I retired my old PC the other day and replaced it with a spiffy new (to me) £140 refurbished machine with Windows 10. Have to say I’m really impressed with its speed compared to my clunky old one.
It’s come with 8GB of RAM on a single stick. There’s a spare slot and I have two 4GB sticks of ram left over from my old PC.
There’s one spare memory expansion slot on the PC so I’m wondering If I have anything to gain (or lose) by sticking in one of my 4GB sticks.
I do vaguely recall that this is far from optimal but would it help/ hinder my PC?
To be honest, it seems to work fine at the moment but I’m just wondering if I can make use of my older memory sticks.
Check memory type and speed. The machine can only operate the RAM, at the slowest of the two memory speeds. If one is CAS5, one is CAS6, it will run both at CAS6 (slower).
And OEM machines run stuff at "slow JEDEC rate", rather than "overclocker rate". Plugging my 2400 RAM into the Dell Optiplex here, it runs 1333. Boo and hiss. Only an Alienware would have XMP.
The memory controllers on modern kit, are of the "flex memory" type.
+------+ --- | | This section is single-channel speed on bandwidth | 8GB | +------+ --- | | | 4GB | This section is dual-channel speed on bandwidth +------+ +------+ ---
I measured this effect years ago, using a modified copy of memtest. On an NForce2.
The dual channel section was 1400MB/sec, the single channel section was 800MB/sec . It's not an exact doubling. But pretty close.
Modern Intel processors have excellent caching, and only the "cache misses" head out to the memory subsystem. You can hardly tell the difference between the slow and the fast part. AMD CPUs were a little more sensitive to memory subsystem.
In the old days, with old enough equipment, you would not even attempt that (mis-matched RAM back then, sucked). On modern equipment, it is "meh", plug it in and away you go. Plenty fast enough.
In any case, you really should use memtest anyway, as you won't know if some memory has developed a defect over time, unless you check. The download files nearer the bottom of the page, should be really tiny, as the standalone memtest has no OS and just launches the code directly. While they call it an ISO, it's not 2GB. It's tiny.
formatting link
*******
And if you do this, this is excellent.
+------+ +------+ --- | | | 4GB | This section is dual-channel speed on bandwidth | 8GB | +------+ --- | | | 4GB | This section is dual-channel speed on bandwidth +------+ +------+ ---
With Flex Memory, you can run three sticks. Two sticks go on one channel, the other (8GB) stick goes on the second channel. Consult your manual, as to which slot goes to which channel.
That's every bit as good as this. One stick per channel.
+------+ +------+ --- | | | | This section is dual-channel speed on bandwidth | 8GB | | 8GB | | | | | +------+ +------+ ---
Not all machines have four slots. Some desktops have four slots, and allow this config. Some cheaper desktops only have two slots.
A laptop with SODIMM, might only have two slots or one slot, and then the first diagram I drew, is your option in the two slot case.
There is a local store here, with "exotic" laptops, where they have four SODIMM slots. That does not happen too often.
Some computers solder down a rank of RAM, and then you're pretty well stuck with some kind of asymmetric config. Which is fine.
Plug in something, and have fun. It's not like the old days. It's better.
Don't you have to worry about the voltage that memory banks need to run at ?. I always thought that you should only use identical memory boards if using 2 or 4 ?.
No don't mix them, they may have different parity. 8GB memory should be enough to give you all you need, my general purpose laptop has 8GB and performs video editing fine and multi-tasks easily. I have two other laptops with 32GB RAM for microwave circuit simulation.
Also a lot depends on the processor, but 8GB should be enough.
DDR3-2400 CL10-12-12-31 1.65V <=== Flip the XMP BIOS switch on the X79 machine 1333 1.50V <=== What my Optiplex would feed it The Optiplex has no DRAM control options.
The RAM is designed to "come up" with ordinary applied voltage. As otherwise, you would not be able to navigate the BIOS menu to the XMP tick box, to click it into high gear.
This means it should really be tested at two operating points. It should be screened by hand-tester, at JEDEC conditions. (The company does not want to waste time doing extended testing on RAM that won't even pass 1333.)
Then it is moved over to the motherboards bench, for testing at 2400. Some companies do the screening for the XMP rate, on a motherboard, rather than a hand tester. They may claim "tested on an Intel system" on the package.
That means, I can safely mix the Dell original DIMM, with my fancy RAM, and it's because the stuff is only going to run at 1333.
The 1.65V limit is established via "damage to processor". There was a range of processors where damage would result, outside of 1.65V. And this is why they just push them to 1.65V, on the package labeling.
That RAM can run 2400 at 1.5V. It's really amazing stuff, and it is second generation chips. When the manufacturer screens, they actually expect 100% success. Whereas some RAM in the past, there was significant fallout at the high rates tested.
But I've never abused that RAM. There's really no point abusing it. You would need a microscope to see the difference abuse makes.
*******
There were some RAMs in the past, that sticking to the higher voltage on the package was a good idea. These were RAMs that were stable at only one point, or had only been tested at one point.
Sometimes you just have to adjust the settings, to get stuff to work. Just because the SPD has timings, doesn't mean the DIMM has been properly tested. There have been some RAMs that shipped with only
5MHz of margin (lift the speed by 5MHz above stock, they throw errors). As long as you can keep the BIOS alive long enough to change tRCD or the voltage, you can probably tune it for better operation. A Dell is unlikely to have a DRAM menu for this tuning process.
Since the OP did not make any "enthusiast noises", my assumption was the DIMMs are all JEDEC rates.
By all means, if buying new RAM, you should match them. That would give a higher resale value later, if you want. What I'm typing on has four identical sticks. But if you're just taking sticks out of the junk room and plugging in, and they're all JEDEC ones (boring speeds), then they should work OK on your Dell. All you can do on a Dell, is just memtest, and see if this was a good idea or not.
If my wonder-RAM did not work, then just unplug and move on. On the Dell, there's no way to make it go faster than 1333, so I'm kinda stuck with the boring speeds.
The slots are keyed. The slots have shapes and dimensions, to prevent mixing wrong memory types.
Some slots are keyed for both UDIMM and RDIMM.
If you install one UDIMM and one RDIMM, the BIOS will beep and will not turn on the power to the RAM or do anything else with the RAM.
If you install two UDIMM, it works.
If you install two RDIMM, it works.
You just can't mix RAM types, on a board keyed for two memory types. You use all of one type, or all of the other type.
For parity, you could plug in an eight chip DIMM and a nine chip (parity) DIMM -- but the BIOS will beep for that.
If the BIOS does not want you using parity, it can beep.
If the parity chip "flaps in the breeze" and is not wired to the bus all the way to the "other end", it will not receive the "special cycle" and will not have been informed as to what the correct CAS cycle count is. It cannot open its little shop, unless it knows the value of CAS. Consequently, it will neuter itself. There is a protocol for bringup of RAM, during POST. And more modern RAM types now have bus calibration processes, so there are even more things that need the twist of a screwdriver... before the RAM chip will do anything. CAS was the only thing the older memories needed (sync memories). On the FPM and EDO, CAS was set by the usage of delay lines.
Once you have checked that the memory is suitable*, and fast enough and fits (none of which are certain) by all means try it. But memory specs seem to change every few years, so it's very possible that wont work.
Another new 8GB stick is around £30 ...
BUT. unless you are doing serious video editing or massive graphics work, its highly unlikely that any extra RAM will result in any noticeable performance increase.
Especially if you have fitted an SSD.
*DDR1,2,3,4&5 are all different. and so are rated speeds. Some memory has ECC error correction. And then laptops take SODIMM instead of DIMM
Memory is used in two ways. To hold large amounts of data on the computer itself, and to speed up disk access by caching some of the disk in memory
Dealing with disk accesses first the use of a solid state drive will already have speed up disk access massively, and unless you are running code that thrashes the disk, adding more RAM wont actually make a huge difference, if any. So, where else is RAM important? When you have huge data objects to manipulate, For example a bit map to transcode. So graphics and video editing of large objects needs RAM. Or many many users on your system. That needs RAM as well. Or running a thousand browser tabs.
Memory is something that doesnt make the computer faster, but lack of it makes it run a lot slower. On this 8GB machine: $ free -m total used free shared buff/cache available Mem: 7837 2494 2101 293 3241 4750 Swap: 2047 16 2031
What that is saying is that its only using about 3GB of RAM, the rest is just tossed into the buffer/cache pool
So *in my current usage*, 8GB is well better than 4GB, which is marginal, but more RAM wouldn't make anything better at all.
Generally there is no issue these days, but I have had issues with slightly different speeds or even different makers sticks not playing well together.
Its suck it and see really. Its a great pity nobody makes an adaptor to turn two smaller sticks into one bigger one, but then its probably not worth it considering the prices. Brian
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.