Speed Matters! Find out why 2400MHz high-speed memory should be your next upgrade with Corsair

[Admetos]

On a modern CPU there is a very high percentage of cache hits from on board memory cache, hence why faster memory normally only give small gains. This is why memory benchmarks will show faster on say an i7 (with 8mb cache) compared to say an i3 (with 3mb cache) as the larger cache on the i7 is increasing the change of a cache hit.

There are some applications such as WinRar that has relative low cache hits hence faster memory shows a good increase here. Faster memory helps with Virtual Machines where the CPU is context switching between VM's.

On the 80286 downwards there was no cache, so the memory timing were important, anyone old enough to remember zero wait state memory where the memory was fast enough so CPU did not have to wait an extra clock cycle!

 

[Rroff]

Did a few quick benchmarks of the opening scene of BF4.

The difference between running my RAM at 2133MHz XMP (11-12-12-30 - sub-timings 8/171/48 type numbers) and 2133 9-10-10-21 with sub-timings down to 6/125/24 type numbers (best of 3 runs each):

Min 72, Max 138, Avg 96.790
Min 72, Max 139, Avg 97.124

Had a few plays with bumping up the frequency a bit to its out the box 2400MHz profile (numbers pretty much identical to the XMP settings), even dropping back to 1600MHz speeds only saw the min drop back to 69 and max to 128 (only did one run) with hardly any change to the average. (all the rest of the runs were bang on 72 min so that 1600 dropped to 69 probably shows a very slight effect from memory performance there).

 

[Darkman]

Well that's what happens when you let the Marketing Dept have free rein

Might as well have changed the end of the OP to:

LAUGH NOW

 

[Rroff]

^^ To be fair I did see a slight drop off at 1600 over 2133 and probably a bit more noticeable if your running a decently high end multi GPU setup. Not really enough to justify a big song and dance about it though.

I'll stick to what I always say the sweet spot for current Intel platforms if you've got a little cash to play with is 2133 @ CAS9 unless your doing something that specfically takes advantage of a different configuration and I only buy 2400 kits as it gives me more headroom to play with tweaking wise.

 

[-Mike-]

If you tune the subs with high perf RAM you can get another few FPS in BF4 etc which need bandwidth. Its not just primary timings at all.

This guy doesn't see much difference

Did a few quick benchmarks of the opening scene of BF4.

The difference between running my RAM at 2133MHz XMP (11-12-12-30 - sub-timings 8/171/48 type numbers) and 2133 9-10-10-21 with sub-timings down to 6/125/24 type numbers (best of 3 runs each):

Min 72, Max 138, Avg 96.790
Min 72, Max 139, Avg 97.124

Had a few plays with bumping up the frequency a bit to its out the box 2400MHz profile (numbers pretty much identical to the XMP settings), even dropping back to 1600MHz speeds only saw the min drop back to 69 and max to 128 (only did one run) with hardly any change to the average. (all the rest of the runs were bang on 72 min so that 1600 dropped to 69 probably shows a very slight effect from memory performance there).

Glade you know best Dice Hunter

Which scent do you prefer?

 

[Akagi]

Which scent do you prefer?

Clearly it is BS?

If you want people to buy overpriced memory get someone to make fancy banners and post them on social networking sites

 

[GoogalyMoogaly]

£320 on new memory (replacing 16GB @ 1866MHz) or £320 on a 3rd R9 290. Which would give the biggest FPS increase?

 

[-Mike-]

£320 on new memory (replacing 16GB @ 1866MHz) or £320 on a 3rd R9 290. Which would give the biggest FPS increase?

I'd imagine the gfx card

 

[Dicehunter]

Here's a good comparison between 1600 and 2400 in a few games and benchmarks done by Lavco Price Tech on Youtube only 3 months ago, The results speak for themselves.

 

[Rroff]

What GPU(s) did they use though? I see people doing these DDR3 roundups... and using a GTX660 or 7870 :| while its not huge there is some gains to be had in gaming with faster than 1600 RAM on higher end GPU setups.

 

[Dicehunter]

What GPU(s) did they use though? I see people doing these DDR3 roundups... and using a GTX660 or 7870 :| while its not huge there is some gains to be had in gaming with faster than 1600 RAM on higher end GPU setups.

They were using a GTX 580 and a Core i5 4670K, When I switched from 1600 to 2400 using a 780 Ti I didn't notice any FPS differences and I'm a monitor whore , By that I mean I always have an overlay up with GPU, CPU, Speeds, Temps etc... on every game.

As said and as most people with any know how and common sense know, It's just a marketing ploy, Bad move trying it on with enthusiasts though

 

[Kaynar]

So graph, many difference, much true. WoW Corsair!

 

[Antony101]

Just turn Windows Aero off, you'll see similar FPS increases.

 

[almighty15]

Only certain games show an improvement... 99% of them don't generally gives a rats ass about memory speed....

Naughty Corsair cherry picking games...

 

[regulus]

If im honest im a little disappointed to see something like this from corsair, especially posted here on OCUK. Its extremely misleading.

Which is why I'll never buy their RAM now. Why try and mislead?

Do they think a forum with tech savvy people won't notice?

 

[mothermachine]

It's not really misleading exactly - it's just a good marketing trick and OCUK and Corsair are businesses that exist solely to make money. I'd expect nothing less from them or any other manufacturer or retailer. This kind of graph is common and clearly does affect sales otherwise they wouldn't bother.

But fortunately we're a community of fairly knowledgable people who can see beyond most marketing nonsense and can quickly conclude that any gains from spending extra cash on faster RAM are absolitely minute at best and you'd be better off in almost all cases by putting that cost saving from going with slightly slower RAM into a better GPU/CPU/SSD, or even a night out with the missus, frankly.

That said, if you've got the cash and no reason or desire to skimp then you'd be daft not to, right? Same with any other computing tech. You buy what is appropriate for your budget and specific use.

 

[leoncao]

samsung green will just do fine :/

 

[David Bisset]

What GPU(s) did they use though? I see people doing these DDR3 roundups... and using a GTX660 or 7870 :| while its not huge there is some gains to be had in gaming with faster than 1600 RAM on higher end GPU setups.

This really.

As I said earlier in the thread the value of upgraded RAM will depend heavily on the rest of the system - if you're already on a very high end system then making it faster in any way is expensive, so the value of this upgrade may be there. In a low end system the money is better spent elsewhere. By low end I'm including any single-GPU setup as there appears little value until you're on several.

No number of benchmarks done with these systems to measure the performance difference will prove if Corsair are telling the truth or not given they're on a multi-GPU system at high res and said as much.

If we want verification we need to beg Kaap or someone similar to test for us

 

[Cleisthenes]

Why not spend the extra 100 quid on a better gpu eh?

 

[setter]

Looking at the res used in the graph, and a high end multi gpu setup. A higher amount of ram at a faster speed might well give a few extra fps. Not many of us here are running as high as this, i guess at that level of detail it could help a bit. I too would like to see someone with such a system here test this out.