Processors
Intel® Processors, Tools, and Utilities
14403 Discussions

Using high speed ram over Intels ratings

TEvan5
Beginner
1,542 Views

What happens when you use ram at speed rated higher then the cpu manufacturers recommendation?

There seems to be a lot of conjecture and difference of opinion in regards to this topic.

I would really like to know Intel`s official position of this as I recently bought a 7700k with 16 gb of ram (4x4) at 2400mhz.

This meant buying a new mainboard, ram and cpu which personally is a large outlay so naturally i want to know did I do the right thing, especially as there are so many people using ram @ 3000mhz and above.

I read an another forum that using xmp IS overclocking and therefore voids any warranty with Intel. Is this correct?

Do these users actually get the speed they are aiming for?

Would this cause some sort of data thrashing/bottleneck?

Will it increase heat of the ram or cpu?

Does it shorten the life of the ram or cpu?

I would have thought that these questions would be easily answered by most retailers but the best I can find can only say 'should be ok' or 'I'm doing it without issues so far...' but never a definitive answer.

All comments appreciated

0 Kudos
1 Solution
n_scott_pearson
Super User
625 Views

Tim,

I am not officially Intel (not anymore; I am retired), but will give the usual spiel...

First of all, Intel does not officially condone overclocking. It warrants its processors to operate at specific clock speeds, voltages, etc. Anything above these levels is considered overclocking. Overclocked processors that fail are not covered under warranty (though, you can purchase overclockers insurance that will see your processor replaced once if it fails as a result). Now, while Intel does consider running your DRAM at speeds above the maximum specified, including via XMP profiles, to be a form of overclocking, this does not automatically void your warranty. The specified maximum is the speed that Intel validates and warrants the processor to support. Operation at any faster speed is not validated and is not guaranteed to work. If it works, great; if it doesn't, well, get slower memory. [Aside: I would add that there is a far higher probability of the motherboard failing to support the higher speed than there is the processor.] Ignoring voltage spikes, shorts, etc., in most cases, a processor that fails is as a result of thermal damage. This can be the result of the owner failing to provide sufficient cooling or as a result of overclocking. Thermal degradation can occur if temperatures spend too much time above the processor's thermal load line. At higher-than-designed clock rates, transistors can be damaged by heating up and cooling off at higher that designed/expected rates.

Ok, to your questions...

  1. Do these users actually get the speed they are aiming for? They can get small gains in performance. It has been shown (at least for Windows-based PCs), however, that increasing the amount of memory can have more of an overall performance impact than does speeding up the memory.
  2. Would this cause some sort of data thrashing/bottleneck? It can most certainly do so. The issue is noise. As clock speeds are raised, the noise generated by the processor, the board, the DIMMs and the DRAM ICs themselves will increase. Motherboards are designed to eliminate the effects of small amounts of noise, but, if this buildup of noise goes beyond that and reaches threshold levels, data errors will begin to occur. This can be recovered from if errors are at the single bit level, but multi-bit errors can result in resets, bus lockups, etc. In addition to this, in all components, more noise will be generated as components age. This can result in situations where memory that worked initially will no longer work unless lower clock speeds are used. If the BIOS detects that memory is not going to operate at specific speeds, it can automatically dial back the bus speed (and other parameters like wait-states) in an attempt to get it working. This can only go so far, however.
  3. Will it increase heat of the ram or cpu? Yes, it can generate additional heat at both ends. In the case of the memory controllers, the additional heat may be dissipated such that the overall increase in heat is not measurable. The problem is that overclocking can result in transistors heating up and cooling at unplanned-for rates and this can damage these transistors over time.
  4. Does it shorten the life of the ram or cpu? Yes, it can, obviously.

Again, I am retired from Intel and not an official voice. These are my opinions only. Do consider that I am a 35+ year veteran of the industry, a 21-year veteran of Intel and I spent a considerable amount of my career at Intel working on system thermals and acoustics.

...S

View solution in original post

4 Replies
n_scott_pearson
Super User
626 Views

Tim,

I am not officially Intel (not anymore; I am retired), but will give the usual spiel...

First of all, Intel does not officially condone overclocking. It warrants its processors to operate at specific clock speeds, voltages, etc. Anything above these levels is considered overclocking. Overclocked processors that fail are not covered under warranty (though, you can purchase overclockers insurance that will see your processor replaced once if it fails as a result). Now, while Intel does consider running your DRAM at speeds above the maximum specified, including via XMP profiles, to be a form of overclocking, this does not automatically void your warranty. The specified maximum is the speed that Intel validates and warrants the processor to support. Operation at any faster speed is not validated and is not guaranteed to work. If it works, great; if it doesn't, well, get slower memory. [Aside: I would add that there is a far higher probability of the motherboard failing to support the higher speed than there is the processor.] Ignoring voltage spikes, shorts, etc., in most cases, a processor that fails is as a result of thermal damage. This can be the result of the owner failing to provide sufficient cooling or as a result of overclocking. Thermal degradation can occur if temperatures spend too much time above the processor's thermal load line. At higher-than-designed clock rates, transistors can be damaged by heating up and cooling off at higher that designed/expected rates.

Ok, to your questions...

  1. Do these users actually get the speed they are aiming for? They can get small gains in performance. It has been shown (at least for Windows-based PCs), however, that increasing the amount of memory can have more of an overall performance impact than does speeding up the memory.
  2. Would this cause some sort of data thrashing/bottleneck? It can most certainly do so. The issue is noise. As clock speeds are raised, the noise generated by the processor, the board, the DIMMs and the DRAM ICs themselves will increase. Motherboards are designed to eliminate the effects of small amounts of noise, but, if this buildup of noise goes beyond that and reaches threshold levels, data errors will begin to occur. This can be recovered from if errors are at the single bit level, but multi-bit errors can result in resets, bus lockups, etc. In addition to this, in all components, more noise will be generated as components age. This can result in situations where memory that worked initially will no longer work unless lower clock speeds are used. If the BIOS detects that memory is not going to operate at specific speeds, it can automatically dial back the bus speed (and other parameters like wait-states) in an attempt to get it working. This can only go so far, however.
  3. Will it increase heat of the ram or cpu? Yes, it can generate additional heat at both ends. In the case of the memory controllers, the additional heat may be dissipated such that the overall increase in heat is not measurable. The problem is that overclocking can result in transistors heating up and cooling at unplanned-for rates and this can damage these transistors over time.
  4. Does it shorten the life of the ram or cpu? Yes, it can, obviously.

Again, I am retired from Intel and not an official voice. These are my opinions only. Do consider that I am a 35+ year veteran of the industry, a 21-year veteran of Intel and I spent a considerable amount of my career at Intel working on system thermals and acoustics.

...S

TEvan5
Beginner
625 Views

N.Scott.Pearson Thank you for your detailed reply. Can i have your permission to use this in a related facebook discussion?

0 Kudos
n_scott_pearson
Super User
625 Views

Sure. Technically speaking, by posting it here, it is in the public domain already.

...S

0 Kudos
TEvan5
Beginner
625 Views
0 Kudos
Reply