Ballistix Sport LT Overclocking

michty_me · 23 Feb 2020 at 09:23

gupsterg said:
Cheers , NP, look forward to reading how it goes for you.

Yesterday I tried more tweaking. It seems on 3800MHz when using 2x16GB (dual sided/rank dimms), setting tRTP below 10 does not take, see right screenie, left screenie is 4x8GB (single sided/rank dimms).

It would have been handy if tRTP could have become 9, in turn this would have allowed me to knock down tRAS and tRC as well.

tRAS = tCL + tRCD + tRTP
tRC = tRAS + tRP

I have instead further tweaked tFAW & tRFC, still keeping to voltages used before, will share how it goes .

Excellent. I'll be hapy with just a nice stable setup anyway. When testing my current config at 3733Mhz, I get a single error usually but never had any issues when gaming or general use.

gupsterg · 23 Feb 2020 at 11:45

michty_me said:
Excellent. I'll be hapy with just a nice stable setup anyway. When testing my current config at 3733Mhz, I get a single error usually but never had any issues when gaming or general use.

I agree. I see from your past posts your using a R5 3600, I couldn't tell which RAM set you have?

TBH my R5 3600 can push higher FCLK than this R9 3900X and this was the best of 3 R9 3900X I had.

There is no point in pushing past an x point TBH, return is low. If I didn't have the time and access to other PCs I'd have stopped tweaking this set for sure. 6 way compare of 3x AIDA64 of timings I've been using.

Timings below just passed ~4000% in Kahru RAM Test, continuing more testing.

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [18]
DRAM RAS# ACT Time [45]
Trc [63]
TrrdL [9]
Tfaw [36]
Twr [24]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [10]

neologan · 23 Feb 2020 at 12:09

Darn, was speccing up a NAS build a few days back and didn't realise this stuff was on sale when i saw it. Didn't buy at the time. Hope it goes back on at some point as once you see it at £50-55 it's hard to grab at £75-80. Looks like nice stuff.

michty_me · 23 Feb 2020 at 13:36

gupsterg said:
I agree. I see from your past posts your using a R5 3600, I couldn't tell which RAM set you have?

TBH my R5 3600 can push higher FCLK than this R9 3900X and this was the best of 3 R9 3900X I had.

There is no point in pushing past an x point TBH, return is low. If I didn't have the time and access to other PCs I'd have stopped tweaking this set for sure. 6 way compare of 3x AIDA64 of timings I've been using.

Timings below just passed ~4000% in Kahru RAM Test, continuing more testing.

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [18]
DRAM RAS# ACT Time [45]
Trc [63]
TrrdL [9]
Tfaw [36]
Twr [24]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [10]

Yes, That is correct, I went for the 3600 for now until the 4XXX chips become cheaper then will likely upgrade to a 12 core part from that.
The RAM I am using is the 2 x 16Gb Ballistix. I don't think my 3600 can cope with 1900 on the FCLK but seems happy at 1867.

gupsterg · 23 Feb 2020 at 14:42

neologan said:
Darn, was speccing up a NAS build a few days back and didn't realise this stuff was on sale when i saw it. Didn't buy at the time. Hope it goes back on at some point as once you see it at £50-55 it's hard to grab at £75-80. Looks like nice stuff.

I would nab it, I'm glad I did. I think the Sport LT is EOL, hence ~£100 for 2x16GB and ~£55 for 2x8GB. Personally don't like the new heat spreaders that Crucial are rolling out, prefer these. Looked more unique besides being same height as dimm PCB.

michty_me said:
Yes, That is correct, I went for the 3600 for now until the 4XXX chips become cheaper then will likely upgrade to a 12 core part from that.
The RAM I am using is the 2 x 16Gb Ballistix. I don't think my 3600 can cope with 1900 on the FCLK but seems happy at 1867.

Ahh, cool

. The R5 3600 has been my fave for sure. I think the R9 3900X is overkill for me, but at times the extra threads come in handy. This one has a stonker of an IMC IMO, depending on what the gains are with 4000 series I may just stick with this til I do a big upgrade.

michty_me · 23 Feb 2020 at 15:44

gupsterg said:
Ahh, cool . The R5 3600 has been my fave for sure. I think the R9 3900X is overkill for me, but at times the extra threads come in handy. This one has a stonker of an IMC IMO, depending on what the gains are with 4000 series I may just stick with this til I do a big upgrade.

It is a fantastic CPU. I'm glad I made the jump over. Looking forward to getting home and finishing it off now in my new case and trying some of the settings you suggest.

gupsterg · 23 Feb 2020 at 16:58

michty_me said:
It is a fantastic CPU. I'm glad I made the jump over. Looking forward to getting home and finishing it off now in my new case and trying some of the settings you suggest.

I posted what I saw with tRTP here, ROG forum and OCN. On OCN a member pointed out that tRTP needs to be even when using gear dowm mode, link.

This prompted me to read a Micron DDR4 white paper and found at end of page 93/top page 94:-

When operating in 2N gear-down mode, the following MR settings apply:

• CAS latency (MR0[6:4,2]): Even number of clocks
• Write recovery and read to precharge (MR0[11:9]): Even number of clocks
• Additive latency (MR1[4:3]): CL - 2
• CAS WRITE latency (MR2 A[5:3]): Even number of clocks
• CS to command/address latency mode (MR4[8:6]): Even number of clocks
• CA parity latency mode (MR5[2:0]): Even number of clocks

So with GDME below must be even:-

• tCL
• tWR
• tCWL
• tRTP

Now I have tRTP of 8 with 3800MHz

, in turn allowed me to drop tRAS from 45 to 43, tRC from 63 to 61. Not yet benched setup, but seems as if stability maybe attainable.

So surprised this ~£100 32GB kit has some more to give

.

alewid · 23 Feb 2020 at 19:48

I know I'm not in everyone else's league but just wanted to know if the below timings are okay. Do I need to adjust any on the secondary timings?
It's the 3000 MHz CL15 memory clocked to 3200 MHz with a Ryzen 3 1200 overclocked to 3.6 MHz all cores.

It's been stable for a few months gaming after initially testing it overnight, only thing I had to do was increase the Soc Voltage to 1.075 from 1.05 after an initial error.

gupsterg · 24 Feb 2020 at 07:55

RAM Kit: BLS2K16G4D32AESB

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [18]
DRAM RAS# ACT Time [43]
Trc [61]
TrrdL [9]
Tfaw [36]
Twr [24]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [8]

alewid said:
I know I'm not in everyone else's league but just wanted to know if the below timings are okay. Do I need to adjust any on the secondary timings?
It's the 3000 MHz CL15 memory clocked to 3200 MHz with a Ryzen 3 1200 overclocked to 3.6 MHz all cores.

It's been stable for a few months gaming after initially testing it overnight, only thing I had to do was increase the Soc Voltage to 1.075 from 1.05 after an initial error.

Only thing that sticks out to me is tRDRDSCL & tWRWRSCL, you should target 2 on each with HW you have. If to gain 2 on those I needed to loosen other timings I may be inclined to, as those 2 timings are significant for gains.

dubster said:
I've got a Ryzen5 3600, and some Crucial Ballistix Sport LT BLS2K16G4D30AESC
I've never done any overclocking, but thought I'd have a look into it by reading in here.
Can I just say - the skills and knowledge in here is incredible.
Great stuff guys, and thanks for the starting pointers.

Enjoy your kit

, I'd remove competitor link.

dubster · 24 Feb 2020 at 09:01

Thanks, didnt even know Id put one in, I was only trying to copy that long text string.

keenan · 24 Feb 2020 at 18:09

@gupsterg

Could you possibly run this DRAM Caclulator Membench to compare the read speed and latency results between 3600Mhz Cl14 Vs. 3800Mhz Cl16?

gupsterg · 25 Feb 2020 at 08:55

3800C16R7 passed ~24hrs mixed testing, link to data ZIP (inc.3x AIDA64), organise by time to see process of testing.

tRTP of 6 did take. I then notched tRAS/tRC down by 2. Also rolled tRP down a notch, thus could also notch tRC a tick down. This setup had many errors in RAM Test. I then got rid of the drop on tRP, pushed tRC up again, had less errors.

I then went back to tRTP 8, set tRAS as needed, knocked tRTP and set tRC as needed, seems to be an attainable timings set.

@dubster

Yeah can happen, if you copy paste text with hyperlink already from another site to here it will apply.

@keenan

Will do, I haven't really done a 3600C14 profile yet though. If you have any pointers on what you've used it may aid me

.

keenan · 25 Feb 2020 at 09:11

gupsterg said:
@keenan

Will do, I haven't really done a 3600C14 profile yet though. If you have any pointers on what you've used it may aid me .

I'll get some screens uploaded later today. As you can see though, my trcdrd is incredibly stubborn and just doesn't want to go below 20 error free at 3600Mhz.

The test I ran above was at 3600Mhz Cl14 with the tightest error free timings possible at 1.425v

I just wanted to compare my 3600Mhz Cl14 results to your 3800 Cl16 results.

gupsterg · 25 Feb 2020 at 16:08

keenan said:
I'll get some screens uploaded later today. As you can see though, my trcdrd is incredibly stubborn and just doesn't want to go below 20 error free at 3600Mhz.

The test I ran above was at 3600Mhz Cl14 with the tightest error free timings possible at 1.425v

Ahh cool I will try those setting ASAP

.

keenan said:
I just wanted to compare my 3600Mhz Cl14 results to your 3800 Cl16 results.

This my MEMBench result, so far profile has done ~7hrs mixed testing, link.

I've never ran MEMBench before TBH, never used RAM profiles from the calculator either TBH. I can't say I've looked at MEMBench results of others before, until today. I know in AIDA64 Matisse will bench slower than Pinnacle/Summit Ridge for like MEMCLK/timings. So I will aim to soon try your shared profile to see what occurs.

In Jan I did a fresh install of W10 Pro x64 1909, fully updated with AMD chipset & GPU drivers current for that date. I was on the 4x8GB Samsung B die at 3800C16, but most timings tighter than the Micron E, R9 3900X was stock, the RX Vega 64 had PP mod applying a mild OC. I did 2 runs of 3DM FS, the best of each for total points is shared in this screenie. The timings where same for each run, but obviously as MEMCLK was higher the true latency was lower as MEMCLK climbed.

From past discussions with The Stilt, Matisse doesn't benefit with timings as much as Pinnacle/Summit Ridge did, by that he or I are not saying reduced timings don't help Matisse; just not the same way as P/SR.

Any runs of Micron E I've done in 3DM, give same/very similar results as Samsung B die, even though the Samsung B die had a lot of lower timings. Again in things like CB15/20, blender benches I see no difference between Samsung B die profiles I used and Micron E.

All in all Micron E has been a smart buy for me. I reckon I will recoup the cost of the Micron E and have ~£75-£90 in my back pocket from sale of the Samsung B die kit they replaced.

NathG79 · 25 Feb 2020 at 20:14

gupsterg said:
RAM Kit: BLS2K16G4D32AESB

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [18]
DRAM RAS# ACT Time [43]
Trc [61]
TrrdL [9]
Tfaw [36]
Twr [24]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [8]

Im running the same ram kit on a 3900X also, at 3800Mhz with your timings as above.
What dram voltage are you using for this.
Mine seems to be stable at 1.4v. (Could I drop Lower?)
Excellent Sticks aren’t they.

keenan · 25 Feb 2020 at 21:07

NathG79 said:
Im running the same ram kit on a 3900X also, at 3800Mhz with your timings as above.
What dram voltage are you using for this.
Mine seems to be stable at 1.4v. (Could I drop Lower?)
Excellent Sticks aren’t they.

I believe he is at stock stock voltage of 1.35v

DeliciousStorage · 26 Feb 2020 at 00:45

I've set my voltage to 1.4v manually, but after restarting and going back into the BIOS there is a value in the left column that shows the current voltage, and it reads between 1.424v-1.428v. After booting into Windows hwinfo64 shows the same reading. Is it normal for it to add an extra 0.024v-0.028v? Motherboard is a B450 Tomahawk Max.

keenan · 26 Feb 2020 at 02:34

DeliciousStorage said:
I've set my voltage to 1.4v manually, but after restarting and going back into the BIOS there is a value in the left column that shows the current voltage, and it reads between 1.424v-1.428v. After booting into Windows hwinfo64 shows the same reading. Is it normal for it to add an extra 0.024v-0.028v? Motherboard is a B450 Tomahawk Max.

Yes, mine does the same.

gupsterg · 26 Feb 2020 at 08:21

NathG79 said:
Im running the same ram kit on a 3900X also, at 3800Mhz with your timings as above.
What dram voltage are you using for this.
Mine seems to be stable at 1.4v. (Could I drop Lower?)
Excellent Sticks aren’t they.

As keenan stated I'm using 1.35V set in UEFI. If I use a digital multimeter to read the voltage it's ~1.355V. This ~+0.005V is it's normal variance from set value from what I have experienced on my C7HWIFI. Originally I started 3800C16 profiling at 1.39V set in UEFI, with looser timings than current. As testing progressed with lowering timings, I found I could lower VDIMM as well.

ASUS TurboV Core shows set values for a lot of the voltages on my setup and then Ryzen Master others, plus ProcODT, CAD Bus, etc. Zen2 Timing Checker shows RAM timings in same order as in the UEFI. In HWINFO the averaged figures for say SOC Voltage (SVI2 TFN), CPU SOC Voltage (VRM) and DRAM Voltage (VRM) are very close to what I read on digital multimeter.

Last night I dropped tWR by 8 in one go

, so current profile:-

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [17]
DRAM RAS# ACT Time [43]
Trc [60]
TrrdL [9]
Tfaw [36]
Twr [16]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [8]

3x AIDA64 bandwidth figures got some gains, ~10hrs stability testing in this ZIP.

Yes these are excellent sticks

, been getting a real buzz from tinkering them

.

Reous has done a great review of the new Micron E 3600C16 bin here, don't like the new heatspreaders TBH, but nice to read the IC is still sweet.

NathG79 · 26 Feb 2020 at 10:54

gupsterg said:
As keenan stated I'm using 1.35V set in UEFI. If I use a digital multimeter to read the voltage it's ~1.355V. This ~+0.005V is it's normal variance from set value from what I have experienced on my C7HWIFI. Originally I started 3800C16 profiling at 1.39V set in UEFI, with looser timings than current. As testing progressed with lowering timings, I found I could lower VDIMM as well.

ASUS TurboV Core shows set values for a lot of the voltages on my setup and then Ryzen Master others, plus ProcODT, CAD Bus, etc. Zen2 Timing Checker shows RAM timings in same order as in the UEFI. In HWINFO the averaged figures for say SOC Voltage (SVI2 TFN), CPU SOC Voltage (VRM) and DRAM Voltage (VRM) are very close to what I read on digital multimeter.

Last night I dropped tWR by 8 in one go , so current profile:-

DRAM CAS# Latency [16]
DRAM RAS# to CAS# Read Delay [19]
DRAM RAS# to CAS# Write Delay [19]
DRAM RAS# PRE Time [17]
DRAM RAS# ACT Time [43]
Trc [60]
TrrdL [9]
Tfaw [36]
Twr [16]
TrdrdScl [4]
TwrwrScl [4]
Trfc [580]
Trtp [8]

3x AIDA64 bandwidth figures got some gains, ~10hrs stability testing in this ZIP.

Yes these are excellent sticks , been getting a real buzz from tinkering them .

Reous has done a great review of the new Micron E 3600C16 bin here, don't like the new heatspreaders TBH, but nice to read the IC is still sweet.

Wow. Stock Voltage. :eek:

, need to drop my Volts then.

They certainly give these 8pack Teamgroup Samsung b-die “Binned” Sticks a run for there money.
Glad I went on the advice of @tamzzy when picking Ram.