Hardware upgrade: RAM DDR3 19200 (2400 MHz)

I have AMD M/B which supports 2000 MHz RAM(OC). I already have 8GBx2 14900 (1866 Mhz) (Kingston Hyper X Fury)RAM in dual channel mode. I want to install another 8GB x2 19200(2400 MHz DDR RAM (Corsair Vengeance Pro) in dual channel mode for total of 32 GB RAM (supported by M/B). Will there be any conflict? I don't intend to OC, but run both at 1866 Mhz. Windows 10 Pro. Gigabyte GA-970A-DS3P  M/B.
Who is Participating?

[Product update] Infrastructure Analysis Tool is now available with Business Accounts.Learn More

I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

David Johnson, CD, MVPOwnerCommented:
the usual way to do this properly would have one set of ram in slots 1 and 3 and the other in 2 and 4. It should run just fine at the slowest speed.

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
Brian MurphyIT ArchitectCommented:
Apologies if I missed this but by support do you mean the bus speed between RAM to Processor supports that as well?

Because this becomes a big issue in the server world.  It can actually make the server slower to add RAM above the threshold where the manufacture states it can handle this bus speed and at 256 or higher you take a 30% hit.  Cisco UCS is the only one I know of that you can exceed that threshold and not take a hit.

And I've experienced the difference.  On my Citrix servers I could get more users on Cisco UCS with the same memory if I added more memory where the manufacturer states that if you go from 32 GB to 64 (just an example) the trade-off is 30% reduction at the bus.  

So, adding more RAM caused a traffic jam.  On that same server, adding more RAM caused a decline in session count.  

And, it was was the # of DIMMS inserted.  So up to 2 DIMM slots most systems you get 1600MHZ but if you install a third one it drops to 1066 MHZ.  Doesn't matter if AMD or Intel.  

But, I don't deal with that much now.  My bottleneck was the drives until flash arrays arrived.

If you already use Solid State drive your still limited by how fast the cable can handle send and receive.

That is why adding a SCSI card with RAM Cache on a workstation class machine is faster than Solid State because a SCSI Card does both IO operations at the same time.  You get a dual port SCSI card with even 8 GB of RAM cache and hardware controller module and 10K RPM drive will outperform a solid state drive off the mother board.  

Same principle as GPU.  The Video Card is far superior relative to the port on the board.  That is why having two works better but not that much better.  You add one to get the extra channel but that second one causes a reduction of 20% of the maximum if running at 100%.  So you get two cards running at 80% is 160% which is better than 100% a 1 card.

But that is the fun in it, maximizing performance by maximizing the bus speed, adding another DIMM might not compare to adding a SCSI Card with 32GB of RAM and hardware chip to write and read at the same time.  Right now that is write or read.  Typical mother board is still one way at a time regardless of SSD or FLASH until you take that and put it in a flash array with two controller cards that has 4 fiber connections split between two fiber switches and active\active.  

That has been my experience.  I build my own PC's and have since the 286 chip was released.  I just don't game any longer so I've got one I built that has lasted 7 years and it is like a side project to keep it living.  It is basically a laptop but no case, no covering at all, took off the laptop monitor and keyboard and glued it to a block of wood and took out the laptop drive and have that running on a standard size drive that sits on a heatsink I got from a Harmon Kardon unit that I tore apart and parts it out.  Then the drive sit on the Heatsink and I have a Fan that is getting power from a powercell used for solar energy.  It is a work of art.  

I love it
i suggest to have the same model of ram for all; only then will it run in dual channel mode
so either the 8GBx4 14900 (1866 Mhz) (Kingston Hyper X Fury

or 8GB x4 19200(2400 MHz DDR RAM

**some boards even refuse to work with different brands/models
The 7 Worst Nightmares of a Sysadmin

Fear not! To defend your business’ IT systems we’re going to shine a light on the seven most sinister terrors that haunt sysadmins. That way you can be sure there’s nothing in your stack waiting to go bump in the night.

Philip ElderTechnical Architect - HA/Compute/StorageCommented:
Both channel's slot 0 is populated.

Install the new memory in the next slot down on each channel. Boot.

It's a lottery as far as two different manufacturer's memory products playing nicely together in the same board.
VakilsDeveloperAuthor Commented:
Thanks for your responses. @Brian, thanks for sharing your exploits. I too build my own PC, started with AMD 386sx CPU(Never bought Intel), It is more like an assembling job these days. Your laptop building project was fun to read. I had that in mind but could not come across any laptop kits, perhaps you can guide me there. And building a Mac is also in the bucket list.
Now to main question:
I do have all 4 slots occupied, which I did not mention before.
One dual channel is 16gb (8gbx2) Kingston HyperX Fury
2nd dual channel channel , same memory but 8GB (4GBx2)
So total 24 GB.
The motherboard manual says that:
Support for DDR3 2000(O.C.)/1866/1600/1333/1066 MHz memory modules
* To support a DDR3 1866 MHz (and above) memory, you must install an AM3+ CPU first.
Since I have AMD FX-6300, it is supported.
As stated in specs: http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series%20FX-6300.html

Memory controller      The number of controllers: 1
Memory channels: 2
Supported memory: DDR3-1866
Maximum memory bandwidth (GB/s): 29.9
Bus speed        One 2600 MHz 16-bit HyperTransport link (What is this?)
On running CPU-Z, it showed FSB 200Mhz, DRAM 667 MHZ ( I think there might be multiplier) and CPU clock at 3.5MHZ, I think CPU directly controls memory unlike old architecture. The DRAM to FSB ratio is 3:10.  What do all these figures translate too? Is your obervation more memory, slow performance still valid? Because I have not observed that going from 8GB to 24 GB, nor I could find any such threads.
I agree with you all that M/B is picky about memory type and adjust with slowest memory. I don't think I am going to get any perceptible performance change going from 24 to 32GB. So, I might leave as is. But I am curious about what you make of above specs.
VakilsDeveloperAuthor Commented:
Oh, and I have O/S on SSD drive ( Samsung EVO 850 500 GB)
David Johnson, CD, MVPOwnerCommented:
the hard drive is immaterial in this discussion
Brian MurphyIT ArchitectCommented:
Good questions but difficult to answer without writing a book.  And I'm not looking for a debate, where it pertains to performance everything on the motherboard is relative.  The company I contract for at present is 2nd or 3rd in the world for MIPS.  

That is simply an acronym that refers to a way we measure instructions per second.  It is an older term and simply means "million instructions per second".  And this, is highly impacted by IO or Hard Drive.

To dispel the first myth, your hardware is only as good or great as the software and drivers.  If you are gaming or using CAD you most likely bottleneck is the hard drive.  Anyone that has read Intel floating point papers or anything on MIPS exposes the logic behind MIPS as the measurement when applications rely heavily on the hard drive.  

MIPS is still a measure when your dealing with IBM Super Computers that process claims for Medicare and Military.  The building I work in requires DoD TS Security Clearance.  When we talk about Mhz, that is the clocking frequency.  So, a CPU performance can deviate substantially on the type of instruction given which is the application.  And, how well that application interacts with any OS from Unix, BSD, Linux flavors, Microsoft and others.  I have solutions on BSD and Microsoft Windows.  Some of the Unix stuff goes back to Cobal.  Every windows OS still uses the concept of paging file created (designed) when the goal was to run multiple applications in a single instance operating system.  

Anyone that does any gaming online knows this already the fact that your hard drive light means that application is caching on disk not RAM and every process you see is swapping.

Maybe I'm the only one but the questions your asking were similar to mine 20 years ago when Ieft Finance for Technology.  Have you ever wondered exactly how Windows is assigning physical memory, how much file data is cached in RAM, or how much RAM is used by the kernel and device drivers?

So, some of this information is just required.  We have 5th amendment, and my first 10 years was a study to get facts.  If someone makes a subjective comment making claims that IO has no bearing it should at least require some fact or reference.  When I first started out I might get 10 different answers from 10 different people until I just learned they don't understand.  It is easier to rule it with one sentence but if I don't get facts I question motive.  So I've spent 20 years working 90 hour weeks to be a polymath and scholar of technology.

So, if you are gaming.  The easiest thing to start with is IO, and Windows.  The design goals for Windows I/O is the abstraction of devices and both physical and software.  Something has to manage or mediate the world between physical, logical, and virtual devices.  For every device, you have a device driver and those exist on a physical hard drive day 1.  All those drivers must interpret high-level commands such as "read" and "write" and issue the low-level command to the device using a entire different language and unless you are running ESXi and Virtual Workstations you need an interpreter  (in most cases).  That I/O Manager is stuck in the middle and doing this for 100's of drivers where low-level commands to the driver go through the manager and the driver issues commands back and the manager is telling the driver this got done, that got done.

So, Plug and Play is driving a bus, bus driver, to allocate drivers to hardware and tracks detection of new and removal of old.  If the driver you added is wrong he escalates to his executive to reassign that to another manager we call  user mode PnP manager.  And this is maybe 1% of what I can share now without another 100 pages.  So this is the OS, I/O Manager and the kernel mode drivers, user-mode drivers, API drivers, Services drivers so everyone takes a hit on this across the board.  Every driver can be a weak link and this starts in software that resides on a solid state drive, in your case.

So real developers are looking at this first.  Right here, it begins.  How do I make this application work on this hardware with this OS and where the OS remains the same but you need a system design that maximizes your application performance.  

Same game, other hardware, the guy with the SCSI controller and cache wins every time due to the nature of read and write in parallel.  You have a solid state but a traffic light where traffic can only go this way or that way but never the same time.  Hard drive.

Gigantic bottleneck until we port OS code in physical RAM everything starts and ends on that hard drive.  Not to mention files.  That is another discussion.

Most I/O operations that applications issue are synchronous.  So, take one thread from APP.exe and a simple readfile function.  

The application thread waits while the device performs the data operation and returns a status code when the I/O is complete. The program can then continue and access the transferred data immediately. When used in their simplest form, the Windows ReadFile and WriteFile functions are executed synchronously. They complete the I/O operation before returning control to the caller.

Then you have your stacks, caching, buffers and you have those nasty scenarios where someone used user-mode versus capturing all input into kernel mode and buffer.

To sum it up here, all Windows file system drivers that manage disk-based file systems are part of a stack of drivers that is at least three layers deep: the file system driver sits at the top, a volume manager in the middle, and a disk driver at the bottom. In addition, any number of filter drivers can be interspersed above and below these drivers.

This is where threading is critical for application development.  Too few or too many you have performance problems.  Do I use a single thread for all processes or 10 threads for ever 1 process.  This is where context switching is excessive or you end up with thread thrashing.  If you have more than one processor, the scheduler may or may not divide those properly.  

We have not made it to memory yet.  Some DIMMS do not perform as well as other, just the facts.

Just like I/O, Windows has a memory manager.  A gaming machine on Windows is not talking to the physical hardware but instead a lot of representatives for the hardware.

Just start Task Manager, and you get a glimpse of it and if you open perfmon there are countless ways to measure memory, disk and so forth.

This is where we start getting into System Commit and the pagefile.  And the addressable range is based on 32 or 64 bit compliant hardware and OS, and the application running in 32 bit you might take a hit.  I just depends.

I just wanted to start with the basics, we can take this as deep in the weeds as you want.

The questions your asking are easy to find.  There are plenty of tools that will show you what is really happening with your memory.  

You probably won't know unless you take a baseline now, look at the fundamentals which is what I started with and build from there so there is complete understanding where all the bottlenecks are and they are everywhere.  All we can do, is learn that information, knowledge is power.

And if anyone says this or that is not on the table, look at that first.
Brian MurphyIT ArchitectCommented:
If you would like to jump start and start looking at I/O download the free Microsoft Tool

This next tool will show you a little higher view, you can look at any process and you will be amazed the dll's attached to that process.  Hint, start look at "Reserved Memory"

If you want to see your hard drive in action, this tool runs in the system tray

This tool is used by developers to maximize memory use.  You can run it as is or have it run a process like any .EXE to start a game or what not and tell you everything it did in RAM

This one will tell you mapping between logical processors and the physical processor, NUMA node, and socket on which they reside, as well as the cache’s assigned to each logical processor.

This one, will show you what the Memory Manager is doing.  I saved this one for last because it will show you what is in memory and tell you how much is cached to your hard drive.   I'm not aware of another tool paid or free that will show you how one of your I/O Managers reacts to instructions given by every application.

On this one, use the memory snaps.  

All free, take your baseline with 1 DIMM, 2 DIMM, 3 DIMM - you might find, the application doesn't care either way.  And if it pulling reserve memory tactics and using a 10th of that you don't want to run that app with another at the same time.

This tool takes about a month to learn and study how your software reacts for your hardware.

That is the point.  You may find that process threads or processor and not memory or a large cache file like every online game does to update the application with new zones or new challenges is downloaded to file cache and your playing the game and your hard drive cannot keep up.

With an ESXi server I can spin up a Windows Virtual machine where each server blade has 768 GB of RAM, Four processors (4 slots) with 12 core each and blow any physical machine away because that Virtual Machine is running on this storage my performance issues then truly do become the processor and memory again.  If I turn up a 60 GB Drive, Win 7 64 Bit allocation the VMWare tools and drivers talk direct to the hardware.  I'm getting 300,000 IOPS.

You are off the motherboard now and this is Cisco UCS blades with XtremeIO Flash Array.

Why do we do this?  Because the storage was the bottleneck.  And that is with HBA fiber cards split between switches back to a standard EMC style storage having numerous controllers and so forth we have an application that every report is a 1GB PDF file.  

The goal is knowledge and understanding how that application works.  You need the facts, what happens with no app, what happens when I kick this off and you will see quickly if it is processor or memory and it is the application causing those things. Now you have to design and build your system to compensate in areas where this application causes problems.

Anything else is a guess.  You can add more and more and maybe it is faster.  Or, you want to know what it is slow or fast.  If you get insane page faults, you start looking at RAM and Hard Drive.

Why am I getting these, the tools I provided will give you a chance to see what is happening at the IO level, every driver, service and so forth.  You might find one service running that if you turn it off and disable it you get 20% better performance due to how that service managed RAM and it caused the application to become competition for processor resources.

And, it's just fun.
...lots of reading for ram upgrade..interesting
VakilsDeveloperAuthor Commented:
All that still doesn't answer a simple question...
David Johnson, CD, MVPOwnerCommented:
the answer is that it should just work. I have mixed speed ram in my i7-2600K and it works just dandy
in many cases, i agree it works
but not always, and not in dual channel
VakilsDeveloperAuthor Commented:
Thanks, All.
It was interesting conversation. Though, I am not upgrading any server class machine, just a personal computer with some gaming, it was interesting to know what's going on. Based on that I also did some additional reading. SCSI interface is going serial like SATA, I thought that was dead 15 yrs ago. China has world's fastest computer, followed by IBM Blue/Gene. The insurance companies still use mainframe (to handle millions of transactions) and no blade server can replace that, and computer on board space crafts have fraction of processing capabilities compared to a home computer. There's no real parallel processing, it's  difficult to have such algorithm. And then there's quantum computer...
I am going to keep system as is. Already have 24 GB. May invest on graphics card.
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today

From novice to tech pro — start learning today.