i5-11400F RAM Speed Benchmarks – How much does RAM Speed Affect Performance?

The procedure online cialis sales involves an instrument known as Foley Catheter. At present, the prostate remedies are mainly as follow: cialis pills drug therapy, physical therapy, surgical treatment and psychological treatment. Ideal dose to start with is 50mg per day. order viagra As soon as you purchase levitra find an attorney who is an expert in Propecia claims.

If you are building a new system, or even upgrading your current one, with an i5 11400F you might be wondering how much you should be spending on your RAM and what speed you need to get the most out of your CPU. Well, I’ve got an 11400F, a B560 motherboard, and 3600MHz CL16 RAM so lets get testing.

To lay out the plan, I’m going to test at this kit’s maximum of 3600MHz, that’s both the fastest I have right now and about as fast as most people would likely get with this kind of chip. I’m also going to test at 3200MHz, as that’s the maximum Intel recommends, and the most you can use with your warranty intact. I’ll also test at 2933MHz – I’d use 3000 here but the board wouldn’t let me so 2933 it is. I’ll do 2666MHz too, along with the base of 2133MHz to cover the range – that last one will also show you how much performance you are missing if you don’t enable XMP.

Got it? Great, lets look at the benchmarks. Starting with Cinebench R20 single threaded, you will probably be surprised to see that it makes absolutely no difference. The peak performance came from 2933MHz, with 3600MHz technically being the lowest, although the total performance spread is only around 1%, or what I’d say is well within margin of error.

In multithreaded there is a bit of an advantage, although again it’s within 1% from slowest to fastest so I wouldn’t call that a resounding victory for the faster speeds. Even in Blender and the BMW scene, the 3600MHz kit runs technically slowest by 2 seconds, but again that’s within 1% of the fastest so temperature and boost characteristics are more of an influence than RAM speed.

But then when we move to the Gooseberry scene, it all starts making sense. The 3600MHz kit is the fastest, then 3200, then 2933, then 2666, with 2133 coming in last and by a reasonable margin. The spread is around 8% here, although the grouping at the top is pretty tight. 2933 is only 1.8% slower than 3600, which considering it’s effectively 23% faster memory isn’t a big gain.

The trend continues in the Puget Bench suite for Premiere, where 3600MHz is fastest, and follows in order down to 2133 which is 9% lower. Again, the difference between 2933 and 3600 is only around 3%, so it’s hardly a massive drop off.

After Effects is a lot tighter, it’s still mostly in order and has a 5% spread, but beyond 2133 they are pretty much within margin of error. Photoshop is much more like Premiere where the spread is a full 12%, and it’s a pretty gradual drop in performance from each step.

To explain what I think might be going on here with the results that seem almost random, versus ones that have a clear benefit to faster RAM, take Blender and the BMW scene. That scene is a 3MB file and Blender, all in, only used about 500MB of RAM while rendering, with 350MB of that being the base blender program. Compare that to the Gooseberry scene which is a 240MB file, and on my PC with 64GB of RAM, uses a full 11GB of that to render, with around 1.8GB of that being the base blender program with that scene open.

In the BMW scene, RAM usage is incredibly low meaning it isn’t transferring much data in and out of memory so the speed of the memory isn’t a big factor. Whereas in Gooseberry there is much more data that needs to be passed around so slower transfer rates will make a bigger difference.

Right, that’s the CPU specific benchmarks, but what about gaming? Lets start with Watchdogs Legion. Technically, 3200MHz takes the lead here by 2 FPS average, although matching the 3600’s 1% low figure. That’s close enough to call basically the same, although 2933, while only 1 FPS lower only average, has 4 FPS less in the 1% lows. The real standouts though are 2666 with 75 FPS average, 8 FPS less than the fastest, and a rather shocking 64 FPS average from 2133, which is almost 20 FPS less. That’s noticeable, as is the 47 FPS in the 1% lows.

In Cyberpunk the speeds are all in order fully, with a little gap to 3200, then 8 FPS to 2933, a further 6 FPS to 2666, then 9 more lost at 2133. That means 2133 is 30% slower than 3600. That’s massive. The 1% lows were also not far off being half that of 3600 too. That’s insane.

In CSGO it’s the same story except for a little inversion with 2666 beating 2933 on average, although 2933 has a sizable advantage in the 1% low figures so call it even between them. Again, 2133 drops over 20% of its average performance compared to 3600, and almost 18% from 3200MHz. Enable XMP kids.

Fortnite is honestly shocking. 3600 and 3200 are closely matched, with 3600 holding slightly better 1% low numbers but close enough, then a small gap to 2933, a bigger jump to 2666, then a chasm between that and 2133. 2133 is 26% slower than 3200 here, that’s a big loss.

MS Flight follows a linear progression down, with 27% spread, although it’s little increments at each speed drop. The biggest gap is actually from 3600 to 3200, then it’s only 1 or 2 FPS lower each time.

So that’s all the data, what’s the takeaway? Well, in some CPU specific applications, especially ones that don’t use much RAM while in use, it really doesn’t matter much what speed you run. In apps that do use more RAM, Intel’s spec of 3200MHz fits pretty well. You can eek out a little more performance with 3600, but it’s not a massive improvement. Only splash out for that over 3200 if it’s a really small price difference, and remember you will be voiding your CPU warranty by enabling it to that speed.

In gaming it’s a little more clear cut. The faster RAM you get, the more performance you will see. Sometimes, sizably. Now would be a good time to open task manager on your PC and check it’s running at the speed you paid for. The jump in performance from 3200 to 3600 isn’t generally massive, so if you want to save a few quid and keep your warranty, 3200 will be fine.

There is one thing I haven’t specifically covered here, which is that I’m using the exact same kit of memory for all these tests and just changing its speed in the BIOS. Now that’s good to see the effect speed alone has, but in the real world if you are buying a slower kit of memory, you might end up with a faster CAS latency which will give you a performance edge over what I’ve tested here, but it seems like CAS 16 is about as low as you can get with either 3000, 3200 or 3600 right now so it’s not a big deal.