Monthly Archives: August 2016
So now you know how to find out which processors are compatible with your system after reading my General CPU Compatibility Guide, but which one is the right one for your usage? Which option will give you the best value and what will work the fastest for you? Which one is going to turn your computer into a space heater and what is going to save you money on your energy bill? There are lots of things to consider when choosing a processor and in many cases, it isn't as simple as buying the most expensive one you can afford. I will guide you through this consideration process and help you make a smart decision
Some people will disagree with this, but I think the first thing to consider about your budget is that it is largely irrelevant. Don't come in with the mindset that you have $300 to spend. Instead, evaluate the task you are doing and try to get the processor that will do the best job for you for the least amount of money. What I mean by this is that if you are just doing standard office tasks like web browsing, excel or word then there is very little to gain from having a $300 dollar processor. You would have your needs met all the same for under $100. On the flipside, a $1000 processor may seem absurd to some. That is more than many people's computers cost now. But if you are someone doing hard rendering work or video encoding or some other intensive task and buying a $1000 processor instead of a $300 processor is going to save you 30 minutes a day and you value your working time at $25 an hour, you will earn that value back in less than a quarter of the year. This is why I tell most people who have a number in their head to throw it out the window. Adhering to an arbitrary budget will likely make you spend too much or too little. The important thing is to get your need met in the most efficient way possible.
In my opinion evaluating your usage is the most important thing to look at when deciding which processor to buy. What I mean by usage is very simply just what you do throughout the day or what you plan to do throughout the day on this computer. Is it a box just for streaming video? Is it a video editing station? Is it a 4k video game rig? Does it need to host piles of virtual machines? Whatever it is doing, consider the software that will be run. Is it something that functions in a good-enough fashion? It either does the job correctly or doesn't. Or is it a speed-benefitting task? Would a faster processor save you time or give you better performance in the task? If, for example, you are playing video games, after a certain point there is no benefit to a faster processor. You have already reached an acceptable frames per second and the game will be enjoyable at the resolution you are using. But if your game is just on the cusp of being playable it may be very worth it to go for the higher clocked processor.
So now you have considered what the machine will be used for and determined whether you really need speed or just something that is able to run things correctly. The next thing to consider is are the programs you use properly multithreaded. Multithreading is when a program uses more than one core of the processor. For a long time, games were only using two or in sometimes four cores, but we are slowly moving beyond that. No matter what your task, it is valuable to look around the forums for that software and try to find out if it would even use the extra cores if you are deciding between two different core count options. Keep in mind that your machine will be able to do other concurrent tasks better if you have more cores, but you may not get any benefit from some of the apps you are using simply because you have more cores. If the biggest thing you are worried about is encoding a video using some software and you see that that software doesn't seem to get a proportional gain to the increased core count then it may be better to look at a different option or just go for the lower core count model.
One of the most common determiner's of speed in days old was GHz. This used to be a pure metric on how fast the processor would be, but now we have so many different models and architectures and manufacturing nodes that you can't simply say that a 3.2GHz chip beats a 2.8GHz chip. Instead, now, a user must simply look at benchmarks and see how an individual processor performs a task because the GHz number does not mean a whole lot. If a processor is of the same generation and the same series and has the same amount of cache (this is basically really fast on-processor RAM) then you may look at GHz as an indicator, but for the beyond that it is really best to simply view benchmarks. Anandtech and Tomshardware are great places to start with these. Processors these days simply have too high of a range of IPCs(Instructions per cycle, basically how fast a processor is in reference to it's GHz). That being said, there is still plenty of room to optimize purely on GHz. Like I said previously, a software may not benefit from increased core count, but it likely does benefit from an increased clock speed. In certain tasks, like gaming, that is why a $300 6700k with 4 cores will almost always outperform the much more expensive 10 core $1700 6950X. The games cannot efficiently take advantage of the additional threads, but they can take advantage of the higher clock speed.
TDP (Thermal Design Power)
TDP or thermal design power is, for our purposes, at least, the amount of power that the processor uses and heat it puts out. This something else to consider when picking your processor. This number has slowly crept down over the years as manufacturers have made more efficient processors, but it is still not insignificant. If you are in a small enclosed room you can definitely increase the ambient temperature by a few degrees if you have a high wattage processor and you run it for full tilt for a while. In many places, the electricity cost is negligible even across an entire year of usage, but it does add up in places with high electricity costs. If you want to see a quick estimate of what a difference between two different wattages of processors will cost you can visit this site and enter in the different wattages of the processors you are considering and your electricity cost and your estimated usage. One other thing to consider is that if you are going to run the AC to counteract the increased temperatures in your room you will actually be paying double for the increased wattage. If this is something you are concerned about then you might want to consider a processor that ends with "T" in the model name. These are Intel's low wattage variants and they have some chips there that perform just as well as their higher wattage siblings.
Overall, what I want you to take from this article is that there are a lot of things to look at with processors, but ultimately it comes down to meeting your needs in the most efficient way possible. Don't come into this process with a set number in mind and instead try to find the correct tool for the job. Figure out what the expected usage is and whether or not the applications you use would benefit from more cores or if they really need higher clock speeds or higher IPC. Then consider if you are in an environment where TDP is important to you or if there is a variant that will save you money on electricity and AC costs while still meeting your needs. If you follow this framework you will likely find you can make a better decision.
RAM Type: DDR3 RDIMMs, LRDIMMs
Max Ram Capacity: RDIMMs: 256GB(16GB DIMMs), LRDIMMs: 512GB (32GB DIMMs)
Processor Socket Count: 2
Compatible Processor Series: E5-2600 V1, E5-1600 V1
Number of Ram Slots: 16
Hard Drive Bays: 4 or 8 2.5" or 3.5" drives.
BIOS Notes: Revision A07 fixed several major issues including:
Non-bootable SAS drive/RAID when LSI 9265 is installed - Hang during POST when two K5000 cards are installed. 5. - RAID ONE BSOD issue.
Revision A04 Fixed an issue with the diagnostic LEDs
The Dell Precision T7600 is a workstation featuring an EATX motherboard with dual LGA 2011 Sockets. These only currently support the E5-2600 first generation processors, but it has the same chipset as the T7610 which received support for the second generation processors so it seems they may soon get support if Dell puts out a BIOS update. The system has 16 DDR3 slots with support for up to 16GB RDIMMs or 32GB LRDIMMs. The T7610 received support for 64GB LRDIMMs so it seems that if the T7600 does get the update for the second generation E5 processors then the 64GB LRDIMM support may come with that update as well. The workstation features 4 hot-swappable bays that can easily accommodate either 3.5" or 2.5" drives of either SAS/SATA. The system comes standard with a PERC H310 RAID controller to handle the hardware RAID configuration.
The Precision T7600 has 5 PCIe slots available if you are only utilizing a one-processor configuration but adds two more PCIe slots if you add in a second processor. All slots are full-height, full-width, but the first and the third slot under the processors only operate at x4 speed instead of the full x16. The workstation features dual gigabit networking ports with Intel adapters: 82574L and 82579LM. The maximum capacity of a single processor configuration with RDIMMs is 128GB and with LRDIMMs 256GB. In a dual processor configuration, the maximum addressable capacity is 256GB with RDIMMs and 512GB with LRDIMMs. Maximum single stick capacity is 16GB RDIMMs or 32GB LRDIMMs.
The Precision T7600 has a few known issues that have been addressed with BIOS updates:
Revision A07 fixed several major issues including:
- Non-bootable SAS drive/RAID when LSI 9265 is installed - Hang during POST when two K5000 cards are installed. 5. - RAID ONE BSOD issue.
Revision A04 Fixed an issue with the diagnostic LEDs
For additional info on this product visit https://www.itconnected.tech/support/dell/dell-precision-t7600.html
What to look at to make sure Memory will be compatible
Memory has a bad rep for sketchy compatibility and random quirks, but in recent years it has actually become very simple because there haven't been many major changes. In this guide, I will explain to you what the various types are, what the various numbers on your memory sticks mean and what will likely be compatible with your system.
ECC Non-ECC FB-DIMM SO-DIMM RDIMM LRDIMM DDR3 DDR4 DDR2
There are a ton of acronyms and names for all of the different kinds of memory so I am going to spend a little bit of time educating you on the different types and then show you that most likely if you are reading this guide that it is very simple and there isn't really anything you can get wrong.
The first distinction is what kind of system are you putting the memory in?
If it is a laptop then almost 100% chance it is the SO-DIMM form factor. These are smaller sticks designed to fit in the smaller body of a laptop. They are roughly half the size of a desktop stick of RAM so they are easy to differentiate.
If it is a desktop then it will almost always be the standard DIMM form factor. In some cases of very small form factor systems, you may see an SO-DIMM used, but rule of thumb is that desktops use the DIMM form factor.
So now that you know about the sizes of memory I can tell you about the types of memory. There is DDR, DDR2, DDR3 and DDR4 all available in both laptop (SO-DIMM) and desktop (DIMM) varieties. This sounds like a lot of different combinations and possibilities, but, in reality, unless you are reading this before I wrote it then you will likely only need to consider DDR3 and DDR4. To start off with some basic party knowledge to impress your friends, DDR stands for double data rate and DDR2, DDR3 and DDR4 are just the second, third and fourth generations of this technology. Each generation has come around and obsoleted the previous generation and since we have now progressed to DDR4, DDR and DDR2 systems are basically extinct. If you find that your system has these in them, it is probably time to buy a new system entirely instead of upgrading the RAM.
Now that you know about the various generations of RAM I can explain what this means for you when you are purchasing memory. The first thing to clear up is that none of the generations are compatible with each other. Some motherboards have both DDR2 and DDR3 sockets or DDR3 and DDR4, but even in this case, you will not be able to run both kinds at the same time. All of the different versions of DDR have different pin layouts so there is no possibility of putting an incompatible type in the wrong socket.
This is true of both desktop and laptop modules
So the most important thing when you are looking for memory is making sure you are purchasing the generation that matches your system. Currently, the majority of systems are DDR3 systems but if you purchased a system in the last 2 years there is a good chance that it uses DDR4 because that has become the new standard. You can easily find this by looking up the part number of the memory you already have by looking up the memory part number that you have pulled by using our custom tool or using CPU-Z and going to the memory tab and it will tell you what type you have.
This is likely the first thing you thought of when you thought of memory. I won't write too much on this because this is something most people already understand. Capacity is simply the amount of information that the stick can store. People generally are already aware how much RAM they have in there system and this is probably why they are considering upgrading. If you open up your task manager and find that the amount of memory being used is almost at 100% of the memory installed you should probably upgrade. Memory comes in all different capacities so you should find an option that comes in a capacity that you think will meet your needs while also being within your budget. One thing I will note about compatibility and sizes is that typically systems will not work with three different sizes of sticks in the same system. So if you have four slots and have 2 4GB sticks and one 8GB stick then it might not be possible to put in a 2GB or 16GB stick.
ECC or Non-ECC
The one thing that most commonly catches people off guard (even season system builders) is ECC memory. ECC memory is memory used for servers and workstations. It stands for error correcting code. This is memory that, as the name states, corrects errors if they come up. This is very useful for servers and workstations that need to be up 100% of the time or run long simulations without errors, but for a typical desktop, it is excessive because you can simply restart a system that is acting up or close and reopen a program that might be acting funny.
If you are reading this guide then you likely do not need and cannot use ECC memory. This is one of the most common mistakes I have seen because counterintuitively these sticks often sell for even below non-ECC prices after they have been out for a while because once a system is retired from a business setting they will likely be sold off and the systems that use these often have many more RAM slots than a typical system and when a business is selling off these it is likely that many other businesses are as well and do not have a need for more because they have moved onto something better. This means they flood the market and prices end up low. If you are not using a system that is designated a workstation or server and has a Xeon or Opteron then you should grab non-ECC memory, sometimes called SDRAM. Now that I have forewarned everyone who might make that mistake, I can explain the types of ECC memory. There are 4 major types and they are all just variations of the same thing.
The first to get out of the way is FB-DIMMs. FB stands for fully buffered. These have basically been phased out since DDR2 so you will not likely encounter these any time soon even if you are using a system that requires ECC memory. These have a different pin layout than even standard DDR2 so they will only fit in sockets that allow FB-DIMMs.
The next type is UDIMM. The U in UDIMM stands for unregistered. This is typically in lower end systems that do not need massive amounts of RAM because UDIMMs limit you to 2 per channel (If you look at the motherboard and see different colored slots, each color is a different channel). This means that they are less commonly used in servers and more often used in Workstations that simply need more reliable options than non-ECC memory but don't need to be able to handle a terabyte of memory. If there is any memory type that will work in a desktop it would be a UDIMM. Some desktops will be able to function just fine with UDIMMs, but they will turn off the error correcting functions so it is still generally better to just go for the non-ECC SDRAM options. On these memory sticks you will likely see the U or E suffix after the speed like on a stick that says 10600U or 10600E.
The final two types of memory are RDIMM and LRDIMM. These stand for registered and load reduced registered memory. Registered, as explained a little bit in the UDIMM explanation is memory that has error correcting functions but also has the ability to be configured with more than two sticks per channel. This is primarily used in high capacity servers that have many slots for RAM.
LRDIMMs are more or less the same thing but are optimized slightly differently so that they can operate at higher frequencies or utilize higher capacity configurations. For example, some systems may max out at 1TB of RDIMMs, but may be able to utilize 2TB of LRDIMMs or they may max out at 1600Mhz in a quad channel configuration with RDIMMs, but may be able to go up to 1866Mhz or 2133Mhz with LRDIMMs.
If you are using a system like this then I highly recommend visiting the manufacturer's site and checking for the most recent documentation to see what these different stipulations are because for each system it is a little different.
Frequencies and Timings
Now that we have talked about the form factors, generations and Types we can move on to the frequencies and timings. Frequency, simply put, is the number of times that a memory stick refreshes the information it is storing. This is the 1600Mhz you might see or the 10600 on the stick. Timings are the number of refreshes that the memory module requires to perform a certain task you will typically see this as 8-8-8-10 or on the memory stick
The fortunate thing about both of these things is that there are standards put out by a group JEDEC so most memory sticks will be the same combination of frequencies and timings. The other thing fortunate thing about memory is that they are all able to run at multiple speeds. This means that if you already have 1333Mhz sticks then they will work with a new set of 1600Mhz sticks that you purchased, but the 1600Mhz will be run at the slower speed of 1333Mhz. This means that if you already have RAM in your system and you are looking to just add new sticks it is likely best to get sticks that are the same speed as what you have unless you can get the faster sticks for cheaper.
Overall I would not worry about frequencies and timings that much other than trying to match with what you have. There have been various benchmarks that have shown mixed results after a certain speed whether there is even any difference at all. Many systems and motherboards do not even officially support the higher speeds and to run them at the RAMs advertised frequency might actually be considered overclocking so it will run at the standard 1600Mhz by default unless you do a little bit of work configuring it in the BIOS.
This is something super simple but also very often overlooked. Sometimes RAM features tall heatsinks or heat spreaders to dissipate the heat produced by the memory sticks and this will hit something in the case or get blocked by the processor's heatsink. It is important to check the inside of your system and find the dimensions of the memory before you purchase it because you might end up with sticks that physically do not fit.
There are plenty of variations of each module of RAM though and it is very unlikely you will be unable to find sticks that do not fit in your system. They also make low profile sticks that are significantly shorter if you do have a system that does not seem to have proper space. These are typically more expensive though so I would recommend going for regular height sticks first.
When looking for memory the first thing you should determine is what form factor of memory your system uses. Either SODIMM if it is a laptop or DIMM if it is a desktop. Then you need to determine what generation of memory it is. You can check this by looking up the documentation of your system or using our custom tool and looking up the part numbers of your current memory or using CPU-Z to tell you. Most likely it is DDR3 or DDR4 at this point. Next, you should determine whether you need ECC or non-ECC memory. By default, everyone should choose non-ECC unless you know that the system you are upgrading is a workstation or server. If you are in the minority and need ECC memory you should choose between UDIMMs, RDIMMs and LRDIMMs based on your use. This is a complex decision to make but, basically, UDIMMs should be for workstations that don't need massive amounts of RAM, RDIMMs for those who need a bit more and LRDIMMs who need the absolute maximum capacity and speed possible. After that, you should determine what frequency and timing you have and try to match that or find the cheapest option that exceeds your current sticks. Then finally you should make sure that the memory will physically fit in your system and has proper clearance. If you follow all of these steps you should have no issues with incompatibility.
What to look for when checking compatibility of a processor
Picking out a processor can be a difficult task. There are many different aspects you need to look at in order to make sure a specific processor will work with your system. Sometimes the documentation is incomplete or you can't find it. Sometimes there is just some weird combination of factors that make things not work and sometimes the information straight from the manufacturer is simply incorrect. In this guide, I am going to explain to you how to sift through the information so that you can determine whether or not a processor will work in your system. In the end, it is very simple. There are only really 4 factors for most modern systems, even less if you are using desktop parts. The factors are socket type, processor generation, motherboard chipset and BIOS revision (and memory type if you are on a server).
This, in the past, was the simplest of aspects to look at when determining a processor for a system because it is essentially just like when you played with blocks when you were a kid and needed to put the correct block in the same hole. This has been complicated in recent years because both Intel and AMD have decided to use very similar socket types across a few generations of processors and only swap out one or two pins or upgrade a chipset without changing the socket. This means that nowadays a processor may still fit in the socket but is still not compatible. This is still the simplest way to quickly disqualify many processors from your potential hunt. Simply put, if a processor has a different socket type from your motherboard then it will not work for you( excluding rare cases like where you can use an LGA775 to LGA771 adapter). There are two quick and easy ways to check the socket type of your motherboard. The first is to go directly to the manufacturer and go to their support area and search up your motherboard or desktop. If you are unsure of what model you have you can use our tool to look up your motherboard or desktop model and then look at the manufacturer's website with that information
Another option is to find the socket type directly by downloading a tool like CPU-Z and letting that do a quick scan of your system.
Once you have found which socket your motherboard has you can start looking at potential processors. One option, if your motherboard uses an Intel processor, is to check the Intel ARK. This website is Intel's official knowledge base which holds just about any processor information you could possibly want. You can do a simple search in the top right for your socket and it will tell you all of the processors that will fit in your socket. This does not guarantee compatibility with your specific system, but it is a start.
If you do not have an AMD motherboard or you could not find any results for your Intel motherboard on the ARK then you can also use a similar tool found at CPU-World. Just find the Socket Type option and then select "contains" and then type in your socket type. After this initial search, you may need to do another search by selecting from a list it gives to you to narrow down the results to your specific socket, but after you do that it should give you a list of potential matches exactly like the Intel ARK. Now you have a list of potential candidates so we can move along to processor generation and motherboard chipset.
Processor Series and Motherboard Chipset
These two next factors go hand-in-hand because generally a new motherboard chipset is released with each series of processors and a processor released with that chipset will be compatible in almost all cases. In the past, it was easy to determine the processor compatibility by looking at the first number of a part number, but recently this has become an unreliable method. For example, relatively recently, Intel released the 4770k and the 4930k at the same time so you would think that they would both work in the same motherboards, but, in fact, they do not even use the same sockets and are incompatible. For a 4930k you need an X79 motherboard and for a 4770k you need a Z97 motherboard. This is why it is dangerous to just look at the leading number. It is best to look at the processor series. In the previous example, you could look up the 4770k and find it was a Haswell processor and then with a search on the Intel ARK you could find a list of the compatible chipsets. To find out which chipset your motherboard uses you can look at the support site from the manufacturer of your desktop again or use CPU-Z again. Many times you do not even need to use these resources if you already know the name of your motherboard though because often the chipset is in the motherboard's name. For example, ASUS Z170-A LGA 1151 Intel Z170 HDMI SATA 6Gb/s USB 3.1 USB 3.0 ATX Intel Motherboard. This tells you in the name that it is a Z170 motherboard and it has an LGA 1151 Socket.
The final factor to look at is the BIOS revision. This is a commonly overlooked factor but can often be a huge frustration point that many people forget about or did not know about in the first place. To simply explain what the BIOS is for those who do not know, it is basically the hardware level settings for everything plugged into your motherboard. In order to run the processor you want, it must have a new enough processor microcode. The microcode is basically the directions for a list of processors and how to run them. If the BIOS does not have the microcode for your processor then it will not boot up because it doesn't have the proper instructions to initiate the processor. Luckily for you, the BIOS is updateable and if a processor comes out after a motherboard is released you can put a newer BIOS on your motherboard and then it will be able to accommodate the new processor. One thing to really emphasize though is that you must update the BIOS before installing the new processor. Some manufacturers now have a tool so that you may update the BIOS via USB without a processor installed but in many cases that may not be possible.
Now that you know all about what the BIOS is and what you can do with it I can talk to you about what you need to look at to confirm compatibility. The first thing to do is to simply look at your motherboard or desktop manufacturer's website. Typically this information will be under support and then maybe downloads and then you will find various BIOS versions there. Some manufacturers will simply list out which processors are compatible and which aren't but for some you will need to delve into their documentation a little bit further. The document name I typically look for is Release Notes somewhere near the BIOS version. In these documents, you can find if there is a version of the BIOS which included an update to the microcode or added compatibility for a certain series of processors. If you find that the processor you want to use was added on by a BIOS update then you need to check for your BIOS revision and make sure it is newer than the required version to run the processor you want. You can find this in CPU-Z under the mainboard tab.
If your BIOS revision is not new enough then you will need to update it before you can use the new processor. You can either find the most recent version off of the manufacturer's website or you may use a tool like BIOS Agent Plus to find the most recent version for you and help you install it. It is very important to make sure that you follow any directions given to you very closely when dealing with the BIOS because you can easily brick your motherboard if you turn off the power when updating or if you update with the wrong version.
TDP( Total Design Power)
TDP is another thing that used to be an important consideration when looking at processors and motherboard compatibility but it has fallen by the waist side as systems have become more efficient and developed the standards to deliver adequate power. I figured it was worth to add in any way for those who may be using older systems. To explain for those who do not know what TDP is, it is basically the total wattage that the CPU will consume and then convert to heat. This is a particularly important consideration on all of the AMD processors because if you look at the highest end processors they have the FX-9590 has a TDP of 220W. This is a gargantuan number in comparison to most current processors and many motherboards likely cannot put that much wattage through their socket without melting their voltage regulators or frying themselves. For reference, 220W is likely more power than your entire system draws if you are on a modern system and do not have a high-end graphics card installed. If you have an AM2, AM2+, AM3, or AM3+ and are looking for compatible parts it is probably wise to also consider the TDP and see if there is any documentation along with your system that says specifically whether it can handle a 135W processors or just a 90W processor and you may also even want to look to see if your power supply can provide ample wattage. I have seen plenty of pre-built systems that have power supplies rated for 225 watts or less and if you plug an FX-9590 into one of those you would really be playing with fire.
To summarize, the first thing you should look at when looking for compatible processors is the socket type. If you do not have the correct socket type then the processor will likely not fit or the pins will not match up. After you determine that it is the correct socket you should look at the CPU series and compare it to the chipset series. There are plenty of resources online to make sure that your motherboard has a chipset that will likely be compatible with the processor you are interested in. Check Wikipedia or the Intel Ark and they will guide you in the right direction. After that, you should look at the manufacturer's website and see if the processor is listed in their documentation anywhere and if it is, does it require a new BIOS revision to run. These are the basic steps that I run through whenever I am uncertain about the compatibility of a processor. If you have any further questions about compatibility or have something you would like to add to this guide, please leave a comment and I will try to address it as soon as I can.
Bonus Cheat Sheet
RAM Type: DDR4 RDIMMs, LRDIMMs
Max Ram Capacity: RDIMMs: 768GB(32GB DIMMs), LRDIMMs: 3072GB (128GB DIMMs)
Processor Socket Count: 2
Compatible Processor Series: E5-2600 v3, v4 or E5-1600 v3, v4
Number of Ram Slots: 24
Hard Drive Bays: 18 3.5" drives or 32 2.5" drives
BIOS Notes: Revision 2.1.5 adds support for E5-2600 V4 and E5-1600 V4 processors
128GB LRDIMMs Supported with V4 processors and BIOS 2.1.5 and newer.
The Dell PowerEdge T630 is a Tower based server that has the ability to be converted into a rackmount server when a mounting bracket and rails are installed. It features a proprietary Dell motherboard which is run by an Intel C612 series chipset (Desktop Similar is the X99 Chipset). This allows for dual Haswell-EP (E5-2600 v3 series) processors or Broadwell (E5-2600 v4 series) processors (If BIOS revision 2.1.5 or newer is installed) and 24 slots for DDR4 RDIMMs or LRDIMMs
When running BIOS revision 2.1.5 or newer the maximum addressable memory for a single socket configuration is 1536GB with Quad or Octo Rank Load-Reduced DIMMs. With a dual processor configuration, this figure is doubled to 3072GB. This is using 24 of the recently released 128GB LRDIMMs. If using a revision older than 2.1.5 then the maximum addressable will be 768GB in a single socket configuration or 1536GB in a dual socket configuration. This is achieved using 24 64GB LRDIMMs. Of note, you cannot utilize 128GB DIMMs until the BIOS is updated and you may not use the memory slots assigned to a non-populated processor. To explain further, you cannot achieve the 768GB single processor maximum by using 24 32GB DIMMs because 12 of the slots are associated with the empty processor socket. The memory can operate at 2400MT/s in all configurations and ranks except for triple channel configurations. With RDIMMs the maximum then becomes 1866MT/s in triple channel and LRDIMMs can still operate at 2133MT/s in triple channel
There are many storage configurations for the Dell PowerEdge T630. One option is 32 2.5" drives without the flexbay or 18 2.5" with the flexbay. Another option is 18 3.5" drives without the flexbay or 8 with the flexbay. The flexbay is essentially a riser card that allows the install of PCIe SSD cards in the front or additional hard drives or solid state drives.
There are a few known issues with the T630 that have been addressed with BIOS updates. The most major being users experiencing the red screen of death (RSOD) when using the E5-2603 or E5-2609 processors and performing a reboot. This was remedied in BIOS revision 2.1.5 so if you are using this configuration it is highly recommended you upgrade.