How to upgrade DDR3 ?

[Menard]

Hi

I am testing some RAM kit 2*8 GB for my machine that can support 1866 MHz
Now I see this situation, it appears that some manufacturers pretend to provide 1866 MHz but in fact it is 1600 MHz that could be (sometimes) overclocked to 1833 MHz, and it is the same problem then if you think "ok so I ll send it back and buy then a 1600 MHz one, yes, but NO !!! because the 1600 MHz will be a 1333 Mhz one !!!

So "add more RAM" they said

I am testing a 16 GB kit at 1600 Mhz, versus my old 8 GB (2x4) not kit though, but dual channel too, 1866 MHz RAM !
What I see is a big improvement in the computer user's experience : my 4 browsers are started and I have some 37% of my RAM used
In the tests
- hardinfo tests : some decreases by around 1%
- Passmark Memory : 4% better SQL Database Operations -3% Latency +6%
- Furmark test : -7 % but stays at 8 fps
other simple Gputest 3D tests : no change
- Glmark2 : -4%

Conclusion : no real change to tests and none reflects the improvement of the user's experience

I am note sure that replacing one 4 GB DIMM of the 2, by a 8 GB one but at 1866 GHz, will give better results

But the subject is the in the beginning : we need to add some ram BUT we face some bad behaviour from the manufacturers
(I could tell this for the hard disk storage too)

[rene]

RAM does not itself determine its clock; the memory-controller (i.e., these days and on x86, the CPU) does. Reputable RAM that is sold as DDR3-1866 (PC3-14900) is tested to be compatible with the standard DDR3-1866 clock and cycle-time settings.

Your worse timings on the 16G kit explain lesser performance in the GPU tests: you have an APU and your previous 8G was already enough by far for your e.g. 1G shared-memory framebuffer use of it so you don't gain in size there, and you lose as to 1600 vs 1866. The "user experience" thing is the context in which you are seeing the difference: you simply have double the room.

Do not mix a 4G and 8G stick; dual-channel (which needs same DIMMs) literally doubles theoretical memory bandwidth; the added 4G vs now will again have you experience better "performance" with a under 12G load but demand will increase until it overflows capability again within approximately one week, and at that point you just have slower memory than you had.

8G isn't roomy as of today and needs a computer user with enough idea of things to keep the load down; 16G as of today is roomy enough to allow for lesser such users. Two same sticks will with dual-channel enabled perform better than no dual-channel, and so will DDR3-1833 vs DDR3-1600 (and DDR3-2133 which your A8-7600 should seemingly be able to do will perform even better, as long as it's stable in your specific board). Within whatever speed setting sticks with lower CAS Latency (CL) allowances also do better.

The stuff about lying sellers all out to get you you'd for your own best interests in any case best forget (or just not shop through wish.com or alike).

[Petermint]

"capable of"
Your motherboard might default to 1600 and have an overclock setting for 1833. Look in your BIOS for options.

The RAM will respond within whatever range it can. If the RAM and motherboard support XMP, the RAM supplies a list of speeds to the motherboard then the motherboard selects something. Your BIOS might list the speeds the RAM can work at, based on XMP, or it might list what the motherboard can do.

[Menard]

Yes ok but they don't say that 1866 MHz is only a sort of automatic overclocking ... that's the problem and i think it's false advertising so it is illegal
because I could logically think it was enough to manually set the frequency at 1866 Mhz, what I ve done
Finally I discover that lines in my Bios, XMP, AMP and I enabled AMP ... so after tests it passes, this RAM is very good, Patriot Memory Viper, for 45 €, knowing a single similar Kingston HyperX Fury CL10 too, costs 63 €
But according CPU-z my timings are 923 MHz 11-11-11-28 I think that Intel was tested at CL10 but not AMD and it is CL11 on this platform

[Menard]

For the moment my "user experience" if I have considered it as very good at the begining of the session, after some hours it seems worse for me than before

[rene]

Clearly then you should downgrade to 4G instead of upgrade to 16G. Be sure to let us know how you fare...

[ThaCrip]

I would not really worry about RAM speed as volume of it is far more important in general. you are never really going to notice MHz differences etc as the differences in real world are too small to notice.

I went from... 8GB (2x 4GB) 1333MHz DDR3 RAM (had from May 2012 until 2020) to 16GB (2x 8GB) 1600MHz DDR3 RAM (in 2020. I bought it used for $45) and in general you won't notice much difference in terms of general browsing etc even though the 16GB does give me more room to breathe and certain tasks it helps. I can't upgrade my current system (which I had since May 2012) any further though as it's maxed out RAM wise. but I figure by the time 16GB of RAM becomes a problem, this system will probably be fairly ancient by then as I don't see 16GB of RAM becoming 'too low' anytime soon given 8GB is still good enough for many people.

Code: Select all

sudo dmidecode -t 17

shows the following for me...

Code: Select all

Getting SMBIOS data from sysfs.
SMBIOS 2.6 present.

Handle 0x0052, DMI type 17, 28 bytes
Memory Device
	Array Handle: 0x0053
	Error Information Handle: 0x0056
	Total Width: 64 bits
	Data Width: 64 bits
	Size: 8 GB
	Form Factor: DIMM
	Set: None
	Locator: ChannelA-DIMM0
	Bank Locator: BANK 0
	Type: DDR3
	Type Detail: Synchronous
	Speed: 1600 MT/s
	Manufacturer: Hynix/Hyundai
	Serial Number: XXXXXXXX
	Asset Tag: XXXXXXXXXX
	Part Number: HMT41GU6MFR8C-PB  
	Rank: 2

Handle 0x0057, DMI type 17, 28 bytes
Memory Device
	Array Handle: 0x0053
	Error Information Handle: No Error
	Total Width: 64 bits
	Data Width: 64 bits
	Size: 8 GB
	Form Factor: DIMM
	Set: None
	Locator: ChannelB-DIMM0
	Bank Locator: BANK 2
	Type: DDR3
	Type Detail: Synchronous
	Speed: 1600 MT/s
	Manufacturer: Hynix/Hyundai
	Serial Number: XXXXXXXX
	Asset Tag: XXXXXXXXXX
	Part Number: HMT41GU6MFR8C-PB  
	Rank: 2

p.s. even if the RAM is capable of higher speeds it could be limited by your motherboard etc. but I would not worry about it. main thing is as long as it's stable/reliable. you could run a RAM checking program like https://memtest.org/ (updated Dec 31st 2022) to make sure everything is in good running order. I would do at least one full pass at the minimum, which will take roughly a hour, but if you got the time I would let it run overnight, say 5-10 passes.

[Menard]

Clearly then you should downgrade to 4G instead of upgrade to 16G. Be sure to let us know how you fare...

You post on hardware section but you don't even know that some RAM are not compatible with some motherboards and some CPU.
The manufacturer, Patriot memory, tested it only on Intel platforms and I am on AMD, MSI stopped displaying the document with RAM compatibility for my MB and I had only made a back up of the page displaying the compatibility for Kingston RAMs

[Menard]

In fact, things are more precise today because more precisely with this RAM I get an issue with the mouse click, probably for all applications and on the web pages ... randomly I must click two or more times to get a response, so imagine as it can be boring ... Nemo, LO Calc, Firefox etc
As you say I imagine that perhaps my motherboard has a problem with dealing the mouse clicks with CAS 11 timings (?) AND this happens with only 8% of the RAM in use, as well ! Sorry but I could understand this only this morning ... because I started this session with a very testing mind (!)

About tests, I ve made the Win10 test on this RAM and it answer Passed, I ve made a Memtest 86+ v6.01, passed, and a Memtest86+ v5.31 all the tests and passed, CPU-z gives as timings 11-11-11 as the DIMM sticks are given as 10-11-10 but tested only with Intel platform, my previous Kingston were tested as 10-11-10 on WinXP 32 bits at 933 MHz (1866)

So I wonder what I am going to do, especially this mouse issue seems to have appeared with this RAM but seems also to enworse with the days (only 3 days though) and it was not clearly present on Win 10 yesterday night


NB : before to enable AMP feature in the bios I ve tried to use this RAM as I was doing with my previous one, setting the frequency on 1866 MHz instead of Auto, and I got that one RAM had many errors at the memtest and the other prevented the startup of the bios

[Tosh]

When this (at the time) 8 year old Toshiba Laptop had become very slow running Windows 10 I did what used to make a difference years ago, I added 4gb of ram taking the total to 8 gb. Ram use went from 50% at idle to about 30% at idle but to say I was disappointed is an understatement, seemed to make little difference to the boot up time and the fan was still running at high speed for a lot of the time.

Move on 6 months and I decided to try a new SSD since they were relatively cheap. Bought a 500 gb one from Crucial and cloned it using their free software. What a difference that made, started in no time and all the apps ran with no lag. Ram use was still about 30% but the fan barely ran. Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL. That is why its got Mint Mate 21.1 on it now and it flies.

[ThaCrip]

When this (at the time) 8 year old Toshiba Laptop had become very slow running Windows 10 I did what used to make a difference years ago, I added 4gb of ram taking the total to 8 gb. Ram use went from 50% at idle to about 30% at idle but to say I was disappointed is an understatement, seemed to make little difference to the boot up time and the fan was still running at high speed for a lot of the time.

Move on 6 months and I decided to try a new SSD since they were relatively cheap. Bought a 500 gb one from Crucial and cloned it using their free software. What a difference that made, started in no time and all the apps ran with no lag. Ram use was still about 30% but the fan barely ran. Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL. That is why its got Mint Mate 21.1 on it now and it flies.

I would say the biggest difference for most people RAM wise would probably be upgrading from 4GB to 8GB (but for a light user I might even say 2GB to 4GB would be the biggest jump. but the 4GB to 8GB of RAM claim is a safer claim to make for most people since 8GB of RAM is generally plenty enough for general usage etc). but if a person is a light user, they might not notice it much.

another way to put it... I would say most people should have at least 8GB of RAM, but 16GB of RAM gives you room to breathe. 4GB of RAM is a bare minimum nowadays for a functional computer. because while technically 2GB of RAM will work on Mint, it will be used very quickly after loading up a browser where as with 4GB of RAM you got some RAM left so it won't be used super quickly.

but yeah, RAM is not going to do anything with general boot times/program loading times (although if you really lack RAM then upgrading RAM can make a noticeable difference). a SSD, as you already know, is much better in this regard. but I imagine with laptop HDD's those are especially slower which makes the SSD upgrade all that much better.

Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL.

If you want Windows 10 back on it, there are ways to get it on there for free and it will be activated to. in short, install Windows 7, 'activate' it, then upgrade to Windows 10 using the official Media Creation Tool from Microsoft, after it's upgraded and you confirm it's activated, clean install Win10 from the newest Win10 ISO and your are good. you can get Win10 Pro this way to. then you will be covered until Oct 2025 which is when Microsoft is dropping support of Windows 10.

edit: I may have slightly misunderstood you. are you saying that, for whatever reason, the laptop just won't run ANY versions of Windows 10 newer than a certain older version that's now not supported as I think the last I knew each release of Win10 was supported I think 18 months (maybe it was 24 months) before they required you to upgrade to a newer version/build.

[Tosh]

Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL.

If you want Windows 10 back on it, there are ways to get it on there for free and it will be activated to. in short, install Windows 7, 'activate' it, then upgrade to Windows 10 using the official Media Creation Tool from Microsoft, after it's upgraded and you confirm it's activated, clean install Win10 from the newest Win10 ISO and your are good. you can get Win10 Pro this way to.

That is exactly the the route this laptop has taken. Bought as Windows 7, upgraded to windows 10 (free) and then a free clean install of Win 10 (all Pro versions)

Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL.

edit: I may have slightly misunderstood you. are you saying that, for whatever reason, the laptop just won't run ANY versions of Windows 10 newer than a certain older version that's now not supported as I think the last I knew each release of Win10 was supported I think 18 months (maybe it was 24 months) before they required you to upgrade to a newer version/build.

That is exactly the issue I found I had. The "Feature Update" on the laptop is 21H1 (June 2021 in other words) supported for 18 months i.e December 2022 (think it was the 16th). There was no way I could get the laptop to install a newer Feature Update so that is why I tested many Linux Distros before settling on Mint 21 Mate in August.

The bizarre thing is I have an older Laptop, a Dell Vostro from 2008. That started life as an XP machine, was dual booted with Win 7 (licence bought from E-Bay) then Win 10 (free) with the original HD being cloned onto an SSD at the same time as the Toshiba. That machine has updated to 22H2 and is supported until August 2024, perhaps even longer.

[Menard]

When this (at the time) 8 year old Toshiba Laptop had become very slow running Windows 10 I did what used to make a difference years ago, I added 4gb of ram taking the total to 8 gb. Ram use went from 50% at idle to about 30% at idle but to say I was disappointed is an understatement, seemed to make little difference to the boot up time and the fan was still running at high speed for a lot of the time.

Move on 6 months and I decided to try a new SSD since they were relatively cheap. Bought a 500 gb one from Crucial and cloned it using their free software. What a difference that made, started in no time and all the apps ran with no lag. Ram use was still about 30% but the fan barely ran. Just a shame the laptop seemed locked on a version of Windows 10 that is now EOL. That is why its got Mint Mate 21.1 on it now and it flies.

I questionned myself a lot to chose between to add RAM or a SSD and first the main point was that I could use the SSD with my next platform, as the RAM not, but finally the main point was that it could use much more time to make my system supporting a SSD and to move it from the hard disks, you know with Linux all is more complicated ... and source of trouble, and adding RAM, was seen as harmless in my mind
Add to this that the timing to start the OS or the applications is not an important concern for me, and I was not expecting the RAM to reduce these timings, because it can't do that.

[Menard]

These mouse issue seem to be independant from my new RAM, but this new RAM seems to enworse the problem, I don't know what to do ... because I can still send it back to Amazon for some weeks
I ve reduced the timing of the double click to the minimum possible and it seems to be better but not perfect, this morning a mouse click on a calc sheet cells, fails in around 1 in 10 cases with my old RAM (fails meaning it make a double click as I click only one time)

[Tosh]

Cannot see in this topic if you have tried a different mouse. Could it be as simple as that?

[Petermint]

chose between to add RAM or a SSD

For notebooks and laptops, SSD also has the advantage of not breaking when you bump the device. I would do SSD first if the disk is rotating rust.

For RAM, I would start system monitor and watch the RAM usage as you start applications. If you do not need the extra RAM, return it and use the refund for SSD.

Can you borrow a mouse to eliminate the mouse as the problem? If it is USB, test with other USB ports.

[Menard]

Cannot see in this topic if you have tried a different mouse. Could it be as simple as that?

Not so simple because the problem is unconstant and we can easily think it is solved by one thing or one else, and if we for example try a Live Linux Iso or a Win10 sessions and there is no mouse problem, it seems to be fixed, or when we disconnect the webcam

[Menard]

RAM does not itself determine its clock; the memory-controller (i.e., these days and on x86, the CPU) does. Reputable RAM that is sold as DDR3-1866 (PC3-14900) is tested to be compatible with the standard DDR3-1866 clock and cycle-time settings.

Your worse timings on the 16G kit explain lesser performance in the GPU tests: you have an APU and your previous 8G was already enough by far for your e.g. 1G shared-memory framebuffer use of it so you don't gain in size there, and you lose as to 1600 vs 1866. The "user experience" thing is the context in which you are seeing the difference: you simply have double the room.

Do not mix a 4G and 8G stick; dual-channel (which needs same DIMMs) literally doubles theoretical memory bandwidth; the added 4G vs now will again have you experience better "performance" with a under 12G load but demand will increase until it overflows capability again within approximately one week, and at that point you just have slower memory than you had.

8G isn't roomy as of today and needs a computer user with enough idea of things to keep the load down; 16G as of today is roomy enough to allow for lesser such users. Two same sticks will with dual-channel enabled perform better than no dual-channel, and so will DDR3-1833 vs DDR3-1600 (and DDR3-2133 which your A8-7600 should seemingly be able to do will perform even better, as long as it's stable in your specific board). Within whatever speed setting sticks with lower CAS Latency (CL) allowances also do better.

The stuff about lying sellers all out to get you you'd for your own best interests in any case best forget (or just not shop through wish.com or alike).

Yet when I read about the subject of dual channel, I always read the specialists explain that the visible gain was very limited, that's why I'm quite surprised.
I have also just done a very worrying test by comparing my old memory to the new one but at the same quantity (8 GB) and it is the collapse of performance on many points (-50%) and therefore I will continue the tests.
And about the subject of the effect of 1866 MHz versus 1600 MHz it is here on this forum that I ve read something like "No real difference"

The test I ve just made (Linux MInt) :
Hardinfo
Test FFT (Software Video Processing) : 1.52 > 1.74 (-13%)
GlMark2 : 1747 > 1067 (-39%)
Furmark : 526 > 285 (-46%) or fps 8 > fps 4
Tessmark x64 and Gimark -11%
CPU Mark - Prime numbers : 14 M > 11 (-21%)
CPU Mark - Physics : 268 > 207 (-23%)
Memory Mark - Write test : 4100 > 3443 (-16%)
Memory Mark - memory threaded : 12578 > 10061 (-20%)

Notice that as I don't know how to get the RAM timings on Linux, so I can only supposed that they are identical to those on Windows, so I am testing in fact more CL11 RAM versus CL10 RAM than single RAM stick versus double stick in dual channel, no ? If right I can say, I was not aware that the difference between Cl11 and CL10 was so huge (though not really visible in user's experience)
Real timings in Windows for these 2 RAMs at 1866 MHz : Kingston (10-11-10-30) and Patriot (11-11-11-28)

[rene]

Yet when I read about the subject of dual channel, I always read the specialists explain that the visible gain was very limited, that's why I'm quite surprised.

That is in fact true for "normal use" but as all tests of this nature highly dependent on usage patterns. Caches in modern CPUs are big enough that in normal use including many gaming contexts the difference is going to be anywhere from negligible to limited, but dual-channel literally doubles bandwidth vs single: under a bandwith-dependent load such as streaming large amounts of use-once data through the CPU in the context of e.g. a humongous scientific dataset you notice. And then scaling that best-case for dual-channel back again to more normal uses you see that you still want it; that single-channel is really just wasted potential.

You seem to want easy answers: there are none. All is dependent on usage patterns -- but that's not to say that more and faster memory is not simply more and faster memory. Yes, sure, if you do ridiculous things, then swapping in and out say 16G takes more time than 8G, but it's not in fact forbidden to have/use a brain and not do ridiculous things.

[Menard]

Yet when I read about the subject of dual channel, I always read the specialists explain that the visible gain was very limited, that's why I'm quite surprised.

That is in fact true for "normal use" but as all tests of this nature highly dependent on usage patterns. Caches in modern CPUs are big enough that in normal use including many gaming contexts the difference is going to be anywhere from negligible to limited, but dual-channel literally doubles bandwidth vs single: under a bandwith-dependent load such as streaming large amounts of use-once data through the CPU in the context of e.g. a humongous scientific dataset you notice. And then scaling that best-case for dual-channel back again to more normal uses you see that you still want it; that single-channel is really just wasted potential.

You seem to want easy answers: there are none. All is dependent on usage patterns -- but that's not to say that more and faster memory is not simply more and faster memory. Yes, sure, if you do ridiculous things, then swapping in and out say 16G takes more time than 8G, but it's not in fact forbidden to have/use a brain and not do ridiculous things.

I've found a way to get the timings of my RAM, following this page found by Google https://unix.stackexchange.com/question ... ings-table
I just had to install i2c-tools and then sudo modprobe eprom then decode-dimms


They are the same than on windows (11-11-11-30) not surprising because for a clock it depends of the plateform, not of the OS (I suppose, very probably)

Code: Select all

 decode-dimms
# decode-dimms version $Revision$

Memory Serial Presence Detect Decoder
By Philip Edelbrock, Christian Zuckschwerdt, Burkart Lingner,
Jean Delvare, Trent Piepho and others


Decoding EEPROM: /sys/bus/i2c/drivers/eeprom/0-0053
Guessing DIMM is in                              bank 4

---=== SPD EEPROM Information ===---
EEPROM CRC of bytes 0-116                        OK (0xBFA2)
# of bytes written to SDRAM EEPROM               176
Total number of bytes in EEPROM                  256
Fundamental Memory type                          DDR3 SDRAM
Module Type                                      UDIMM

---=== Memory Characteristics ===---
Maximum module speed                             1600 MHz (PC3-12800)
Size                                             8192 MB
Banks x Rows x Columns x Bits                    8 x 16 x 10 x 64
Ranks                                            2
SDRAM Device Width                               8 bits
Bus Width Extension                              0 bits
tCL-tRCD-tRP-tRAS                                9-9-9-24
Supported CAS Latencies (tCL)                    9T, 8T, 7T, 6T, 5T

---=== Timings at Standard Speeds ===---
tCL-tRCD-tRP-tRAS as DDR3-1600                   9-9-9-24
tCL-tRCD-tRP-tRAS as DDR3-1333                   8-8-8-20
tCL-tRCD-tRP-tRAS as DDR3-1066                   6-6-6-16
tCL-tRCD-tRP-tRAS as DDR3-800                    5-5-5-12

---=== Timing Parameters ===---
Minimum Cycle Time (tCK)                         1.250 ns
Minimum CAS Latency Time (tAA)                   11.250 ns
Minimum Write Recovery time (tWR)                15.000 ns
Minimum RAS# to CAS# Delay (tRCD)                11.250 ns
Minimum Row Active to Row Active Delay (tRRD)    6.000 ns
Minimum Row Precharge Delay (tRP)                11.250 ns
Minimum Active to Precharge Delay (tRAS)         30.000 ns
Minimum Active to Auto-Refresh Delay (tRC)       48.125 ns
Minimum Recovery Delay (tRFC)                    300.000 ns
Minimum Write to Read CMD Delay (tWTR)           7.500 ns
Minimum Read to Pre-charge CMD Delay (tRTP)      7.500 ns
Minimum Four Activate Window Delay (tFAW)        30.000 ns

---=== Optional Features ===---
Operable voltages                                1.5V
RZQ/6 supported?                                 Yes
RZQ/7 supported?                                 Yes
DLL-Off Mode supported?                          Yes
Operating temperature range                      0-95 degrees C
Refresh Rate in extended temp range              2X
Auto Self-Refresh?                               No
On-Die Thermal Sensor readout?                   No
Partial Array Self-Refresh?                      No
Module Thermal Sensor                            No
SDRAM Device Type                                Standard Monolithic

---=== Physical Characteristics ===---
Module Height                                    30 mm
Module Thickness                                 2 mm front, 2 mm back
Module Width                                     133.35 mm
Module Reference Card                            B revision 1
Rank 1 Mapping                                   Mirrored

---=== Manufacturer Data ===---
Module Manufacturer                              Patriot Memory
Manufacturing Location Code                      0x02
Manufacturing Date                               2022-W43
Assembly Serial Number                           0x102404DA
Part Number                                      1866 CL10 Series


Number of SDRAM DIMMs detected and decoded: 1