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

Tips for non-kit memory stability? / Two "bad" 64Gb DDR3 kits

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
For 7 years, I've been trying to solve this mystery. I've tried many things, and now completely ran out of ideas.

I bought two identical PCs: i7-3820, 64Gb DDR3 Corsair Vengeance GMZ64GX3M8A1600C9 , Gigabyte GA-X79-UP4 . And both systems are unstable - segmentation faults and memory corruption. Everything points to bad memory - but I can't believe both kits are bad, so I also suspect everything else. I could have returned the kits 4 years ago, if by that time I was sure that I should return the RAM (both of them) - and not both CPUs, or both mainboards, or maybe I'm just doing something wrong... Segmentation faults can happen for a lot of reasons.

Is there anything I can do to make it stable? By now, I would be fine with lower speeds/timings, or missing one or two sticks in each kit. The problem is, NONE of that helps: lower speeds/timings do not help even a little bit, and removing one or two sticks help a little. With only half the kit, it's *almost* stable, but not still stable enough for production (medium-load ERP servers). Some sticks show better "compatibility" than others. I think we can treat this as a "mixed kits" case for the purposes of system tuning.

- - -

Now, the details and the journey.

The test I'm using is building the Linux kernel with 8 threads, on a "ramdisk". That's easy to script, it emulates the real workload, and it seems to catch errors much faster than OCCT. Each build takes 15G "disk" space and some RAM usage, so building it three times tests over 45 Gb - good for testing 6-8 sticks. I have "Channel Interleave" and "Rank Interleave" enabled, so that should be a pretty good test. I test it on one of the two PCs, the other showing exactly the same behavior.

At first, I thought the kits might have been mixed, but then I've found numbered warranty labels on each stick, and their numbering makes sense. All sticks are the same - batch (apparently) 130502368 , serial 236xxx , so I'll refer to them by the last three digits of the serial.

Here are the kits - warranty stickers in the first column, serials in the second column:

Kit 1:
27 351
28 346
29 305
30 352
31 353
32 291
33 508
34 354

Kit 2:
35 262
36 506
37 507
38 226
39 289
40 227
41 225
42 288

I can't entirely dismiss the possibility that the shop employee was a complete idiot and went out of his way to mix up the kits before putting the stickers on them, but if we look at the first kit in sequence of the stickers, we see: 351-...-...-352---353-...-...-354 . Whatever pattern they were using, that seems to be really one kit. And the numbers on the second kit also follow this pattern: ...-506-507-...---289-...-...-288 . Also, 225, 226 and 227 belong to one kit. Based on that, I think those are not mixed.

Of course, I've tried them in all kinds of other combinations, which did not help. If I put two "better" half-kits together (351-352-353-354-506-507-288-289), then it lasts for a few builds (2-3 hours), but the sticks left are completely unusable as a kit. Both kits fail under 5 minutes in seemingly any stick order.

I once had an opportunity to test it with a known good memory kit (another 64Gb Corsair Vengeance) and a known good system (i7-4930K, GA-X79S-UP5). The faulty system worked with known good memory! (Though I do not remember if I tested it long enough and well enough...) And faulty memory was "much less faulty" in the "good system", but that was before I noticed the warranty stickers, so I could not test the kits as intended - and it was still not good enough.

I have once managed to get it to run seemingly stable with 7 sticks, but then decided to continue the pursuit of perfection, and did not write that combination down. %)

After that, remembering that i7-4930 helped somewhat in the past, I've tried this system with an i7-4820. It is said to have a better IMC - maybe this was the reason?.. But it did not make any difference at all.

I've tried a better power supply. That did not make any difference at all. I've tried another one (600W this time) - again, no difference. Or maybe a little bit - I'm still testing this. But surely 500W is enough? 150W for the CPU, 100W for the mainboard+memory, 30W for a GeForce GT610, and I'm testing with only one SSD attached.

I've tried raising voltages and lowering speeds and timings. No difference at all. But I've never done any overclocking other than simply raising the "turbo limit" and CPU core voltage, so maybe there is something else I can try?..

The results I have right now are roughly this: with each kit, 8 sticks fail under 5 minutes, 7 sticks fail under an hour, 6 sticks fail under two or three hours, and 4 sticks are not enough to run the intended application (and if I remember correctly, is not completely stable either).
 

sneekypeet

Super Moderator
Staff member
Joined
Apr 12, 2006
Messages
28,472 (5.36/day)
System Name His
Processor Intel i9 7920X
Motherboard Asus Prime X299 Deluxe
Cooling Corsair H115i RGB Platinum
Memory G.Skill TridentZ RGB 32GB @ 3600MHz
Video Card(s) nVidia RTX 2080ti
Storage Samsung 960 Pro / Crucial MX300 750GB / Seagate 1TB Spinner
Display(s) Samsung 55" 4K 60hz
Case Cooler Master Cosmos C700P (Inverted Layout)
Audio Device(s) Realtek on board > Sony Receiver > Cerwin Vega's
Power Supply Thermaltake TPSG 1050W
Mouse Always Changing
Keyboard Always Changing
Software Windows 10 Pro 64
Run them at 1333MHz or add some IMC voltage. TBH, the only way of really knowing if it is density versus a bad stick is to test them all individually, and if you find no faults, it is more likely the IMC failing to handle the load.

That is a lot of density for either systems IMC to deal with.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
Actually, I could try them individually. I remember some of them being unable to work with some others even in a less-than-four configuration. But I'm testing with "the best sticks", leaving the more suspicious ones out for now.

Running them at 1333 with timings reduced to 10-10-10-30 makes no difference. Raising the IMC voltage or DRAM voltage seems to make things worse; apparently, because I have no idea about what voltages to try (and I know the mainboard usually suggests really stupid values as "normal", so that's no help either). I remember there are a lot of rules about "keep IMC voltage exactly this much higher than DRAM voltage, while DRAM voltage must be around this, and the CPU voltage not higher than this much from IMC voltage"... Is there any rules like that?

I've tried running the CPU at "stock speed" with 1.330 V , which AFAIK is way too much, but it made no difference either.

EDIT: Also, if it's the IMC (which I think is the only logical explanation), then shouldn't a 4820 have helped?.. And also I'd expect the 3820-s to be rated "not supporting 64Gb", if their IMC is not up to the task...
 
Last edited:

sneekypeet

Super Moderator
Staff member
Joined
Apr 12, 2006
Messages
28,472 (5.36/day)
System Name His
Processor Intel i9 7920X
Motherboard Asus Prime X299 Deluxe
Cooling Corsair H115i RGB Platinum
Memory G.Skill TridentZ RGB 32GB @ 3600MHz
Video Card(s) nVidia RTX 2080ti
Storage Samsung 960 Pro / Crucial MX300 750GB / Seagate 1TB Spinner
Display(s) Samsung 55" 4K 60hz
Case Cooler Master Cosmos C700P (Inverted Layout)
Audio Device(s) Realtek on board > Sony Receiver > Cerwin Vega's
Power Supply Thermaltake TPSG 1050W
Mouse Always Changing
Keyboard Always Changing
Software Windows 10 Pro 64
IMC is a crapshoot, a lottery if you will. There is not guarantee that changing a CPU to a newer generation will always give you a better IMC.

When dealing with densities over 32GB on the older systems, going by the QVL becomes more important, as many times you are left to fiddle with something that may not be tuned to work in the first place.

I honestly hope you didn't mix any of them up as you may even be dealing with various ICs on all of the sticks, making your quest even tougher.

It also seems you covered this at Corsair, and are getting the same answers here as there.

However, they noted that a slot might be the issue. have you inspected the slot to ensure no fingers are damaged, is it possible you might have a bent pin or more in the socket?
 
Last edited:

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
Those were actually 8-pack kits! And this is Corsair, so I expected them to "just work" (on XMP!) as long as I don't do anything "above stock". The batch is the same, and even the serials seem to be "well picked"!

If I remember correctly, the QVL for this mainboard has ONE model. %)

Thank you!
 

sneekypeet

Super Moderator
Staff member
Joined
Apr 12, 2006
Messages
28,472 (5.36/day)
System Name His
Processor Intel i9 7920X
Motherboard Asus Prime X299 Deluxe
Cooling Corsair H115i RGB Platinum
Memory G.Skill TridentZ RGB 32GB @ 3600MHz
Video Card(s) nVidia RTX 2080ti
Storage Samsung 960 Pro / Crucial MX300 750GB / Seagate 1TB Spinner
Display(s) Samsung 55" 4K 60hz
Case Cooler Master Cosmos C700P (Inverted Layout)
Audio Device(s) Realtek on board > Sony Receiver > Cerwin Vega's
Power Supply Thermaltake TPSG 1050W
Mouse Always Changing
Keyboard Always Changing
Software Windows 10 Pro 64
Those were actually 8-pack kits! And this is Corsair, so I expected them to "just work" (on XMP!) as long as I don't do anything "above stock". The batch is the same, and even the serials seem to be "well picked"!

If I remember correctly, the QVL for this mainboard has ONE model. %)

Thank you!
I did a few edits in the above post. Both of your sets have the same revision number?
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
Yes, same everything - which made it very hard to make sure the kits are not mixed, until I noticed the warranty labels. :) Not marking the 64Gb kits is a "brilliant" idea on their part =/ It could be that I even saw the warranty labels back then, but since it did not help, I decided not to trust them.

The Corsair post was when I could try it on the third system with a 4930K. Turned out one of the sets I got as a result worked with 56Gb, and the other one was failing even with 48, so I definitely did not test long enough... And so I've decided to try again to get more.

I've brushed the slots in case there's dust, but the slots and the modules seem to be not damaged: it does not matter which slots I'm using, otherwise I would have noticed it (I'm watching carefully for any patterns to give me more ideas, trying different slots and different stick order).

I think nothing is "broken" - it's just "not good enough", so "de-overclocking" could help, but I did not manage to get any results with it. I'm setting the CPU voltage manually, because I don't trust the mainboard's default on this, but all other voltages on "Auto" or "Normal", because I don't know what to set them to. And if I touch them, it makes things worse. %) Disabling interleaving seems to help. but that's probably because this way I'm not testing all the modules at the same time; when it works with interleaving, then I might turn it off for even more stability. Since running at 1333 makes no difference, I'm trying at the only available XMP profile - 1600. Also, I'm still testing on 1.330V on the CPU, even though I'm almost sure 1.3 is well enough, and the mainboard suggests 1.250 or 1.285.

It seems like "Normal" voltages are actually more stable than "Auto" - I'll see if I can confirm this.
 
Joined
Sep 16, 2018
Messages
727 (0.94/day)
Processor X5690 ES
Motherboard Asus Rampage III Formula
Cooling Thermalright Le Grand Macho RT, TY143
Memory 10Gb G.Skill mix
Video Card(s) EVGA GTX 980 Classified 3988 @ 1495/2001
Storage Intel 545S
Display(s) Hisense 55"
Case Fractal Design Meshify C TG, White
Audio Device(s) Asus Xonar Essence ST
Power Supply CM Silent Pro M2 850
Mouse Zowie EC2 Evo
Keyboard Logitech G15
Software Yes
I had a similar problem with my x58. Surprisingly an older bios than the one I was using was the key. Really weird. I had just tried it out a couple of months ago. Should have tried that like 8 years ago lol.
 
Joined
Mar 18, 2015
Messages
2,784 (1.36/day)
Location
Long Island
1. The time to address this was back when everything is new .... tho can we assume that the Corsair sticks have a lifetime warranty ?

2. Not knowing whether the each PC has its own kit makes it harder.

3. I don't think about the PSU wattage in instances like this but more with regard to voltage stability and electrical noise. Asking the IMC to handle 8 sticks is a bigger ask then 4 or 2. When we build 32 GB systems back then, we always went 4 x 8GB to make life easier on the IMC. Today on desktop builds, we do 2 x 16GB rather than 4 x 8GB. That decision came from experience when upgrading from 2 to 4 sticks often resulted in reducing the OC that was present before the upgrade.

4. Has anything changed since 2019 ? I went looking for the specs on those kit and the only two hits I got was this thread and and the one below. There you had one system working with 8 and one with 7.

5. You tried raising voltages .... how far ? I'm running two sticks of 2400 at 1.71 Auto didn't work

6. I once ran into an issue like this, tho in a much simpler scenario. My system has 2 x 8GB DDR 3-2400 CAS 10 Mushkin Redlines (Hynix 10-12-12-28) .... another build I did for a buddy had the same specs w/ Corsair ... about 18 months later, he bought another pair (same model / series ) and added, taking his chances using 2 sticks from different sets. It was a no go. I got involved and noticed that Corsair had changed memory suppliers and after version 4.51, they had timing of 10-12-12-31. We tried voltages up to 1.725 .... (staying within the +5%) w/o luck, manually setting all voltages, taking timing down to 10-12-12-31 .... and lower even tried various CAS 11 settings.... nothing worked. Then on a lark, I tried my Mushkin set with his old Corsairs and everything was fine. We checked and Mushkin was still using Hynix, so he bought a set of those ... no issues. I have seen folks who spend their days trying to get on OC Leader boards, running that Mushkin Redline set up to 1.94 v.

7. Having you tried minimizing the variables and running memtest86+ ?

8. As far as rules ... the web is full of misinformation. Back in DDR3 days we say 4-5 rants per thread on how DRAM voltage should never exceed 1,5 v .... yest most hi speed RAM has an XMP setting in the SPD of 1.65 and Intel stated +5% was no worry. In addition, Intel published XMP compatability lists which showed voltages over 1.65. Again with DDR4, we see posts all the time exclaiming that intels max is 1.35 (even 1.20 sometimes) and yet Intels response is "1.5v is the absolute max we allow for XMP certifications. However, good DDR4 memory will run at 1.35v up to 3200. " And yet, if you go to Intel's XMP page ... they have ***certified*** many RAM / MoBo combos up to 1.55


As i said in Item 4 above, I was unable to check to find the specs on your kits to see what the XMP voltage was set at. Take a peek at the voltage of your kits by looking at the SPD tab w/ CPUz. Personally, the limit of my bravado is SPD tab + 5% ... Might also wanna check Intel's XMP files at the links above .... I checked the on-line PDF and didn't foind your MoBo listed. Might be something close tho.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
I've bought all components (at the same shop) and built the systems myself. I did not know what to return - CPUs, RAMs, motherboards, or something else... and I thought it was most likely a user error rather than two "bad" somethings. The shop warranty for the RAM was three years, but I don't know about Corsair warranty - I'm not even sure there was any warranty paper included with the kits...

Since that Corsair forums thread, I've found out that those systems actually did not work - instead, one worked with 7 sticks, and the other one did not even with 6. And, of course, I did not write down the working 7-stick combination, because I was certain that this time I'll get it better - now that I've realized what the correct kits are... %)

I've tried 1.35 on the CPU, 1.6 in the DRAM and 1.2 on the IMC. That only made things worse, so I did not do that again. :)

Memtest is much worse at detecting this problem - I think because it does not load the CPU enough, and does things very sequentially.

As far as I understand, my mainboard takes the "recommended" voltages from the SPD, and it was showing 1.5 , so I believe it.

- - -
(Oh, this forum concatenates posts... so I should do this to mark the beginning of a new one)

Seems like someone had this problem almost exactly: https://forum.corsair.com/forums/showthread.php?p=643727 . (Except he says he has faulty channels, and I do not - I checked that to be sure. Maybe he's mistaken about it, or maybe his problem really is different...)

I never thought the mainboard could be a problem - the IMC is in the CPU, so its only job is to connect the wires together!.. When I was testing it on another system, that one had a better mainboard. So apparently this specific mainboard has a grudge against this specific kit model...

And when I was testing another memory kit, I think it was CL10. And mine here is CL9. Could it be possible to "turn this one into a CL10"? I'm trying to modify the timings to 10-10-10-30, but all the other timings' defaults do not change, and I do not know how should I change them.

Also, does anyone know how "interleaving" works? If I leave empty, for example, the second half of the first channel, how does it "interleave" the rest of it then?
 
Last edited:
Joined
Dec 14, 2019
Messages
222 (0.70/day)
Location
Loose in space
System Name "The black one in the dining room"
Processor Intel i7 5960X @4.6GHz
Motherboard Asus X99 Deluxe
Cooling Corsair H105 w/4 Silverstone FM121 fans @110cfm ea on the radiator
Memory 64GB G.Skill Ripjaws V DDR4 2600 (8x8)
Video Card(s) XFX R9 Fury X
Storage 960GB MyDigital NVMe. 480GB MyDigital NVMe, 1TB Samsung 860, 4 Western Digital 2TB Black
Display(s) 40" Samsung 8000 series 4K
Case Modded Corsair Carbide 500R
Audio Device(s) Asus Xonar D2X
Power Supply Corsair AX1200i
Mouse Logitech Performance MX
Keyboard Logitech K750 Solar, DiNovo Edge
Software Win 10 Enterprise LTSC
It's been 5-6 years but I used to have long telephone conversations with The Ram Guy at Corsair. On the subject of large amounts of RAM (64GB and greater) he told me to run DDR3 at 1.65 volts. He said memory controllers can get wonky under heavy loads and some motherboards & CPU's work far better than others. I've never had a single issue running 64GB of RAM on any of my Asus Sabertooth X79's with 3930K, 4960X and Xeon E5 1680 V2 CPU's. All were/are overclocked to a minimum of 4.5Ghz, 1.4 core volts and the RAM at 1.65 volts. .
Run Memtest and see what comes up.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
When I raise DRAM voltage, leaving all other voltages on "Auto", the system crashes almost immediately. Apparently, Gigabyte BIOS is really bad at guessing the correct voltages when some of them are changed, and I don't know what to specify there.

My backup option is to give up on those two and buying a new system with 64Gb or even 128Gb DDR4. But DDR4 has this same problem with density too, doesn't it?..
 
Joined
Jul 9, 2016
Messages
514 (0.33/day)
System Name Main System
Processor i9-10940x
Motherboard MSI X299 Xpower Gaming AC
Cooling Noctua NH-D15
Memory G.Skill 64GB @3600MHz XMP
Video Card(s) EVGA GTX 1080 ti FTW3
Storage 2TB Samsung 970 EVO Plus; 2TB Corsair Force MP600
Display(s) LG 43MU79-B
Case Corsair A540
Audio Device(s) Creative Lab SoundBlaster ZX-R
Power Supply EVGA G2 1300
Mouse Logitech MK550
Keyboard Logitech MK550
Software Windows 10 Pro
Benchmark Scores Cinebench R20 - 6910
1. Did you test your system with Passmark Memtest86? To me that is the de facto tool to test for faulty memory stick.
2. Your memory kits have mixed sticks or maybe even some individual 8gb sticks. The serial number is supposed to be sequential. I have had Corsair vengeance 64gb kit in the past and still have some 32gb kits now and the serial numbers were sequential. Does each stick have the right model number of the 64gb kit on the sticker?
3. Which Corsair kit is it? Does it have XMP? If so enable it and it should present with the right voltage. I did have to raise the DIMM voltage from the XMP 1.5v for one of my systems to 1.55 with 8 sticks of vengeance pro and a 4930k.
4. There is another voltage that you can raise but I forgot what it is called so I will have to look it up.

Really, try the memtest first and see if you have any bad sticks in the group. You can test one by one or in a group.
 
Last edited:
Joined
Dec 14, 2013
Messages
1,624 (0.65/day)
Location
Alabama
Processor Ryzen 2700X
Motherboard X470 Tachi Ultimate
Cooling Scythe Big Shuriken 3
Memory C.R.S.
Video Card(s) Radeon VII
Software Win 7
Benchmark Scores Never high enough
I did a few edits in the above post. Both of your sets have the same revision number?
The thing about Corsair is in the past they've been known to use different IC's across sticks, even of the same revision/model and so on.
It's a crapshoot with Corsair at times.
No one except Corsair themselves really knows "Why" but it's like they use whatever IC's are available (In the assembly line's IC bin?) at the time a set is made just to make them.

You can check on what IC's are in what and Corsair is just all over the place and they have so many revisions too.....

I'm fairly sure they haven't changed along the way in how they do things and this may even account for why so many seem to have problems with them concerning DDR4 sets in a setup.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
@LFaWolf

I tried memtest86+ (not Passmark - the GPL one) a long time ago, but since it does not load the CPU, it takes MUCH longer to detect failures than any other test I've used. And the problem here is not a single faulty stick - most sticks work unless there is more than four of them, and they are combined with specific other sticks they "don't like". I was using OCCTPT with Linpack to test the memory under load, which seems to find errors almost as fast as building kernels. But Linpack provides constant load, and I'm suspecting power among other things, so building kernels is better in my case.

I've never seen a kit of 8 sticks with all of them being sequential numbers. Even 2-stick kits are very often non-sequential. My kits have the same everything on them (version, voltage, model, batch) except the three last digits in the serial number. And those were in two 8-kit blister packs. And they have sequential warranty stickers on them (this shop puts warranty stickers on each module at the moment of purchase).

It is 64Gb Corsair Vengeance GMZ64GX3M8A1600C9. It has XMP (one profile for 1600MHz), but enabling or disabling it makes no difference at all. the problem with voltages is, there are 5 or 6 voltages for each channel listed in my BIOS, and changing only one of them makes the test crash almost immediately (Gigabyte seems to be really bad at auto-tuning the rest of voltages set to "Auto").

@Bones

Ummm according to this, they are not supposed to: https://forum.corsair.com/forums/showthread.php?t=68811
And certainly not across sticks in one kit, I'd hope... (They have those for DDR2 and DDR1 as well.)
BTW, I found that the "known good" kit I've tested it with was the same model, not CL10. But I think it had a different version - meaning, ICs. Which leads me to believe that this mainboard hates Microns...
Is there any software I could use to identify the IC manufacturer? (Windows or Linux.) I saw Thaiphoon Burner recommended, but it is "This program is not intended to run in your country or region". %) So yeah, probably something GPL would work better. %)

I don't want to touch Corsairs anymore - out of three kits I've bought, two were broken out of the box. Are there any 64Gb/128Gb DDR4 kits from other manufacturers that are more reliable in this regard? (In case I end up having to buy a new system after all.)

- - -
NEWS:
I've found one combination of sticks (4 from one kit and 4 from the other one) that lasts for about 4 to 6 hours! But only if I set the CPU voltage to 1.330, which is kinda too much for a 3820, and... if I overclock it, simply raising the turbo limits to 44/44/43/43 (the locked maximum). If I DON'T overclock it (i.e. return the turbo limit to the default 38-38-37-37), then tests crash almost immediately (and often the system won't even boot)! What does this tell us?.. I can't make sense of this...
 
Joined
Jul 9, 2016
Messages
514 (0.33/day)
System Name Main System
Processor i9-10940x
Motherboard MSI X299 Xpower Gaming AC
Cooling Noctua NH-D15
Memory G.Skill 64GB @3600MHz XMP
Video Card(s) EVGA GTX 1080 ti FTW3
Storage 2TB Samsung 970 EVO Plus; 2TB Corsair Force MP600
Display(s) LG 43MU79-B
Case Corsair A540
Audio Device(s) Creative Lab SoundBlaster ZX-R
Power Supply EVGA G2 1300
Mouse Logitech MK550
Keyboard Logitech MK550
Software Windows 10 Pro
Benchmark Scores Cinebench R20 - 6910
Use memTest86 from Passmark. It is free. Why are you using an ancient memtest86+ which was no longer supported back in 2012? Use the right tool for the right thing. Change your testing tool and stop using OCCTPT or whatever that is.

Nah, all the kits if certified for Dual or Quad Channels will have sequential serial numbers. Look at the Corsair DDR3 I have attached. The G.Skill 64GB DDR3 kit also has sequential serial numbers but I don't have the box with me at the moment to take a picture.

You mentioned you get the memory from a shop. Is that your local computer shop and where are you located? I have never heard of a shop putting the warranty sticker on when they sell it to you, because the warranty sticker should have been on the memory stick and the retail box, put on by the manufacturer. This tells me you probably did not get a kit of 8 but just some random sticks, because a kit of 8 is expensive as the manufacturers specifically test them for compatibility to work together as kit.

Edit: Did you try to increase the CPU VCCSA voltage to 1.250 or above?
 

Attachments

Last edited:
Joined
Dec 14, 2013
Messages
1,624 (0.65/day)
Location
Alabama
Processor Ryzen 2700X
Motherboard X470 Tachi Ultimate
Cooling Scythe Big Shuriken 3
Memory C.R.S.
Video Card(s) Radeon VII
Software Win 7
Benchmark Scores Never high enough
@the-ghost That's true, they're not supposed to but reality says different.
I've been around this stuff long enough comparing IC's to have seen it for myself with sticks I have personally.

I have two sets that are identical except for revision, (1.1 and 1.2 sticks) and the chart shows each having a different IC chip BUT that's not entirely accurate.
My 1.1's are supposed to be BH5 and they are, the 1.2's are supposed to be BH6 but they are in fact BH5 sticks. I know that because a heatspreader finally fell off one and the IC's inside were indeed BH5 instead of BH6 like the charts says.
Corsair is very inconsistent at times with this. I even have a set or two that's not in any chart and no telling what they are.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
(For some reason, I did not get a notification about new answers... sorry for "disappearing".)

I've tried raising the "CPU voltage" to 1.350, which seems to help, but when I touch any other voltage in the "CPU voltages" tab, it makes things much worse. So setting them on "Normal" seems to be best.

I've used the "memtest86+" that comes with many Linux installation media; and it's not really "unsupported", it's libre software, and I kind of already had it. But it did not provide enough load to quickly produce errors.

- - -

However, now I've found stressapptest, which does exactly what I need: several CPU-intensive threads doing memory load. This tool catches errors MUCH faster than anything else (in a few seconds or minutes). And so I could confirm that every single one of those sticks is bad even individually. (I did not actually test each one, but I've tested the most stable ones, and also tried it in another system on a different mainboard. And in that one, even the most stable sticks failed almost immediately.)

Now I had to do the only remaining option - get a new kit (on a second hand market). It was also Corsair Vengeance, but 1866 MHz and different version. And with that kit, there was no errors at all. Even on 1866 MHz with XMP (with an i7-3820!). Even though it's actually two 32Gb kits, and their numbering is not sequential either. So apparently the mainboard and the CPU are not to blame.

So now the question is, how could I possibly break every stick in two 8-stick kits. %) The "shop" is a very big tech store for consumer and professional hardware, and those kits came in proper huge 8-stick packaging, so I don't think anything could have been wrong with them. Also, at least four of those sticks were sequential, so definitely from one kit, and now those fail too. The only explanation is that I've managed to damage them. And I'd like to avoid doing the same with a new kit...

I might have left DRAM voltages on "Auto" at some point. And since Gigabyte UEFI seems to be not very good at auto-guessing voltages, that's one thing I can suspect. I've tried taking the heatsinks off one stick (the very worst), and found nothing strange inside. It was not even dirty. Could I have damaged the soldering by pushing sideways on the heatsinks when removing the sticks?..

I'll try those sticks in other systems to see if maybe Micron chips don't play well with some mainboards...
 
Joined
Apr 21, 2007
Messages
29 (0.01/day)
> Corsair Vengeance GMZ64GX3M8A1600C9

1600 MHz, CL 9? Unless it's made from Nanya chips, it has to be overclocked because nobody else makes 1600 MHz DDR3 faster than CL 11.

The best DDR3 is probably the used stuff on eBay that's branded Micron (not Crucial), Samsung, Hynix, SKHynix, Elpida, Dell, Nanya. Avoid registered or anything where the product photo doesn't show a valid bar code number and legible part numbers on the chip packages.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
Hm, this is strange. It's definitely CL9, but it's Micron chips, judging by their "official list".

Could it help to change the timings to CL11? What would the "other" three timings be in that case?

The XMP timings were different from "normal" timings, but I don't remember what they both were...
 
Joined
Dec 14, 2019
Messages
222 (0.70/day)
Location
Loose in space
System Name "The black one in the dining room"
Processor Intel i7 5960X @4.6GHz
Motherboard Asus X99 Deluxe
Cooling Corsair H105 w/4 Silverstone FM121 fans @110cfm ea on the radiator
Memory 64GB G.Skill Ripjaws V DDR4 2600 (8x8)
Video Card(s) XFX R9 Fury X
Storage 960GB MyDigital NVMe. 480GB MyDigital NVMe, 1TB Samsung 860, 4 Western Digital 2TB Black
Display(s) 40" Samsung 8000 series 4K
Case Modded Corsair Carbide 500R
Audio Device(s) Asus Xonar D2X
Power Supply Corsair AX1200i
Mouse Logitech Performance MX
Keyboard Logitech K750 Solar, DiNovo Edge
Software Win 10 Enterprise LTSC
There is the possibility that the mobo has issues. Have you tested the RAM in other boards?
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
I did test it in one other board (Huanan). The results were even worse. On the other hand, this board works just fine with another Vengeance kit (or rather, two 32Gb kits mixed together).
 
Joined
Apr 21, 2007
Messages
29 (0.01/day)
It's definitely CL9, but it's Micron chips, judging by their "official list".

Could it help to change the timings to CL11? What would the "other" three timings be in that case?

The XMP timings were different from "normal" timings, but I don't remember what they both were...
Did you remove the heatsink to read the writing on the chip packages? (if not don't do it -- chips sometimes tear off). Some reviewers have found the writing on the chip packages not matching the SPD even for the slow JEDEC timings.

I looked at a very few Micron DDR3 datasheets and didn't see any 1600 MHz rated CL 9, just CL11 ("-125") and CL10 ("-125E"). Often Crucial/Ballistix or another company's module will be made with Micron branded chips that don't have the speed grade included with the part number, and Crucial couldn't explain what that meant, but I doubt it's good. Ballistix modules have been shipped with those or Spectek chips, which is Micron's subprime recycled brand, and one Ballistix Sport rated 1600 MHz had Spectek -15e 1333 MHz chips on it. Also Spectek's website once had a flash video of modules with "CORSAIR" printed on each chip.

I think it was very common for heatsinked DDR3 modules to be made with 1333 MHz chips, even when the module was rated for 2133 MHz, and I think none were ever made with 1600 MHz chips, so I was surprised when some 1600 MHz Kingston HyperX had them. However the SPD ran the CL11 chips at CL 9 and CL 10.
 

the-ghost

New Member
Joined
Sep 24, 2020
Messages
12 (0.38/day)
I did take off the heatsink on the "worst" stick - to see if maybe there is dirt inside or something like that. It was clean and otherwise normal. Here is what the ships look like - I don't know how to identify them, they don't have the name on them...
 

Attachments

Joined
Jul 9, 2016
Messages
514 (0.33/day)
System Name Main System
Processor i9-10940x
Motherboard MSI X299 Xpower Gaming AC
Cooling Noctua NH-D15
Memory G.Skill 64GB @3600MHz XMP
Video Card(s) EVGA GTX 1080 ti FTW3
Storage 2TB Samsung 970 EVO Plus; 2TB Corsair Force MP600
Display(s) LG 43MU79-B
Case Corsair A540
Audio Device(s) Creative Lab SoundBlaster ZX-R
Power Supply EVGA G2 1300
Mouse Logitech MK550
Keyboard Logitech MK550
Software Windows 10 Pro
Benchmark Scores Cinebench R20 - 6910
You have an original sets that didn't work too well. You got 2 new 32GB sets now working fine. So what is the ongoing problem or is this still a problem?
 
Top