It's the next big thing to improve for automatic boosting. We got GPUs and CPUs covered on the AMD eco system, now just getting good results on auto is missing. ^_^
I just used this video as leverage to tune my Patriot Viper DDR3 @ 2100 Mhz. Definitely recommend this channel for newcomers. I learned something new about the uses of timings & performance (regardless of clock speeds) and doing calculations to figure out the max bandwidth.
Verry good vid! Could you make one explaining how to set tight timings? What are the rules for this, what to do on some chipsets, what to not do etc :D
It is possible to calculate the optimal combination for speed and timings for a given memory specification. I have a sheet in my channel that calculates the timings for a specific speed, but to calculate the optimal combination the limits in mathematics can be used. The problem is that I totally forgot how to calculate the limits regarding the lowest speed differential maybe you remember those calculations.
+Grey Goose He means in terms of maximum bandwidth. In terms of maximum performance then yes it'll depend on the application, but for a bandwidth limited application there *is* a optimal combo. And with the APUs that is worth looking at.
@@peppybocan Simple. You just set your RAM for the highest frequency AND lowest timings you possibly can, with the system still booting up and running stable, and tweak the frequency & timings accordingly, so that they will give you the best results in your chosen benchmark(s). Did you not watch the video?
Infrequently works but I think the word you were looking for is seldom. Once again stellar Post. There are a lot of mirrors and smoke out there in consumer land. Thanks for going down the rabbit hole for us and getting real world information.
In a very simple way but complete if possible. Not everybody is that good at english. There are some things that I still haven't understood like can I go under with timings on my modules? There is a way to know how long an udimm could last under a certain setting?
@@temet_nosce You can certainly set lower timings for your RAM modules than what's shown on the sticker. That's If the RAM sticks are capable of it and you know what you're doing of course. UDIMM will last you for a long long time as long as you are not running it at a very high voltage for a long period of time. This voltage threshold is different for the different kinds of chips used by the RAM vendor in your particular memory kit.
Well, I havn't bothered doing it for awhile, but back in the days, you could find the chips timings in ns on the vendors website. So it was all about rounding them up to a clock number, depending on the speed you were running it at. Then, it's all about how much voltage you're willing to run through it. I've lost countless hardware parts to oc, so now I just limit myself to specs
You can check you maximum theoretical memory bandwidth in AIDA even without a calculator. In AIDA go to: Motherboard -> Motherboard -> Memory Bus Properties -> Bandwidth
New to PC, never OCd my RAM (well first build had 16gb Crucial Ballistixs 2666 at 3200 with no timings adjustment on a i5-7600k at 5ghz. Now I have an i7-8700k at 5ghz on an Aorus Z370 G7 with 16gb of Corsair Vengeance RGB 3466(B-die) with pretty sloppy timings as you've now shown me, about to go tinker. Thanks Buildzoid.
So for gaming I know that latency matters more so reducing timings is better than getting higher frequency, but how much frequency is it worth it to give up?
No you idiot speed matters way more then any latency on games you get more more FPS but benchmark look at latency also filling you dims with memory gives you more bandwidth then 2 dims slot Also a 40000 GHZ memory has a true latency of 13.50
@@MILSIM530 calls someone asking a reasonable question an idiot, then proceeds to talk about memory ranks without taking to account the topology of the motherboard, says 40,000Ghz, and spouts latency figures without timings. Who's the actual idiot? Lol
Awesome vid.Since I have a crap binned 2400mhz cl17 micron bdie sticks, and I couldn't get more than 2666 out of em on my ryzen, I pretty much maxed every timing i have access to, putting as low is it can go and remain stable. In geekbench, I actually beat some 2933 and 3200 MHz sticks in copy and bandwidth (my only setback is my latency) , so thats awesome.
Personally I like to use the RightMark Memory Tester (which is a module of the Memory Analyzer). It's really outdated, but still gets the job done and has some customization options for the tests. You have to set test block size. If you fit in L1d you can measure L1 cache, if you fit in L2... and so on. On a side note Write NT test matches AIDA write tests really closely.
If only there were a way to simplify x * 2^8 / 2^3 :P I appreciate explaining it the first time though. How relevant is memory bandwidth for 2400g; where would the sweat spot between frequency for CPU and bandwidth for GPU be, in most scenarios?
+Dominik Liberda right, i forgot there is a difference between cpu frequency and memory controller frequency. Damn SOC's making everything more complicated.
Memory bandwidth is the most important factor for the APUs when it comes to gaming. After that it's the GPU, and lastly the CPU which gets a pretty good job done by Precision Boost 2 so doesn't really need to be touched tbh.
I’m using userbenchmark. I have a Ryzen 3700x. My board is Asus crosshair VIII formula X570. The memory I am using is F4-3600C14Q-32GTZNB is it B-Die. My ram volts is 1.45 My timings the first 6 r, 14 14 14 14 30 48 My tRFC is 344. The ram score I get on user benchmark is 113. I know this ram will do way better I am running the ram at 3800mhz.
I've noticed that sometimes you get better performance on looser timings on GPU benchmarks and games than some CPU benches. Also games seem to be smoother with looser timings, around 3800mhz with the right timings seems to be optimum with Kaby & Coffee Lake when running around 1.55v on the sticks. Ryzen is a bit trickier as it alters the northbridge (mesh) speed with the memory speed so best results vary between 2933mhz & 3200mhz for me depending on the CPU cooling as a higher mesh will make the cpu hotter and affect performance.
just bought a 2x8GB kit of Team Group T-Force Xtreem ARGB C14s to replace my Trident Z Royal Silver 3200C16-18-18-18s for my X570 Taichi like 4 hours ago
Hi there, what does affect the temp of the memory most (exept for voltage) Ofc. i mean when you change the frequency howmutch extra heat does it produce or if you chage some timings what extra does that do.
Throwing a question out Package Wattage, I have two identical system except the mobo. They are I7-6800K's using Asus X99-A II and X99-A\USB3.1 boards. Overclocked to 3,9 GHz all cores VID is 1.171 V with idle watts at 29.85 and the other is 1.205 V with idle watts 26.04. Funny I would think the higher VID would draw more watts or is there is missing piece here.
Arent timings just estimation by the controller on how fast the thing is done? And the actual thing is depndant only on freq and voltage (and silicon)?
how do you get over 100k read on x299 with quadchannel when I only get like 60k max on x99 with 3333 cl 13 13 13 36 cr1 260? x99 has quadchannel too or what is tanking the bandwidth so much?
General max guidelines for cpu voltage on Ryzen. 1.42v Gen 1, 1.4V Gen 2. Vsoc = 1.2V except for APU's = 1.15V. But, you have to be using a quality motherboard with quality components, enough VRM's, and heatsinks to manage the heat. Use a garbage motherboard and\or a case with inadequate cooling at these settings for everyday use and their will be consequences.
Because it doesn't matter Once the OC is stable at a given voltage, the voltage won't affect performance. And to get the same timings and clocks, any two memory kits will likely need different voltage, so where's the point in knowing what he used? You might as well buy some B-Die and dial in the timings he used, then see what voltage you need to sustain them.
Need some info.. there are are settings in asus bios that i cannot find any info on ... trefix9 and oref_ri.. should they be higher or lower for better performance and what are they in the first place? thank you
So theoretically if we have DDR3 2133 and DDR4 2133, DDR3 and DDR4 will have the same RAM Bandwidth if they have the same clock because DDR3 2133*64*2/8=34GB/s and DDR4 2133*64*2/8=34GB/s that is in dual channel, correct me if i am wrong please?
Just try to match the highest MHz with the lowest timings and that is the best you can do. No such thing as perfect settings when it comes to memory timings.
You tweaked the timings of the lower speed kits, so why not see what you could get out of the 4000 kit? Maybe I missed that part or is the video missing that part?
So ideally the best one might get for memory bandwidth on a 2400G would be ~64GB/s at 4000MHz? (has anyone gotten memory higher? ) If they make a X399 APU then they could get some pretty nice bandwidth.
That's pretty informative for me. Also, I would 100% watch if you made a video on trying to get the best bandwidth like you mention in this video. Off topic question: I do video editing, gaming, and transcoding primarily on my 7700k w/ 16GB of RAM. It's currently @ 3000 MHz going 14-15-15-33. I was thinking about bumping up the frequency to 3200 MHz to see if it's stable but at this point, am I really going to benefit from it or is it just a nice number to achieve?
Hey interesting video! But wait a minute...lets say I get a RRD4 3600 CL15-(...) that comes out-of-the-box like this and use a good oc motherboard, and set that ram exactly as its max stated freq and timings are by XMS, is it possible to have a stable system with an i5 or i7 without any bottle-neck issue or worst incompatibility without having to overclock the cpu?? Stock cpu freq and ram at 3600... Thankx!
Could you please tell me how to calculate the bandwidth of let's say a 2 Dual Channel Kit in a system? Is it going to be the same 256 or 128 separately? I commented on your Buying "DDR 4 Basic video" regarding my 2 Dual Channel HyperX Fury 2133Mhz that I kinda got sped up to 2933Mhz with 14-16-16-34 timing on 1.36V. The stock 2133Mhz timing is 14-14-14-35. Once again thank you for sharing the knowledge.
Awesome video as always, I have a question/looking for advice... not really relevant to this video however so I apologise for that. I have started building a custom cooling solution and was wondering is there any negative side effect from running RAM at sub zero or close to zero (Celsius)? I understand SSD's lifespan is degraded when running too cold... does RAM suffer in the same way? Endless Google searches just bring up LN2 CPU overclocking info. Thanks in advance!!! :D
I can't help with first hand experience, but RAM should not give a dimm about low temperature, same like CPU. But condensing water could be problematic.
Thanks, i have only tried briefly overclocking my new ram, but i hope to see more videos from you helping me figure out how to get the most out of my ram :D Just to be clear is it save to put 1.5 volt into my ddr4?
thanks, it Samsung B, i am trying 1.5 volt at 3200mhz with 15 17 17 35 timing, but i cannot boot with lower timing and if i go higher in clockspeed then games start crashing randomly.
If you're on Ryzen try www.overclock.net/forum/11-amd-motherboards/1624603-rog-crosshair-vi-overclocking-thread-2067.html#post26178558 or www.overclock.net/t/1640919/ryzen-dram-calculator-overclocking-dram and run HCI Memtest to 500% to be sure about stability. If either don't work, move to the latest BIOS and try again. Memory OC is a pain, hopefully these help a little bit. Good luck haha.
I just got some G Skill Trident Z 15 15 15 35 3000MHz @1.35 on a z170 extreme4 /6600k. What timings and frequency should i try while keeping at stock 1.35 v? Would you do a dedicated RAM OC guide? I'm sure with you skill and knowledge it would be the best on the net!
Je5tercore Intel has listed the max XMP2 profile voltage as 1.5. I've seen lots with 1.45v, none with 1.5v, but that's where I'm at right now. So far so good.
so Kingston HyperX Savage 16GB (2 X 8GB) DDR4 2400MHz CL12 same as Team T-Force Night Hawk White RGB 16GB (2 X 8GB) DDR4 3200 CL16 1.35V Desktop Memory (TF1D416G3200HC16CDC01)
Is this just for bragging rights, or does it translate to, say, faster rendering times in Premiere? By how much, over oob 2400 kit? (Not trolling, just asking as the real world benefit is not mentioned in the video).
What are your recommendations for setting up 3466 MHz RAM. Currently at CL18 on the asus rog x370-i and the 1800x. I only know like the first 5 timings lol.
You didn't talk about voltage which you have used for those settings...your 3.2 and 3.6GHz looked like it would need at least 1.5V where your 4GHz looked like it can work on 1.35V no problem. I mean if you set 1.4 or 1.45V as your absolute maximum for 24/7 then 4GHz with adjusted timings would be the most effective giving the best results in my opinion.
Memory is certainly not in my wheelhouse of expertise so - if tREFI basically just unilaterally reduces Memory performance period, what does it even do? why does it exist, does it need to exist? it sounds like a scan to remove data that's no longer needed to be stored, i think?
Nice Video:). So to get it right....my 2x16GB 3200 CL14 KIt with optimised sub timings is doing 50000 read in Aida.....therefore I'm close to the maximum Bandwith my dual channel Ryzen 7 can do?
BhaalTheFleischgolem Nice, I've got similiar results with my 3600MHz CL16 kit (it costed me same as 3200MHz CL14) clocked at 3200MHz 14-13-13-27, what is your board and your primaries, secondaries? Sadly, I can't run this profile, cause I updated my BIOS and my mb now freaks out when I try to set RAM properly, so settings that were 100% stable are now no where near to that. And I can't flash older BIOS, cause of "security", stupid MSI.
I use a Crosshair VI Hero with Trident Z. It was quiet a hassle to get it to run on an older Bios like 1403. With CPU@3.8GHz, BCLK@100.2MHz Ineeded: DRam@1.355V, soc@1.05V, ProcODT@60 Ohm, Geardown enabled, Power down disabled, Bankgroupswap enabled(only for dual ranked modules), RttNom disabled, RttWr RZQ/3, RttPark RZQ/1 and the four Cad Bus Strenght values @ 20 20 40 60 Ohm. With less frequenzy on Ram or CPU, they need to be in the direction of 20 20 30 30. But with a new Bios this is not nessesary anymore! Now the Auto settings do work just fine and you only need to set Dram voltage and soc voltage and it is just as stable and fast. But I'm not sure what values are used with auto, so I force my old settings. Latencies are: DRAM CAS# Latency [14] DRAM RAS# to CAS# Read Delay [13] DRAM RAS# to CAS# Write Delay [13] DRAM RAS# PRE Time [13] DRAM RAS# ACT Time [26] Trc_SM [42] TrrdS_SM [5] TrrdL_SM [7] Tfaw_SM [29] TwtrS_SM [4] TwtrL_SM [8] Twr_SM [10] Trcpage_SM [Auto] TrdrdScl_SM [2] TwrwrScl_SM [2] Trfc_SM [304] Trfc2_SM [203] Trfc4_SM [145] Tcwl_SM [14] Trtp_SM [8] Trdwr_SM [6] Twrrd_SM [1] TwrwrSc_SM [1] TwrwrSd_SM [6] TwrwrDd_SM [6] TrdrdSc_SM [1] TrdrdSd_SM [4] TrdrdDd_SM [4] It may be possible to get better results with more voltage, but it is my daily system so I want it to last;)
nice:) If DRam would be cheaper again, I would go for high voltages and try some more frequency. Maybe not for daily use, but just to try out how much performance there is to gain in games.
yes? there is slight hiss but pretty sure that's common in his videos... if you are using a dynamics compressor enhancement, then it would make the loud parts quieter & the quiet parts louder, making the hiss louder, ruining the sound, so never use anything other than an EQ (& when EQ'ing never raise, only lower)
If XMP is trash why is it there? I have 3200MHZ CL14-14-14-34 and 4000MHz CL19-19-19-39. 3200/14=228 and 4000/19=210. So in reality the 3200MHz RAM is faster. Price to performance makes no sense than.
Sure, blame the "stupid calculator"! Hey, good stuff. I see that out of the billions of people on the planet, only 23K even have a clue of what you are talking about! Keep up the great work!
It's also hypnotic to listen to the smart people talk, even if you have no idea what he's saying. A while back I got quite drunk and about to fall asleep, but I managed to put up a Buildzoid video. Too pissed to follow what he was saying, dozing off... I haven't slept that well since.
Learning CPU 5 minutes learning GPU 5 minutes learning Ram lifetime
It's the next big thing to improve for automatic boosting. We got GPUs and CPUs covered on the AMD eco system, now just getting good results on auto is missing. ^_^
@@Traumglanz A good first stage is a FOSS BIOS that works with your MB.
I just used this video as leverage to tune my Patriot Viper DDR3 @ 2100 Mhz. Definitely recommend this channel for newcomers. I learned something new about the uses of timings & performance (regardless of clock speeds) and doing calculations to figure out the max bandwidth.
Verry good vid! Could you make one explaining how to set tight timings? What are the rules for this, what to do on some chipsets, what to not do etc :D
It is possible to calculate the optimal combination for speed and timings for a given memory specification. I have a sheet in my channel that calculates the timings for a specific speed, but to calculate the optimal combination the limits in mathematics can be used. The problem is that I totally forgot how to calculate the limits regarding the lowest speed differential maybe you remember those calculations.
David Martins What do you mean with "optimal"? Doesn't that also depend on the application?
+Grey Goose
He means in terms of maximum bandwidth.
In terms of maximum performance then yes it'll depend on the application, but for a bandwidth limited application there *is* a optimal combo.
And with the APUs that is worth looking at.
Learnt this during the Athlon 64/X2 OC'ing days. DFI Lan Party board for days. When world records were set on ram timings.
More elaboration on the relationship of timings to frequency for optimal bandwidth. Thanks for the detailed explanations.
first
how do you set the RAM timings properly?
As a complete noob to ram, these videos are really helpful
yo man big fan btw. please post these screenshots to your blog if you have time :)
Love your channel dude. Now, if I could just get you make out a parts list for my first build....😁
@@peppybocan Simple. You just set your RAM for the highest frequency AND lowest timings you possibly can, with the system still booting up and running stable, and tweak the frequency & timings accordingly, so that they will give you the best results in your chosen benchmark(s). Did you not watch the video?
Infrequently works but I think the word you were looking for is seldom.
Once again stellar Post. There are a lot of mirrors and smoke out there in consumer land. Thanks for going down the rabbit hole for us and getting real world information.
nice video, thanks and thumbs up.
i would love to see your attempt of hitting max bandwidth, especally to see, wich timings are responsible for this
@13:56
opposite of often => rarely..?
lmao
or seldom
Infrequently did the job as well, although it took a few tries to figure it out! :D lol
Can you do a video explaining what all these timings are?
In a very simple way but complete if possible.
Not everybody is that good at english.
There are some things that I still haven't understood like can I go under with timings on my modules?
There is a way to know how long an udimm could last under a certain setting?
@@temet_nosce You can certainly set lower timings for your RAM modules than what's shown on the sticker. That's If the RAM sticks are capable of it and you know what you're doing of course.
UDIMM will last you for a long long time as long as you are not running it at a very high voltage for a long period of time. This voltage threshold is different for the different kinds of chips used by the RAM vendor in your particular memory kit.
Well, I havn't bothered doing it for awhile, but back in the days, you could find the chips timings in ns on the vendors website.
So it was all about rounding them up to a clock number, depending on the speed you were running it at.
Then, it's all about how much voltage you're willing to run through it.
I've lost countless hardware parts to oc, so now I just limit myself to specs
The sheet will now give you the memory sweet spots in a graphical view.
You can check you maximum theoretical memory bandwidth in AIDA even without a calculator.
In AIDA go to: Motherboard -> Motherboard -> Memory Bus Properties -> Bandwidth
New to PC, never OCd my RAM (well first build had 16gb Crucial Ballistixs 2666 at 3200 with no timings adjustment on a i5-7600k at 5ghz. Now I have an i7-8700k at 5ghz on an Aorus Z370 G7 with 16gb of Corsair Vengeance RGB 3466(B-die) with pretty sloppy timings as you've now shown me, about to go tinker. Thanks Buildzoid.
Great vid and good idea the bandwith test, but could you do it with dual channel and like more mainstream cpu
So for gaming I know that latency matters more so reducing timings is better than getting higher frequency, but how much frequency is it worth it to give up?
No you idiot speed matters way more then any latency on games you get more more FPS but benchmark look at latency also filling you dims with memory gives you more bandwidth then 2 dims slot
Also a 40000 GHZ memory has a true latency of 13.50
@@MILSIM530 40 thousand Gigaherz you say???
@@MILSIM530 calls someone asking a reasonable question an idiot, then proceeds to talk about memory ranks without taking to account the topology of the motherboard, says 40,000Ghz, and spouts latency figures without timings. Who's the actual idiot? Lol
Awesome vid.Since I have a crap binned 2400mhz cl17 micron bdie sticks, and I couldn't get more than 2666 out of em on my ryzen, I pretty much maxed every timing i have access to, putting as low is it can go and remain stable. In geekbench, I actually beat some 2933 and 3200 MHz sticks in copy and bandwidth (my only setback is my latency) , so thats awesome.
Anyone know a free tool for memory benching? (AIDA expires)
Personally I like to use the RightMark Memory Tester (which is a module of the Memory Analyzer). It's really outdated, but still gets the job done and has some customization options for the tests. You have to set test block size. If you fit in L1d you can measure L1 cache, if you fit in L2... and so on. On a side note Write NT test matches AIDA write tests really closely.
Free as in freedom, too, nice! Thanks man.
Search for AIDA64 key on TH-cam and click on the first video. You may have to also look through the comments, though.
Thanks but stealing isn't my thing.
Sisoftware Sandra lite got some memory benchmarks, and it´s free. dont know if it is what you are looking for
Mmmmm! X299 DARK!
If only there were a way to simplify x * 2^8 / 2^3 :P I appreciate explaining it the first time though. How relevant is memory bandwidth for 2400g; where would the sweat spot between frequency for CPU and bandwidth for GPU be, in most scenarios?
Mihai Bica CPU freq and memory bandwidth are completly different things.
+Dominik Liberda right, i forgot there is a difference between cpu frequency and memory controller frequency. Damn SOC's making everything more complicated.
Memory bandwidth is the most important factor for the APUs when it comes to gaming. After that it's the GPU, and lastly the CPU which gets a pretty good job done by Precision Boost 2 so doesn't really need to be touched tbh.
how to calculate the theoretical bandwidth including the MHz and Clocks? not using these softwares?
I’m using userbenchmark. I have a Ryzen 3700x. My board is Asus crosshair VIII formula X570. The memory I am using is F4-3600C14Q-32GTZNB is it B-Die. My ram volts is 1.45
My timings the first 6 r, 14 14 14 14 30 48
My tRFC is 344. The ram score I get on user benchmark is 113. I know this ram will do way better I am running the ram at 3800mhz.
I've noticed that sometimes you get better performance on looser timings on GPU benchmarks and games than some CPU benches. Also games seem to be smoother with looser timings, around 3800mhz with the right timings seems to be optimum with Kaby & Coffee Lake when running around 1.55v on the sticks.
Ryzen is a bit trickier as it alters the northbridge (mesh) speed with the memory speed so best results vary between 2933mhz & 3200mhz for me depending on the CPU cooling as a higher mesh will make the cpu hotter and affect performance.
just bought a 2x8GB kit of Team Group T-Force Xtreem ARGB C14s to replace my Trident Z Royal Silver 3200C16-18-18-18s for my X570 Taichi like 4 hours ago
Hi there, what does affect the temp of the memory most (exept for voltage)
Ofc. i mean when you change the frequency howmutch extra heat does it produce or if you chage some timings what extra does that do.
Ram doesn't get hot unless you run the kinda voltages buildzoid does, 1.8-2.0v
Throwing a question out Package Wattage, I have two identical system except the mobo. They are I7-6800K's using Asus X99-A II and X99-A\USB3.1 boards. Overclocked to 3,9 GHz all cores VID is 1.171 V with idle watts at 29.85 and the other is 1.205 V with idle watts 26.04. Funny I would think the higher VID would draw more watts or is there is missing piece here.
Great video!
Arent timings just estimation by the controller on how fast the thing is done? And the actual thing is depndant only on freq and voltage (and silicon)?
how do you get over 100k read on x299 with quadchannel when I only get like 60k max on x99 with 3333 cl 13 13 13 36 cr1 260? x99 has quadchannel too or what is tanking the bandwidth so much?
Thank you
General max guidelines for cpu voltage on Ryzen. 1.42v Gen 1, 1.4V Gen 2. Vsoc = 1.2V except for APU's = 1.15V. But, you have to be using a quality motherboard with quality components, enough VRM's, and heatsinks to manage the heat. Use a garbage motherboard and\or a case with inadequate cooling at these settings for everyday use and their will be consequences.
Why does he keeps the voltage secret on all these overclocks?
Will it hurt him telling us what it is?
He doesn't, its always his XTREEM DDR4 8gb DC Kit at ~1.8-2v
Because it doesn't matter
Once the OC is stable at a given voltage, the voltage won't affect performance.
And to get the same timings and clocks, any two memory kits will likely need different voltage, so where's the point in knowing what he used?
You might as well buy some B-Die and dial in the timings he used, then see what voltage you need to sustain them.
I noticed the command rate for the memory wasn't consistent across tests, it changes between 1T and 2T. How does this setting affect the results?
Need some info.. there are are settings in asus bios that i cannot find any info on ... trefix9 and oref_ri.. should they be higher or lower for better performance and what are they in the first place? thank you
So theoretically if we have DDR3 2133 and DDR4 2133, DDR3 and DDR4 will have the same RAM Bandwidth if they have the same clock because DDR3 2133*64*2/8=34GB/s and DDR4 2133*64*2/8=34GB/s that is in dual channel, correct me if i am wrong please?
where can i find guides for perfect settings?
there is no such thing as perfect
Just try to match the highest MHz with the lowest timings and that is the best you can do. No such thing as perfect settings when it comes to memory timings.
@@kn00tcn +1
You tweaked the timings of the lower speed kits, so why not see what you could get out of the 4000 kit? Maybe I missed that part or is the video missing that part?
So whats better than. 3200 @cl16, or 2933@cl14?
now I'm interested in seeing what you can do with a 2200g/2400g
owen barrett would love to see that too. Want to push my 2200g for stable 24/7 mitx build.
So ideally the best one might get for memory bandwidth on a 2400G would be ~64GB/s at 4000MHz? (has anyone gotten memory higher? ) If they make a X399 APU then they could get some pretty nice bandwidth.
That's pretty informative for me. Also, I would 100% watch if you made a video on trying to get the best bandwidth like you mention in this video.
Off topic question: I do video editing, gaming, and transcoding primarily on my 7700k w/ 16GB of RAM. It's currently @ 3000 MHz going 14-15-15-33. I was thinking about bumping up the frequency to 3200 MHz to see if it's stable but at this point, am I really going to benefit from it or is it just a nice number to achieve?
have you ever seen any encoding benchmark change much from such a minor ram difference?
Hey interesting video!
But wait a minute...lets say I get a RRD4 3600 CL15-(...) that comes out-of-the-box like this and use a good oc motherboard, and set that ram exactly as its max stated freq and timings are by XMS, is it possible to have a stable system with an i5 or i7 without any bottle-neck issue or worst incompatibility without having to overclock the cpu?? Stock cpu freq and ram at 3600... Thankx!
What memory kit is overall better?
3200MHz CL 14-14-14-34
or
3600MHz CL 16-16-16-36
Jonas Sjöman 800mhz 3-3-3-11 :)
3600
whats the maximum voltage on hynix AFR chips for 24/7 use without degrading them?
1.5V
+Actually Hardcore Overclocking thank you :D btw, your videos are awesome and very helpful. Keep up the good work.
Could you please tell me how to calculate the bandwidth of let's say a 2 Dual Channel Kit in a system? Is it going to be the same 256 or 128 separately? I commented on your Buying "DDR 4 Basic video" regarding my 2 Dual Channel HyperX Fury 2133Mhz that I kinda got sped up to 2933Mhz with 14-16-16-34 timing on 1.36V. The stock 2133Mhz timing is 14-14-14-35. Once again thank you for sharing the knowledge.
Do you have any guides? Would love to maximize my z370 and x99 systems
Does memory speed affect usage?
It can matter...Depend on what you do.
Can you do a video on how to stabilise 7700 k past 5.2 gigahertz
You know its purely luck of the draw right
liquid nitrogen
have to be very fucking lucky with the silicon lottery, i lost that one, cant get past 4.7 on air, no matter what.. i got a fucking DUD!
Wow 4.7 on a 7700K is absolute shit sorry for your loss...lol
feelsbadman.jpeg
What was the latency for all the speeds that you tested?
If i'm running 2x8GB DDR4 Sticks, setup in Dual Channel. Do I use 256 or 128?
so google dailiable settings? Guessing you mean dont run it at the max stable oc that you can manage if you going 24/7
Awesome video as always, I have a question/looking for advice... not really relevant to this video however so I apologise for that. I have started building a custom cooling solution and was wondering is there any negative side effect from running RAM at sub zero or close to zero (Celsius)? I understand SSD's lifespan is degraded when running too cold... does RAM suffer in the same way? Endless Google searches just bring up LN2 CPU overclocking info. Thanks in advance!!! :D
I can't help with first hand experience, but RAM should not give a dimm about low temperature, same like CPU. But condensing water could be problematic.
Thanks, i have only tried briefly overclocking my new ram, but i hope to see more videos from you helping me figure out how to get the most out of my ram :D Just to be clear is it save to put 1.5 volt into my ddr4?
If it's Samsung then yes, but if it's Hynix then no that'll only handle up to 1.45V.
If you're unsure, stick to 1.45V.
thanks, it Samsung B, i am trying 1.5 volt at 3200mhz with 15 17 17 35 timing, but i cannot boot with lower timing and if i go higher in clockspeed then games start crashing randomly.
If you're on Ryzen try www.overclock.net/forum/11-amd-motherboards/1624603-rog-crosshair-vi-overclocking-thread-2067.html#post26178558 or www.overclock.net/t/1640919/ryzen-dram-calculator-overclocking-dram and run HCI Memtest to 500% to be sure about stability. If either don't work, move to the latest BIOS and try again.
Memory OC is a pain, hopefully these help a little bit. Good luck haha.
no, i am on an asus z370 f motherboard with an intel i5 8600k paired with crucial ballistix elite 3466mhz ddr4 ram.
Ah. Welp maybe give them a shot anyway. Memory's worse on Ryzen so it might still work on Intel, you never know.
Could you do the same for Ryzen? The platform seems to be more sensible about RAM speed and latency.
What timings you messed with? The only 5 timings as seen on CPU-Z the we can see, or did you go down further into a rabbit hole? :)
I just got some G Skill Trident Z 15 15 15 35 3000MHz @1.35 on a z170 extreme4 /6600k. What timings and frequency should i try while keeping at stock 1.35 v?
Would you do a dedicated RAM OC guide? I'm sure with you skill and knowledge it would be the best on the net!
tweaks are trial & error, you will have to find out yourself by seeing what is stable
What does single/dual rank RAM mean?
What's a max safe voltage for daily use for DDR4?
I'm gonna start contributing to your channel through Patreon, I've learned a lot from you.
Je5tercore Intel has listed the max XMP2 profile voltage as 1.5. I've seen lots with 1.45v, none with 1.5v, but that's where I'm at right now. So far so good.
whats the highest you've seen on x399? Would love to know how the infinity fabric latency impacts memory bandwidth
I don't have X399
Actually Hardcore Overclocking that would make testing it rather difficult lol
so Kingston HyperX Savage 16GB (2 X 8GB) DDR4 2400MHz CL12 same as Team T-Force Night Hawk White RGB 16GB (2 X 8GB) DDR4 3200 CL16 1.35V Desktop Memory (TF1D416G3200HC16CDC01)
Is this just for bragging rights, or does it translate to, say, faster rendering times in Premiere? By how much, over oob 2400 kit?
(Not trolling, just asking as the real world benefit is not mentioned in the video).
Where can I buy that ‘[ TRIAL VERSION ]’ motherboard you have? I have never heard of that brand but it looks interesting....
What are your recommendations for setting up 3466 MHz RAM. Currently at CL18 on the asus rog x370-i and the 1800x. I only know like the first 5 timings lol.
Use this to find out www.overclock.net/forum/13-amd-general/1640919-ryzen-dram-calculator-1-1-0-beta-1-overclocking-dram-am4.html
You didn't talk about voltage which you have used for those settings...your 3.2 and 3.6GHz looked like it would need at least 1.5V where your 4GHz looked like it can work on 1.35V no problem. I mean if you set 1.4 or 1.45V as your absolute maximum for 24/7 then 4GHz with adjusted timings would be the most effective giving the best results in my opinion.
How much you need to lower a 2666 or increase the 4000 in latencies for the 2666MHz to win? OwO
2 years ago, I bought Corsair 16GBx4 Platinum memory at XMP 15-17-17-35 3000Mhz. Can't overclock the ram over 3100Mhz. Timing is no good either.
you didnt mention voltage or which memory controller is being used
kn00tcn
Doesn't matter, a single 16GB is hard.
Any specific RAM kits with B-die that you would recommend?
G.SKILL 3600MHz CL14 or very expensive 4000MHz CL14.
Memory is certainly not in my wheelhouse of expertise so - if tREFI basically just unilaterally reduces Memory performance period, what does it even do? why does it exist, does it need to exist?
it sounds like a scan to remove data that's no longer needed to be stored, i think?
greaat vid
Nice Video:).
So to get it right....my 2x16GB 3200 CL14 KIt with optimised sub timings is doing 50000 read in Aida.....therefore I'm close to the maximum Bandwith my dual channel Ryzen 7 can do?
BhaalTheFleischgolem Nice, I've got similiar results with my 3600MHz CL16 kit (it costed me same as 3200MHz CL14) clocked at 3200MHz 14-13-13-27, what is your board and your primaries, secondaries?
Sadly, I can't run this profile, cause I updated my BIOS and my mb now freaks out when I try to set RAM properly, so settings that were 100% stable are now no where near to that. And I can't flash older BIOS, cause of "security", stupid MSI.
I use a Crosshair VI Hero with Trident Z. It was quiet a hassle to get it to run on an older Bios like 1403.
With CPU@3.8GHz, BCLK@100.2MHz Ineeded:
DRam@1.355V, soc@1.05V, ProcODT@60 Ohm, Geardown enabled, Power down disabled, Bankgroupswap enabled(only for dual ranked modules), RttNom disabled, RttWr RZQ/3, RttPark RZQ/1 and the four Cad Bus Strenght values @ 20 20 40 60 Ohm. With less frequenzy on Ram or CPU, they need to be in the direction of 20 20 30 30.
But with a new Bios this is not nessesary anymore! Now the Auto settings do work just fine and you only need to set Dram voltage and soc voltage and it is just as stable and fast. But I'm not sure what values are used with auto, so I force my old settings.
Latencies are:
DRAM CAS# Latency [14]
DRAM RAS# to CAS# Read Delay [13]
DRAM RAS# to CAS# Write Delay [13]
DRAM RAS# PRE Time [13]
DRAM RAS# ACT Time [26]
Trc_SM [42]
TrrdS_SM [5]
TrrdL_SM [7]
Tfaw_SM [29]
TwtrS_SM [4]
TwtrL_SM [8]
Twr_SM [10]
Trcpage_SM [Auto]
TrdrdScl_SM [2]
TwrwrScl_SM [2]
Trfc_SM [304]
Trfc2_SM [203]
Trfc4_SM [145]
Tcwl_SM [14]
Trtp_SM [8]
Trdwr_SM [6]
Twrrd_SM [1]
TwrwrSc_SM [1]
TwrwrSd_SM [6]
TwrwrDd_SM [6]
TrdrdSc_SM [1]
TrdrdSd_SM [4]
TrdrdDd_SM [4]
It may be possible to get better results with more voltage, but it is my daily system so I want it to last;)
your max theoretical bandwidth is 51200,so yea, you're really close:
3200 * 8 * 2 =51200
frequency * bytes per chanel(always 8 on ddr4) * channels
My RAM timings with my shitty MSI board, getting something like 50500 read Aida:
Freq: 3200MHz
Command Rate: 1T
tCL: 14
tRCDRD: 13
tRCDWR: 13
tRP: 13
tRAS: 27
tRC: 42
tRCPAGE: 0
tRFC: 260
tWR: 10
tWTR_S: 4
tWTR_L: 12
tRRD_S: 6
tRRD_L: 9
tRTP: 8
tFAW: 28
tCWL: 14
tCKE: 1
tRDRDSC_L: 3
tRDRDSC: 1
tRDRDSD: 5
tRDRDDD: 5
tWRWRSC_L: 3
tWRWRSC: 1
tWRWRSD: 7
tWRWRDD: 7
tRDWR: 7
tWRRD: 2
nice:)
If DRam would be cheaper again, I would go for high voltages and try some more frequency. Maybe not for daily use, but just to try out how much performance there is to gain in games.
am i the only one hearing the distortion/fuzz in the audio?
yes? there is slight hiss but pretty sure that's common in his videos...
if you are using a dynamics compressor enhancement, then it would make the loud parts quieter & the quiet parts louder, making the hiss louder, ruining the sound, so never use anything other than an EQ (& when EQ'ing never raise, only lower)
No paypal. i've found only patreon.. but i like paypal just for a tantum.
MY DUDE! when is the next stream?!?!?!
Got 59gb out of 3600 max theoretical 57600 lol
I'll bet 90% of these settings shit out alot of errors in memtest86+.
It’s Intel duh.
If XMP is trash why is it there? I have 3200MHZ CL14-14-14-34 and 4000MHz CL19-19-19-39. 3200/14=228 and 4000/19=210. So in reality the 3200MHz RAM is faster. Price to performance makes no sense than.
Sure, blame the "stupid calculator"! Hey, good stuff. I see that out of the billions of people on the planet, only 23K even have a clue of what you are talking about! Keep up the great work!
It's also hypnotic to listen to the smart people talk, even if you have no idea what he's saying.
A while back I got quite drunk and about to fall asleep, but I managed to put up a Buildzoid video. Too pissed to follow what he was saying, dozing off... I haven't slept that well since.
tRFC -> HIGHER = BETTER iimns....
u never answer questions do you
Go to his twitch and ask.
Aida key UUFB1-4RYD6-9NDNG-7DM94-UBIKV
Really really thanks mate! :)
I might be interested to watch more of your content if you didn't have a "I don't want subscribers, I'm doing this for fun" attitude.
Second :(
rambling 17 minute video... could have been much shorter if you did any editing or even just organized your thoughts before recording