First time oc samsung b die

[Deleted member 185158]

That’s pretty much how everyone does it

I guess I do live in a different world!!!

My world (home)

Home | Warp9-systems

Visit our forum at: warp9-systems.proboards.com

warp9-systems.proboards.com

There's not a single person from the above link that would suggest running b-die at cl20.

Haha. I gotta get outta here...

 

[ThrashZone]

What do u mean it went opposite ?
cl16 xmp was 65ns in aida and now it’s 52ns

Hi,
Although those are improvements dual channel ddr4 is usually mid 30ns not mid 40 or 60 for that matter guess that where you're trying to go
45ns you're finally in quad channel x299 land but then again all the ddr5 memory numbers I've seen, latency wise are shit to me so guess ddr4 on this platform is just as bad so carry on the fight

 

[ir_cow]

Cl 20 at 3600mhz? Your world is way crazy lol.

Not to hijack this thread, but we are troubleshooting here. Need to get back to the basics. Besides technically the memory being binned for only 3600 and the CPU only supports up to 3600 as well, we all know memory can clock much higher.

While typing in low values someone says to do is much quicker, when it doesn't work (as in now), that is when you need to go back to the slow method. First you need to eliminate variables. Can the MB in question actually run this, can the CPU IMC do it as well, can the memory run the desired freq? Unfortunately most people don't have a bunch of memory kits with various bins, CPUs or motherboards to test. I am lucky to have all of that. The best way I have found for memory overclocking when you have nothing else to test against is to eliminate one variable at a time.

For Intel DDR4 memory overclocking (11th, 12th 13th Gen CPU) First make sure XMP values work. Next loosen the timings, change command rate to 2T and use Gear 2 and set to the highest DRAM voltage YOU are willing to use 24/7. leave CPU at auto voltages. Test again, should be fine (sometimes its not though). Now raise the freq one step at a time until it no longer boots or reaches the desired freq. If it doesn't reach the desired freq, you will need to figure out if its a voltage issue, motherboard limitation, IMC limitation or just the memory itself (after all this is overclocking...). This is why lose timings helps, far less likely to be a memory limitation. If you do manage to get to where you want to be. First change the Gear Ratio back to 1, passes test, now lower the Command Rate to 1T. Passes tests, now lower the primary timings. Start with just CAS. Repeat until you reach the bottom. If you get stuck at Gear Ratio, or Command Rate, play with the CPU voltages. If your having problems with timings, more voltage is probably necessary or that is JUST the limit of that particular memory kit / ICs. I've had Samsung B-die that couldn't even do 4000 CL19. They exist....

For motherboard reviews since I know what the CPU can do and have a binned memory kit (say DDR4-46000 32GB), I can start at the top and work down.
For memory reviews, it is a little more tricky, but once I get a motherboard and CPU that can do what I want, overclocking is the same process, but I can skip the motherboard limits, CPU IMC / voltage portion and just focus on the memory voltage and timings.

When you have only one of everything, there is no other way to narrow down a memory overclocking problem. Typing is voltages and timings, could speed things up, but often it just end up putting you in circles not knowing what the real problem is.

This is the best help I can provide, I have nothing more to add.

 

[Redwoodz]

I guess I do live in a different world!!!

My world (home)

Home | Warp9-systems

Visit our forum at: warp9-systems.proboards.com

warp9-systems.proboards.com

There's not a single person from the above link that would suggest running b-die at cl20.

Haha. I gotta get outta here...

Young ones don't know about overclocking before there were any guides.

 

[Johnthedon1]

Not to hijack this thread, but we are troubleshooting here. Need to get back to the basics. Besides technically the memory being binned for only 3600 and the CPU only supports up to 3600 as well, we all know memory can clock much higher.

While typing in low values someone says to do is much quicker, when it doesn't work (as in now), that is when you need to go back to the slow method. First you need to eliminate variables. Can the MB in question actually run this, can the CPU IMC do it as well, can the memory run the desired freq? Unfortunately most people don't have a bunch of memory kits with various bins, CPUs or motherboards to test. I am lucky to have all of that. The best way I have found for memory overclocking when you have nothing else to test against is to eliminate one variable at a time.

For Intel DDR4 memory overclocking (11th, 12th 13th Gen CPU) First make sure XMP values work. Next loosen the timings, change command rate to 2T and use Gear 2 and set to the highest DRAM voltage YOU are willing to use 24/7. leave CPU at auto voltages. Test again, should be fine (sometimes its not though). Now raise the freq one step at a time until it no longer boots or reaches the desired freq. If it doesn't reach the desired freq, you will need to figure out if its a voltage issue, motherboard limitation, IMC limitation or just the memory itself (after all this is overclocking...). This is why lose timings helps, far less likely to be a memory limitation. If you do manage to get to where you want to be. First change the Gear Ratio back to 1, passes test, now lower the Command Rate to 1T. Passes tests, now lower the primary timings. Start with just CAS. Repeat until you reach the bottom. If you get stuck at Gear Ratio, or Command Rate, play with the CPU voltages. If your having problems with timings, more voltage is probably necessary or that is JUST the limit of that particular memory kit / ICs. I've had Samsung B-die that couldn't even do 4000 CL19. They exist....

For motherboard reviews since I know what the CPU can do and have a binned memory kit (say DDR4-46000 32GB), I can start at the top and work down.
For memory reviews, it is a little more tricky, but once I get a motherboard and CPU that can do what I want, overclocking is the same process, but I can skip the motherboard limits, CPU IMC / voltage portion and just focus on the memory voltage and timings.

When you have only one of everything, there is no other way to narrow down a memory overclocking problem. Typing is voltages and timings, could speed things up, but often it just end up putting you in circles not knowing what the real problem is.

This is the best help I can provide, I have nothing more to add.

What test should I be using to validate improvemts ? Aida ? Pyprime ? Imlc ?

 

[thesmokingman]

These timings from the SS are kind of whacky. What's the brand/model density?

 

[ir_cow]

What test should I be using to validate improvemts ? Aida ? Pyprime ? Imlc ?

I do the basics to save time until I get near the end. y-cruncher 2.5b and either MemtestPro or Karhu Software for 5 minutes. Both cost money so a alterative is TestMem5. I used TM5 a few times and it seems to do the job. The hardest on the IMC is y-cruncher and Karhu Software I think. You will find out really quickly if it doesn't work, either a BSOD, a system freeze. Sometimes you will get a the normal error if its on the verge of stability. Once I am nearing my goal, I will run the memory test longer, until the final push of 1 hr.

 

[Johnthedon1]

These timings from the SS are kind of whacky. What's the brand/model density?

What’s wacky about them ?
It’s t force dark pro 8 pack edition
32gb dual rank

 

[ThrashZone]

Hi,
Think he's asking exact set
3600c16 2x16gb
3600c16 2x32gb
..

 

[Johnthedon1]

I do the basics to save time until I get near the end. y-cruncher 2.5b and either MemtestPro or Karhu Software for 5 minutes. Both cost money so a alterative is TestMem5. I used TM5 a few times and it seems to do the job. The hardest on the IMC is y-cruncher and Karhu Software I think. You will find out really quickly if it doesn't work, either a BSOD, a system freeze. Sometimes you will get a the normal error if its on the verge of stability. Once I am nearing my goal, I will run the memory test longer, until the final push of 1 hr.

I passed 8 hours of memtest5 anta777 extream so I’m pretty confidant it’s stable

i ment how do I validate performance gains ? I know I’m aida my bandwidth went up by 8gbs and my latency went down by 14ns compared to the xmp

but is there something better to see performance gains ?

 

[ir_cow]

I literally went through this same process with DDR5 last weekend trying to get DDR5-8000 to work. Same situation, I didn't know if the memory, motherboard or CPU IMC could do it. I ended up at DDR5-7800 until I was sent a MB, CPU and Ram binned for DDR5-8000. Turns out, it was just the memory kits I had that couldn't quite make the cut. Well now one of my original kits can once I copied the XMP profile and raised the voltage more (beyond what I wanted).

In total it was about 20 hours of troubleshooting because I had nothing to compare it to. Now its really quick because I have a reference starting point.

i ment how do I validate performance gains ? I know I’m aida my bandwidth went up by 8gbs and my latency went down by 14ns compared to the xmp

Real world gains would be seen in the games you play. higher frame rate. I believe AID64 shows the peak read and write and avg latency. Besides synthetic benchmarks, overclocking memory does nothing outside of gaming.

 

[Johnthedon1]

I literally went through this same process with DDR5 last weekend trying to get DDR5-8000 to work. Same situation, I didn't know if the memory, motherboard or CPU IMC could do it. I ended up at DDR5-7800 until I was sent a MB, CPU and Ram binned for DDR5-8000. Turns out, it was just the memory kits I had that couldn't quite make the cut. Well now one of my original kits can once I copied the XMP profile and raised the voltage more (beyond what I wanted).

In total it was about 20 hours of troubleshooting because I had nothing to compare it to. Now its really quick because I have a reference starting point.


Real world gains would be seen in the games you play. higher frame rate. I believe AID64 shows the peak read and write and avg latency. Besides synthetic benchmarks, overclocking memory does nothing outside of gaming.

Ok so I go back to xmp run shadow of tombraider at 720p then with the settings iv dialed in and if they gone up I done good ?
I found this benchmark running different ram speeds and I got 196 minimums and 267 average

 

[ir_cow]

I never saw the point in benchmarking games at resolution you don't play it. Unless I'm submitting it for a record, I don't waste my time on stuff that gives me no real world gains.

sorry to pop your overclocking bubble lol. Only reason to OC memory is for personal system gain or submitting to HWbot.

 

[Johnthedon1]

I never saw the point in benchmarking games at resolution you don't play it. Unless I'm submitting it for a record, I don't waste my time on stuff that gives me no real world gains.

sorry to pop your overclocking bubble lol. Only reason to OC memory is for personal system gain or submitting to HWbot.

I’m overclocking it and running low res to see what kind of gains I will have from moving to a faster in the future

 

[ThrashZone]

Hi,
Must be shit kit or shit board frankly
4000c16 2x16gb only really needs 1.40v if it's a good kit and that's using 16-16-16-36 play with it later :boom:

 

[ir_cow]

Hi,
Must be shit kit or shit board frankly
4000c16 2x16gb only really needs 1.40v if it's a good kit and that's using 16-16-16-36 play with it later :boom:

or a crap IMC, or a CPU mounting issue. Why do people think memory overclocks are guaranteed when its not binned for it?

 

[ThrashZone]

or a crap IMC, or a CPU mounting issue. Why do people think memory overclocks are guaranteed when its not binned for it?

Hi,
I doubt it's the chip frankly

But the system as a whole has not really been posted here either so issues inserting memory or mounting is a possibility to :/

 

[Johnthedon1]

Hi,
I doubt it's the chip frankly

But the system as a whole has not really been posted here either so issues inserting memory or mounting is a possibility to :/

How could Me seating memory be a problem ? Surely if I put the ram In wrong it just wouldn’t work ? Anyway first slot left to right is empty then a stick then a space then a stick

 

[ir_cow]

I know my sample size is one, but the Tomahawk Z790 I had for review worked just fine with dual rank DDR4-4000. It auto configured to Gear 2 and Command Rate 2 by default. Changed it to Gear 1 and couldn't get CR 1 to work. That was with a 13900K and 12900K and memory binned for it. There is always bad motherboards, they exist. It is far more likely either the IMC cannot do it or the memory itself. I would try my method for overclocking before jumping to conclusions like the motherboard is at fault, or CPU, etc.

If you want to eliminate the motherboard DIMM slots and memory, use 1 DIMM at a time. Such a hassle.... Once again, this may not work because you don't have DDR4-4000 memory to compare to. It could be that one DIMM cannot do it and the other can.

 

[ThrashZone]

Hi,
At this point and low frequency I'd be checking/ reinstalling all psu/ board power and memory/ chip connections
Also look for bent cpu pins while you're at it because something is off.

One stick at a time is indeed the next test.
slots 2 & 4 is how asus does 2 stick configs not sure about msi though.

 

[Deleted member 185158]

Young ones don't know about overclocking before there were any guides.

Yes, I often forget to take that into account.

Already pointed young blood in the right direction.

But lack of feed back makes helping difficult.

But there's no way 14-14-8-14 is a logical B-Die timing set.

Then tell whiper snapper to run CL20 "real loose" when this is opposite of the capability of b-die.

Reading on, it's just turned into a mess.

 

[Johnthedon1]

Yes, I often forget to take that into account.

Already pointed young blood in the right direction.

But lack of feed back makes helping difficult.

But there's no way 14-14-8-14 is a logical B-Die timing set.

Then tell whiper snapper to run CL20 "real loose" when this is opposite of the capability of b-die.

Reading on, it's just turned into a mess.

There are plenty of other sub timings lengthening out write commands for trcdw for 8 to be stable so for ur benafit think of my primarys as 14-14-14-28 which is perfectly normal timing for b die boomer

 

[Deleted member 185158]

There are plenty of other sub timings lengthening out write commands for trcdw for 8 to be stable so for ur benafit think of my primarys as 14-14-14-28 which is perfectly normal timing for b die boomer

lol XD. My old man was the boomer. I'm not quite that old.

Yes and no.

Typically, you start at the top of the timing set. Since you're new.... that'd be my first suggestion which I made, you said was gonna try, and it didn't happen.

So as a semi experienced clocker, I'd say you're on the write track, just with the wrong timing/s.

So yes you have it stable in some sort of theory, but it's not quite the way to approaching OC B-die kits in most general aspects of "overclocking and tweaking"

____

Out on a limb here, just to put this into perspective, You should be a little bit higher in frequency. A sweet spot on most "daily" setups would be running 4000mhz (or higher) with Intel platforms, since oh.... 7th Gen.

I'm simply trying to start the OC over again, but with a different approach.

If you wish to keep the current timings, then do so. I'll step out and watch from the side lines.

 

[ThrashZone]

There are plenty of other sub timings lengthening out write commands for trcdw for 8 to be stable so for ur benafit think of my primarys as 14-14-14-28 which is perfectly normal timing for b die

Hi,
14-14-14-34 is xmp timing so no 14-..28 is not reasonable or normal at all but then again we aren't having any issues either you are

 

[Johnthedon1]

Hi,
14-14-14-34 is xmp timing so no 14-..28 is not reasonable or normal at all but then again we aren't having any issues either you are

Xmp 3600 was 16-16-16-38