• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Having problems setting 1T for 5950X

Joined
May 12, 2006
Messages
1,616 (0.23/day)
Location
The Gulag Casino
System Name ROG 9800X3d By purecain
Processor AMD Ryzen 7 9800X3D
Motherboard ASUS Crosshair X670E Hero
Cooling Noctua NH U12A
Memory 64Gb G.Skill Trident Z5 neo RGB 6400@6000mhz@1.41v
Video Card(s) Aorus RTX4090 Extreme Waterforce
Storage 990Pro2Tb-1TbSamsung Evo M.2/ 2TbSamsung QVO/ 1TbSamsung Evo780/ 120gbKingston Now
Display(s) LG 65UN85006LA 65" Smart 4K Ultra HD HDR LED TV
Case Thermaltake CoreX71 Limited Edition Etched Tempered Glass Door
Audio Device(s) On board/NIcomplete audio 6
Power Supply Seasonic FOCUS 1000w 80+
Mouse M65 RGB Elite
Keyboard K95 RGB Platinum
Software Windows11pro
Benchmark Scores [url=https://valid.x86.fr/gtle1y][img]https://valid.x86.fr/cache/banner/gtle1y-6.png[/img][/url]
When i initially bought this ram, i remember having to use it at 2T as it simply would not boot up with the 1800X otherwise.

I had the same experience with the 2700X, then the 3900X.

This problem has now carried across to my new CPU the 5950X. I recognise the performance cost and after spending the last 4hrs on my 3rd attempt or so trying to get it working, I'm once again finding hard walls.

After a discussion on this forum earlier in the week I was told by another user that 1T was pretty much essential for good performance.

I've checked a few threads of his and they were all correct which leads me to think there is a work around for my problem. A number of timings etc set a certain way.

If these work arounds exist, embarrassingly I have not found them when researching the matter over the years.

I'm pretty much resigned to using 2T but I'm open to any suggestions as i seemingly know very little compared to some of the members here when it comes to ram timings.

This is my memory config
G.Skill Trident Z RGB F4-3466C1-8GTZR (4x8gb B-Die)




If I am unable to run at 1T then should I be able to run tighter settings in 2T.
When I'm memory overclocking i usually get a spare SSD and install windows 10, avoid any corruption on my main install while I'm playing around.
If anyone can point me in the right direction, a little reading material or previous thread, I'd very much appreciate it.

I feel like I need to buy new memory, I just cant bring myself to do it as i know these sticks have more in them.

2 of the sticks are rated for 4000mhz, I flashed the spd's to the same as the 3466mhz sticks I already had.

Notice i didnt give you the benchmark results in CPU-Z. They are really low, so theres my motivation for learning everything I can about memory so I can get my pc performing properly.
I'm also big enough to admit I need some assistance.
What say ye???
ZenTimings5950X.jpg
MemoryNot2T.jpg
 
Last edited:
Since switching the command rate reduces the amount of time for the the chip select to occur, you might need to run the memory at a lower frequency to get 1T to work. It almost might just not work at all. Not all DIMMs can handle 1T and it gets a lot harder when there are more DRAM chips and when frequencies are on the high side.
 
Wait, two of the sticks are 4000mhz and you flashed them to the lower speed ones? Sorry I don't get why one would do that. If it were me I would flash them back and then go get two more 4000mhz matching dimms and go from there.
 
Since switching the command rate reduces the amount of time for the the chip select to occur, you might need to run the memory at a lower frequency to get 1T to work. It almost might just not work at all. Not all DIMMs can handle 1T and it gets a lot harder when there are more DRAM chips and when frequencies are on the high side.
Thanks for the assist, It says 1067mhz @ input clock latency of 0.938ns. Is this not the minimum standard for DDR4. Is there anyone who can read a little deeper into the info. I'm hoping someone notices or knows something, that ive been missing.

@thesmokingman My 3466mhz sticks were identical in every way to the 4000mhz rated sticks. I still have the bios and may flash the chips the other way, they are all B-Die.

At the time there was a shortage of memory, 4000mhz was the second fastest you could buy at the time. It was all the same ram overclocked.

If this is a hard limitation would it be worth considering upgrading to some newer IC's. Maybe try to find some binned very low latency B-Die.

I'll see what I can find out.
 
Last edited:
Pump the voltage up. B die is completely safe daily driving right up to 1.7v if its kept cool (under 40c).
 
VDDG and Soc voltages can help too.
 
Dont worry I know enough about ram not to burn them or cause a problem. I'm just reading through some of the other threads. I need to visit this part of the forum more often.
I thought about setting 1.1 for soc voltage. Ive left it on auto for now. :toast:
 
Also your tRFC is too high for B die, it's extremely important fo Ryzen. Increase your ProcODT to at least 40, and your VDDG voltages to at least 1.0v, Vsoc at 1.1 can help stability, higher doesn't always guarantee scaling.

3.png
4.png


There's plenty of people who've run B die at 1.7 for literal years, and some at higher, most important is it's kept cool when you increase over 1.5v, have a ram fan. B die is stablest at 20c range, but it doesn't get seriously bad until 45c+. Aim for 40c and below.

5950x should do at least 3800/16 3600/14 with 4 sticks of B die.
 
@thesmokingman My 3466mhz sticks were identical in every way to the 4000mhz rated sticks. I still have the bios and may flash the chips the other way, they are all B-Die.

At the time there was a shortage of memory, 4000mhz was the second fastest you could buy at the time. It was all the same ram overclocked.

If this is a hard limitation would it be worth considering upgrading to some newer IC's. Maybe try to find some binned very low latency B-Die.

I'll see what I can find out.
I don't know if that is true or even how you could prove this especially in respect to binning.

Technically speaking all the ic's are the same, ok sure. Except they are not due to the binning process. A lot of guys will say buy the lower speed b-dies cuz the stuffs all the same, however they are conveniently ignoring the binning process.

For me, if they were the same, you wouldn't be in this situation. The 3466 kit have mismatched timings. This tells me they could not run matched timings, ie. not top grade. What were the advertised timings of the 4000mhz dimms?
 
Loads of people were getting good results overclocking the 3466mhz memory and above. I dont know about binning below that as i didnt own any. I know my 3466mhz set has performed exactly like my 4000mhz set. What does that tell you. They added more voltage and found more head room. Programmed the spd to apply these speeds and they had a new high end product. Of course binning goes on. You just cant deny this happened. Its just business.

3466MhzSPD.jpg
4000MhzSPD.jpg
 
Last edited:
@purecain

For starters, you gotta stop flashing your DIMMs (unless to get around the RGB bugs). There's no problem with buying a 4000CL18/19 kit and taking the time to learn about the timings to be able to set and test any custom profile you want at any speed. Why you would voluntarily go out of your way to sabotage your own warranty and possibly affect the integrity of your sticks is beyond me :confused: it doesn't do anything for the potential or performance of your sticks.

Your tRFC is fine. 140-180ns is okay for B-die, once you go below ridiculous (200ns) tRFC the differences aren't that noticeable anymore but you start needing to add V to keep extreme tRFC stable.

Tighten up your RRDS/RRDL/FAW/WTRS/WTRL/WR. That's going to hold back your memory benches, not your 180ns tRFC. Do at least 4/6/16 4/12/12. Can go down to 4/4/16 4/8/10, but be advised that will likely require more voltage.

Wouldn't go so far as 1.7V especially with a hot 4-DIMM setup like yours, but 1.5V is certainly safe as long as you keep them cool.

As for 2T vs 1T, 1T should always be run with Geardown on unless you're chasing that elusive bench result that's 0.3ns away. It's an AMD exclusive and essentially gives you close to 1T performance while not requiring much more voltage than 2T.

What VDRAM are you running currently? No reason why 3533 16-17-17 at loose timings should be unstable 1T GDM unless you're trying to run like 1.2V.
 
Loads of people were getting good results overclocking the 3466mhz memory and above. I dont know about binning below that as i didnt own any. I know my 3466mhz set has performed exactly like my 4000mhz set. What does that tell you. They added more voltage and found more head room. Programmed the spd to apply these speeds and they had a new high end product. Of course binning goes on. You just cant deny this happened. Its just business.

View attachment 183436
All I can go buy is what you post. The timings are not great. If they were b-dies they should run 1T in their sleep w/o any added voltage.
 
I only flashed them to the same timings so that DOCP would detect them all as an identical set.

Ive done no harm whatsoever, i do however appreciate the advice.

Luckily i'm not that much of a noob with ram and that statement, 'If they were b-dies they should run 1T in their sleep w/o any added voltage' is what had me start this thread.

I must of spent hours reading about memory timings over the years. I just always had trouble with both sets installed. I didnt familiarise myself with the sqew settings for the memory channels but i may still get around to those settings as my interest has been piqued. I have medium knowledge of memory timings. I can understand technical advice. np.
ps i wont ever be running 1.7V :toast:
 
You don't have to run 1.7v, that's just the upper safe limit, personally I use 1.62v.
 
First reboot changed to run 3600Mhz 1800IF
3600Mhz.jpg


Time to lower some timings. i'll post my terrible results from the cpu-z benchmark
BTW this is using an Noctua A12 air cooler. I have a Thermaltake 360rad which I can use if this air cooler cant keep up.
badScore.jpg

I havnt overclocked the CPU iether yet but the memory not being optimised will definitely be having an effect.
I saw someone score 680 single thread in another thread using a 240 aio...
 
I only flashed them to the same timings so that DOCP would detect them all as an identical set.

Ive done no harm whatsoever, i do however appreciate the advice.

Luckily i'm not that much of a noob with ram and that statement, 'If they were b-dies they should run 1T in their sleep w/o any added voltage' is what had me start this thread.

I must of spent hours reading about memory timings over the years. I just always had trouble with both sets installed. I didnt familiarise myself with the sqew settings for the memory channels but i may still get around to those settings as my interest has been piqued. I have medium knowledge of memory timings. I can understand technical advice. np.
ps i wont ever be running 1.7V :toast:

I can tell you that the 4000 kit can run anything reasonable in its sleep, but I can't tell you if the 3466 kit is "shitty B-die". The point about not flashing is that it doesn't improve what the ICs are like from the factory, if the 3466 kit really is shitty then no amount of flashing will change the voltage they need.

Seeing as you are most likely limited by the lower end kit, you can try bumping voltage to see if you can run 1T GDM.

Like i said before, your screenshot shows 1T GDM off. Unless you have a really good kit, happened to settle on unicorn CADBUS settings that allow you to do so, and are willing to pump extra V to compensate, don't ever go there, keep GDM on when on 1T. I didn't think it a problem either until my entire Windows install gave up after 6 months of unstable GDM off on my CJR kit.

You need to take care of GDM before you try anything else.
 
3466Mhz is not a low end kit, at the time it was the best you could get. I've tested both kits separately as when I first bought them i expected differant performance. they both behaved like identical ic's and what one couldnt make neither could the other.
There wasnt much ram about at the time, a shortage and the 4000mhz kit was announced shortly after the 3466mhz, then the 4133mhz kit so on and so forth. I bought the kit for the same price as i bought the 3466mhz and at the time it was a no brainer. Take it from me, they behave very similar if not identically.

set 1800mhzIF mclk fclk
1800mhzMCLK.jpg

I also lowered the timings suggested by tabascosauz... and gear down mode on.

The pc will hand with Q-code 64(CPU Initialisation) for about a minute then boot.
Performance has gone down in the CPU-Z Bench.
I'll back the IF back down until the next bios hopefully fixes this issue with the Crosshair VIII boards
 
Last edited:
3466Mhz is not a low end kit, at the time it was the best you could get. I've tested both kits separately as when I first bought them i expected differant performance. they both behaved like identical ic's and what one couldnt make neither could the other.
There wasnt much ram about at the time, a shortage and the 4000mhz kit was announced shortly after the 3466mhz, then the 4133mhz kit so on and so forth. I bought the kit for the same price as i bought the 3466mhz and at the time it was a no brainer. Take it from me, they behave very similar if not identically.

set 1800mhzIF mclk fclk View attachment 183446
I also lowered the timings suggested by tabascosauz...

The pc will hand with Q-code 64(CPU Initialisation) for about a minute then boot.

I dunno which post you read but those aren't the timings I suggested...

RRDS: 4
RRDL: 6
FAW: 16
WTRS: 4
WTRL: 12
WR: 12

Also get your RDRDSCL and WRWRSCL to 5 for now.

Aside from your primaries (CL to RC), the above timings, setting GDM, and tRFC, leave all the other timings on Auto.

Start from 1.35V VDRAM and 1.1V VSOC. Leave your CLDO minor voltages on Auto.

I don't disagree with your assessment, but I have no idea the extent to which you have tested these sticks, whether they're from different vendors and product lines, and whether they are A0 or A2 PCB.

Don't mess with your procODT unless you're dedicated to testing every option exhaustively. There's no "sliding scale". procODT is wildly different depending on whether you have single rank or dual rank sticks and how many sticks you have.
 
I've spent many an hour with an 1800X/ 2700X /3900X trying to find the best ProcODT.
 
Personally I wouldn’t run them more than 1.45~1.48V unless I’m after some extreme OC or super tight timings. Even if B-dies can tolerate 2.0V with proper cooling.
If I’m not mistaken the main concern of this topic is stability first at reasonable speed and timings with 1T.

My suggestion for a START would be.
3600MHz (1800MHz FCLK, and UCLK=MEMCLK)
DOCP: On/Enabled (for now)
DRAM: 1.4V
DRAM VTT: 0.7V (always half the Ram Voltage)
GDM: Enabled
Cm2: 1T
16-18-18-18-36-54 (first 6 of ZenTimings)
tRFC: 432 (tRC x8) (I know it’s high for the ICs, but bare...)
tRRDS: 6
tRRDL: 9
tFAW: 24 (tRRDS x4)
tWR: 18
ProcODT: 43~48
Leave every other timings on Auto (for now)

Power savings settings
SoC/Uncore OC Mode: Enabled
PowerDownMode: Disabled
DF C-states: Disabled

VSOC: 1.11~1.12V (should land around 1.09~1.1V actual)
CLDO VDDP/VDDG: Auto

CLDO VDDP is UCLK voltage and VDDG is FCLK voltage.

————————————

Check your current(2T) stable AIDA64 memory benchmark score to have baseline for future reference.

Since all 4 sticks don’t have temp sensors and a dense configuration like yours with minimum or no space between them can heat up rather quickly it’s best to have a fan over them for additional air flow.
 
Theres an issue with the bios, i used to run my 3900X at 3600mhz 1900 IF. With this bios and CPU the pc starts acting up at 1800 IF.
These are the best settings ive had for memory so far. I'm using Super Pi to test my settings...
Does Gear down mode affect the 1T 2T setting???
As it says the memory is set to 1t here and I am almost certain it is set to 2T.
I'll buy another Noctua Fan and set that in place. I may also buy one for the rear fan to help increase the airflow through my case.
The 3090 blows straight into the case.
MemSettings.jpg
 
If you're on a Ryzen 5000, you probably want the latest AGESA 1190 BIOS. IF struggles with earlier AGESA. Plenty of Ryzen 5000 owners have had trouble running 3600-3800 on AGESA 1100-1180 and it's not been the fault of the memory.

Geardown overrides command rate, so it's more like 1.5T when it's on. Just leave CR on 1T and set GDM on. Not sure why you're worrying.

Use HCI Memtest, Karhu, or TM5 w/ anta777 extreme. Round it off at the end if it's stable, with P95 Large FFT or LinpackXtreme. Never heard of SuperPI used to twst memory.
 
I'm not worrying, i just wanted to learn is all. :toast:
 
Hello!

I have 2 packs of 2 modules each F4-4000C18-8GTZR I mean a total of 32gb in a z490 maximus formula and I have no way to make it stable at anything other than the XMP frequency.

Can you help me?

First, what software can I use to monitor those times? I tried some but they didn't show me anything.

I would love to at least be able to make 4200 or 4300 stable to cl18
 
Back
Top