3200Mhz RAM

[Redjo]

Hello my friends here is quick question that borders me:

Will i have heat issues with this memory because of 3200mhz? ---> G.Skill RipJaws V Black DIMM 2x4 3200Mhz

Do i need additional cooler for it ? 

[ybriK]

No, pretty sure computer hardware companies wouldn't sell a product that has a major flaw in it...

[Enderman]

nah

[Bajantechnician]

8 minutes ago, Redjo said:

Hello my friends here is quick question that borders me:

Will i have heat issues with this memory because of 3200mhz? ---> G.Skill RipJaws V Black DIMM 2x4 3200Mhz

Do i need additional cooler for it ? 

if you didnt buy it already, u dont need 3 ghz.

2 is more than plenty

[MuggenZifter]

why the hell would you get such high clocked memory. Just get 1866 or 2133

[porina]

I have the 2x8GB version of that kit. No need for cooling. Ram speeds can help some use cases and for my uses, dual channel dual rank 3200 is still not fast enough it is limiting my 6700k at stock turbo.

[MuggenZifter]

2 minutes ago, DEcobra11 said:

Ripjaws V are DDR4, they aren't that high clocked from that perspective

i understand but do you realy need the 3200? I would pick a lower clocked. Then you will be definitely safe with the temps.

[Redjo]

Thank you all for the help! 

The reason i go 3200mhz is because it cost 50€ which is good i guess.

 

Edit: My CPU is 5820k with MSI X99A SLI mobo 

[TheFlyingSquirrel]

1 hour ago, Redjo said:

Thank you all for the help! 

The reason i go 3200mhz is because it cost 50€ which is good i guess.

 

Edit: My CPU is 5820k with MSI X99A SLI mobo 

Running their DDR3 predecessor with 2400mhz.  You will not have a heat problem if you can get it stable, but at the same time you will not get a performance gain going up to 3200mhz as there is no use case for it (Though nice to know that your RAM can handle anything). 

[SteveGrabowski0]

1 hour ago, MuggenZifter said:

why the hell would you get such high clocked memory. Just get 1866 or 2133

Because it performs better

[TheFlyingSquirrel]

5 minutes ago, SteveGrabowski0 said:

Because it performs better

Actually in real world scenarios, there is little difference between 2133 and upwards due to latency trade offs higher up the MHz scale

[SteveGrabowski0]

23 minutes ago, kidanime3d said:

Actually in real world scenarios, there is little difference between 2133 and upwards due to latency trade offs higher up the MHz scale

You don't trade off latency, it's usually pretty flat as you go to higher speeds. Latency is a time delay, and it's inversely proportional to clockspeed. Higher CAS number doesn't necessarily mean higher latency unless you're comparing two memory kits running at the same clock. Meanwhile higher clockspeed gives you more throughput.

[TheFlyingSquirrel]

1 minute ago, SteveGrabowski0 said:

You don't trade off latency, it's usually pretty flat as you go to higher speeds. Latency is a time delay, and it's inversely proportional to clockspeed. Meanwhile higher clockspeed gives you more throughput.

 

If the time delay is higher I call that a trade off. Unless you have a real world test that proves higher clockspeed to be of real benefit, my point is still valid.   I have same G. Skill 2400Mhz RAM (different series) as you and found no benefit during my research.

[SteveGrabowski0]

5 minutes ago, kidanime3d said:

 

If the time delay is higher I call that a trade off. Unless you have a real world test that proves higher clockspeed to be of real benefit, my point is still valid.   I have same G. Skill 2400Mhz RAM (different series) as you and found no benefit during my research.

I found a lot of benefit in GTA V and Fallout 4 going from DDR3-1600 to DDR3-2400 in terms of better minimums at very specific places on the maps that always dragged my framerate down on DDR3-1600.

 

And the time delay isn't always higher, and when it is it's often really minimal, much less then just saying CAS 13 > CAS 15 would imply.

[TheFlyingSquirrel]

3 minutes ago, SteveGrabowski0 said:

I found a lot of benefit in GTA V and Fallout 4 going from DDR3-1600 to DDR3-2400 in terms of better minimums at very specific places on the maps that always dragged my framerate down on DDR3-1600.

Yes the benefit from 1600 to 1866 is noticeable and well known, but we are talking about 2133 vs 2133 upwards here.  You stated 1866Mhz andd 2133Mhz are inferior to higher speeds, but no test has proved this; if anything they prove the higher speeds have no extra value for everyday use.

[SteveGrabowski0]

1 minute ago, kidanime3d said:

Yes the benefit from 1600 to 1866 is noticeable and well known, but we are talking about 2133 vs 2133 upwards here.  You stated 1866Mhz andd 2133Mhz are inferior to higher speeds, but no test has proved this.

Digital Foundry has proved it for DDR4-2666 vs DDR4-2133 and even for DDR4-3000 vs DDR4-2666.

[SteveGrabowski0]

12 minutes ago, kidanime3d said:

 

If the time delay is higher I call that a trade off.

Here is an example to show CAS number is not the same as latency:

 

A CAS 9 DDR3-1600 kit will have a column address strobe latency of

 

(9 cycle) / (800 * 10^6 cycle / second) = 11.25 nanosecond

 

A CAS 11 DDR3-2400 kit will have a column address strobe latency of

 

(11 cycle) / (1200 * 10^6 cycle / second) = 9.17 nanosecond

[porina]

2 hours ago, kidanime3d said:

Yes the benefit from 1600 to 1866 is noticeable and well known, but we are talking about 2133 vs 2133 upwards here.  You stated 1866Mhz andd 2133Mhz are inferior to higher speeds, but no test has proved this; if anything they prove the higher speeds have no extra value for everyday use.

It isn't even as simple as that. How fast the CPU is will affect the demands it makes on the ram. As CPUs get faster, the ram needs to get faster (in bandwidth) to keep up with it. There will be tasks that are efficiently served by the CPU cache that ram performance makes little impact. If you start exceeding this then ram performance starts making bigger impacts. Problem is there is no simple way to know if ram will help or not short of testing it. In my interest area of finding prime numbers, Intel CPUs from Haswell onwards are significantly limited by lack of memory bandwidth. For a quad core dual channel ram system, I worked out a rule of thumb you want to aim for ram speeds equal to the CPU clock to run >90% IPC efficiency compared to having unlimited ram bandwidth. Basically my Skylake systems are all ram speed limited to some degree. I didn't work out until after I got them, that Haswell-E starts to make sense. In this application, latency doesn't have anywhere near a significant impact as the bandwidth. Intel CPUs older than Haswell and anything from AMD are slow enough to not be as limited.

 

But is the above "real world"? Maybe not for 99% of people, but it is no synthetic benchmark either. In a more general case, the system might be limited outside the CPU/ram. For example, gaming might be more GPU limited. This seems to be the case for many games, but there seems to be more exceptions as people start looking at it more.

[TheFlyingSquirrel]

16 minutes ago, porina said:

It isn't even as simple as that. How fast the CPU is will affect the demands it makes on the ram. As CPUs get faster, the ram needs to get faster (in bandwidth) to keep up with it. There will be tasks that are efficiently served by the CPU cache that ram performance makes little impact. If you start exceeding this then ram performance starts making bigger impacts. Problem is there is no simple way to know if ram will help or not short of testing it. In my interest area of finding prime numbers, Intel CPUs from Haswell onwards are significantly limited by lack of memory bandwidth. For a quad core dual channel ram system, I worked out a rule of thumb you want to aim for ram speeds equal to the CPU clock to run >90% IPC efficiency compared to having unlimited ram bandwidth. Basically my Skylake systems are all ram speed limited to some degree. I didn't work out until after I got them, that Haswell-E starts to make sense. In this application, latency doesn't have anywhere near a significant impact as the bandwidth. Intel CPUs older than Haswell and anything from AMD are slow enough to not be as limited.

 

But is the above "real world"? Maybe not for 99% of people, but it is no synthetic benchmark either. In a more general case, the system might be limited outside the CPU/ram. For example, gaming might be more GPU limited. This seems to be the case for many games, but there seems to be more exceptions as people start looking at it more.

 

You're first statment is incorrect, information retrieval from RAM is based on cycles, not bandwidth, the bandwidth on mobo is already adequate. If tasks are served by the Cache, the RAM has no impact as the CPU does not need to make a request. My i7 4770k memory bandwidth is 25.6 GB/s a second, aand I have never seen that filled.

These aren't benchmarks these are timed tests forr video editing, photo editing, and fps for gaming where the improvement was around 1-2%, using what was at the time a high end card.   So yes in real world scenarios high clock speed RAM has little to no benefit, over 1866Mhz and 2133Mhz.  Buy G Skill stuff because it is reliable and cheap, not because you think 2400Mhz+ is superior

[MageTank]

14 minutes ago, kidanime3d said:

 

You're first statment is incorrect, information retrieval from RAM is based on cycles, not bandwidth, the bandwidth on mobo is already adequate. If tasks are served by the Cache, the RAM has no impact as the CPU does not need to make a request. My i7 4770k memory bandwidth is 25.6 GB/s a second, aand I have never seen that filled.

These aren't benchmarks these are timed tests forr video editing, photo editing, and fps for gaming where the improvement was around 1-2%, using what was at the time a high end card.   So yes in real world scenarios high clock speed RAM has little to no benefit, over 1866Mhz and 2133Mhz.  Buy G Skill stuff because it is reliable and cheap, not because you think 2400Mhz+ is superior

Alright, I am tagging in. I need to put an end to your nonsense before people start believing you. First of all, just because ram speed increases, it doesn't mean latency gets higher. This is a myth that has NEVER been true. Sure, you get your morons that run XMP kits that trade higher clock speed for terrible tertiary timings, but real overclockers get both. Want proof?

 

 

4.4ghz CPU, 3050mhz memory. Look at my bandwidth and latency above.

 

 

4.4ghz CPU, 3500mhz memory. Look at how my bandwidth AND latency improved. Woah, it's almost as if people don't have to trade speed for latency. Mind boggling, right? If you have the slightest clue as to how memory works, not really. 

 

Your statements on real world improvement only being "1-2%" is false, and I've already proven that with my own tests. If you want me to test it again with a 6600T at 4.4ghz and a GTX 770, I'll gladly do so. I have yet to even mention how important memory speed and latency is for multi-GPU setups either. We are talking a worlds difference in frametime and minimum FPS. Next you will tell me that isn't real world either, because "SLI/Crossfire is uncommon". 

 

All the more reason for me to kickstart this memory guide I've been working on. 

[TheFlyingSquirrel]

10 hours ago, MageTank said:

Alright, I am tagging in. I need to put an end to your nonsense before people start believing you. First of all, just because ram speed increases, it doesn't mean latency gets higher. This is a myth that has NEVER been true. Sure, you get your morons that run XMP kits that trade higher clock speed for terrible tertiary timings, but real overclockers get both. Want proof?

 

 

4.4ghz CPU, 3050mhz memory. Look at my bandwidth and latency above.

 

 

4.4ghz CPU, 3500mhz memory. Look at how my bandwidth AND latency improved. Woah, it's almost as if people don't have to trade speed for latency. Mind boggling, right? If you have the slightest clue as to how memory works, not really. 

 

Your statements on real world improvement only being "1-2%" is false, and I've already proven that with my own tests. If you want me to test it again with a 6600T at 4.4ghz and a GTX 770, I'll gladly do so. I have yet to even mention how important memory speed and latency is for multi-GPU setups either. We are talking a worlds difference in frametime and minimum FPS. Next you will tell me that isn't real world either, because "SLI/Crossfire is uncommon". 

 

All the more reason for me to kickstart this memory guide I've been working on. 

Thank you for your input, the tests I saw were using XMP, and wasn't aware that manual overclocking was all that different.  RAM speed is generally limited by latency due to stability a limitation of the way the I/O bus clock cycle functions.  So your obnoxious "Mind boogling, right?" is simply full of it.

 

It isn't exactly false it is based on others findings who used controlled tests.  I would be interested to see your tests, ie fps comparisons, complex algorithm calculations time fft if you have any.  Power vs necessary is the whole point of this discussion.

[Nena Trinity]

23 hours ago, MuggenZifter said:

why the hell would you get such high clocked memory. Just get 1866 or 2133

2133MHz DDR4 is slow you need/want higher MHz on DDR4!