sense. The other anandtech article just posts an advertisement/announcement of availability of other G.skill memory kits, in particular 32GB per single DIMM sets (reaching system totals of up to 256GB), whereas this advertisement/announcement of availability is for a 4x8GB=32GB total memory kit.The people who are buying low-latency high-speed RAM aren't necessarily the same ones buying kits with as much capacity as possible.
Furthermore, reread the original post:
"...but would be fun to see if AMD or Intel see much benefit to such high speed low latency memory. These would provide a nice far edge case."
The poster is asking how big of a difference is there between using default speed memory vs the fastest available memory kit on the market, and if there is a differential between the two, is it worth it to buy "fast memory".
My response is essentially tl;dr: Yes, particularly for Ryzen systems these days, fast, low latency memory kits and manually tuning subtimings can provide ~10% gains or so, and I provide graphs and sources.
Your response to me is a link to another G.Skill anandtech article of a completely different memory kit, that wasn't reviewed at Anandtech, and you're supposedly saying only these different high capacity memory kits can provide that kind of 10% improvement, which is both wrong and an unsubstantiated claim on your part.
Dude, keep up, your reading comprehension is really garbage at following the conversation.
close - Tuesday, October 22, 2019 - link
Really, really slow... I mean even literally quoting what you should look at didn't light up a candle in that darkness. Ok, let me spell it out for you: Anton Shilov's articles are garbage, they are full of embarrassing errors that never get corrected, and they deserve to be shamed with every opportunity.Fitting name Joey Jo Jo. Now stop struggling so much to insult me or explain stuff to me. You're not really up to either task.
JoeyJoJo123 - Tuesday, October 22, 2019 - link
Now it's clear. I'm only taking the information/source you're presenting at face value, because that's exactly what I was doing in my reply, providing an honest source that didn't have any errors (typographical or otherwise). The problem is that you're presenting an irrelevant (and incorrect total memory capacity) article in response to an earnest question and earnest reply in the comments, and then you're surprised that people aren't picking up on what is (apparently) unexplained internet sarcasm.Normally people expect sarcastic replies in response to sarcastic comments -- Neither my answer or FreckledTrout's question were sarcastic, which is why your "4x32GB=256GB" kits bring "4x2=10%" didn't make any sense since you're coming so far out of left field.
>Fitting name Joey Jo Jo. Now stop struggling so much to insult me or explain stuff to me. You're not really up to either task.
Or in other words:
>Oh no! Somebody's not picking up on my uninitiated internet sarcasm! You're hurting my feelings stop insulting me, and insult Anton Shilov instead!
rahvin - Tuesday, October 22, 2019 - link
Joey, your response was to the original posters question and was well done. Significant performance uplift, particularly on Ryzen platforms, is obtainable by getting higher speed memory and sub-timings. I've seen charts showing almost 25% uplift going from 2133 to 4000ish if the sub-timings weren't nerfed to make that happen.Whether that justifies the price premium GSkill and others demand for these binned modules on the other hand is an entirely different question.
MikeJason77 - Monday, April 6, 2020 - link
I completely agree with the author of the blog I want to advise everyone who is engaged in writing an excellent essay https://ca.payforessay.net/ and article writing service that will help you professionally write an article on any topicerotomania - Tuesday, October 22, 2019 - link
4x32 = 256 ???timecop1818 - Tuesday, October 22, 2019 - link
lmao, retards linking to fucking youtube videos to prove a point. I ain't gonna click that shit. you are dumb, and overpriced overclocked memory is fucking dumb. enjoy ricing out your AMD until it burns your house down, I'll stick with intel and air coolers.JoeyJoJo123 - Tuesday, October 22, 2019 - link
>posts proof>"i don't like that proof! therefore it's invalid!"
Enjoy being ignorant I guess?
umano - Thursday, October 24, 2019 - link
Don't forget the infinity fabric overclock (as shown by the good Antony at LTT).Reading and viewing video around the best ram performance with ryzen is achieved by
having 4 ranks
switching to 1T (from tom's hw) (dual rank with 1T duhh?)
Lowest timings possible (common knowledge)
Infinity fabric OC to get the highest ram frequency maintaining 1:1 ratio
So In my case (itx) I should buy a 2x16 kit capable to reaching 3800mhz Cas 14 1T.
It doesn't exist and in my little knowledge, dual rank memory is not 2T by design?