Just how much difference is there in different ddr5 cas latency and speed for 13th gen in terms of performance?

[kitnoman]

So I'm about to upgrade to 13600k. and I saw TEAMGROUP T-Create Classic DDR5 CL46 32GB Kit (2 x 16GB) 5600MHz. I'm working on no to less RGB build, so a 32gb ddr5 for $97 is a great deal for me. And how does that compare to G.Skill RipJaws S5 Series DDR5 6000 CL36 32GB (2 x 16GB) or RipJaws S5 Series DDR5 6000 CL30 32GB (2 x 16GB) or TEAMGROUP T-Force Vulcan Alpha DDR5 CL38 32GB Kit (2x16GB) 6000MHz or TEAMGROUP T-Create Expert Overclocking 10L DDR5 CL40 32GB Kit (2 x 16GB) 6400MHz and orsair Vengeance DDR5 C36 32GB (2x16GB) 6000MHz . Normally I always go with kingston/hyperx, but with their current pricing, the only thing I would consider is Kingston Fury Renegade Silver DDR5 CL32 32GB (2x16GB) 6400MT/s which is about $70-80 difference than the one I'm currently looking at, while the other ones are between $20-30 more.

[Shimejii]

If you can tune the memory you can expect it to be like in this video, or others around the same thing. On average tighter timings will be much better then faster speeds. a tight timing 6000 kit will be  pretty much all you need.

[Dedayog]

8 minutes ago, kitnoman said:

So I'm about to upgrade to 13600k. and I saw TEAMGROUP T-Create Classic DDR5 CL46 32GB Kit (2 x 16GB) 5600MHz. I'm working on no to less RGB build, so a 32gb ddr5 for $97 is a great deal for me. And how does that compare to G.Skill RipJaws S5 Series DDR5 6000 CL36 32GB (2 x 16GB) or RipJaws S5 Series DDR5 6000 CL30 32GB (2 x 16GB) or TEAMGROUP T-Force Vulcan Alpha DDR5 CL38 32GB Kit (2x16GB) 6000MHz or TEAMGROUP T-Create Expert Overclocking 10L DDR5 CL40 32GB Kit (2 x 16GB) 6400MHz and orsair Vengeance DDR5 C36 32GB (2x16GB) 6000MHz . Normally I always go with kingston/hyperx, but with their current pricing, the only thing I would consider is Kingston Fury Renegade Silver DDR5 CL32 32GB (2x16GB) 6400MT/s which is about $70-80 difference than the one I'm currently looking at, while the other ones are between $20-30 more.

Small difference.  Linus and others have done videos about.

 

Check YouTube.

[RONOTHAN##]

With 13th gen, frequency is everything, so with nearly every benchmark higher frequency is better. A 6400 CL40 kit should be better than a 6000 CL30 kit near exclusively, so you generally want to go for the highest rated kit you can find. 

 

In terms of timings, the number you actually want to look at for the overall latency is (tCL + tRCD + 8) * 2000 / data rate. That should be the overall first word latency. Going by that formula, a CL30-38-38 kit is only about 5% faster than a CL36-36-36 kit, completely unnoticeable, but a CL36 kit is 10% faster than a CL40-40-40 kit. The CL30 kits will overclock better because they're Hynix memory, odds are if you've got a good board you can get 7200+MHz, so if you want to do some overclocking in the future spend the extra on the CL30 kit, but if you're only ever gonna enable XMP and call it a day stick to the CL36 kit because you won't notice the difference. 

 

Out of those, I'd get the TeamGroup CL40 kit, it's the best value. 

[philmar]

52 minutes ago, RONOTHAN## said:

In terms of timings, the number you actually want to look at for the overall latency is (tCL + tRCD + 8) * 2000 / data rate. That should be the overall first word latency. Going by that formula, a CL30-38-38 kit is only about 5% faster than a CL36-36-36 kit, completely unnoticeable, but a CL36 kit is 10% faster than a CL40-40-40 kit.

I had to google it, so for those who ask what is tCL or tRCD:

[kitnoman]

This would be my first ddr5, but I think I have enough experience overclocking ddr4, probably just around amateur level. So I'm now thinking of getting either the one with the lowest cas latency or highest rated speed. Then I will try to overclock and tighten Both. In your experience, which is more likely, getting a CL30 kit with 6000mhz and try to oveclocking it to the highest allowed by the motherboard, which making sure that timings are tight and as close to cl30, or is it easier to get a 6400mhz/MTs(I barely understand the difference and which is more correct), then try to both overlock it and tighten the timings?
 

This is because the way everyone have said and after watching the review, it seems my first choice, the team group classic cl46 5600mhz is the worst for me. So depending on the game, it may have a big to little impact. However, just the review alone have some of the games I have and play does improved. The only thing, unlike the review, I have 1440p and 3080 and not 1080p and 4090(I wonder how those would translate when resolution is increase). Basically, which do you guys think is more achievable to the ram I've mentioned, 7200mhz 34-44-44-28 or 6000mhz 30-38-38-28? Why would I target this two settings? Because in the review above, This are the best tuned settings that gave the best performance.

 

Also I did mention that while I try to tune and overclock rams, I'm an amateur right. So can anyone tell why did the guys in hardware unboxed have a lower T-ras, compared to the other settings? I think it''s a bit late, but what is the general rule to tighten timings again? I mean I know lower is better, but I've always just paired t-rcd,t-rp and tras at the same digit or t-rcd and t-rp the same, but Tras was never lower than the rest. By the way thanks for  the replies. This actually gave me a lot of insights.

[RONOTHAN##]

First off, I want to say that you want to quote so we get a notification. I just happened to come back here. 

 

10 hours ago, kitnoman said:

In your experience, which is more likely, getting a CL30 kit with 6000mhz and try to oveclocking it to the highest allowed by the motherboard, which making sure that timings are tight and as close to cl30, or is it easier to get a 6400mhz/MTs(I barely understand the difference and which is more correct), then try to both overlock it and tighten the timings?

If you're gonna overclock them, all that matters is that you get Hynix-based DDR5. Both 6000 CL30 and 6400 CL40 are guaranteed Hynix, and that 5600 CL46 kit is about 50/50 whether it will be Hynix based. They're binned about the same, and as long as the kit was made within the past 4 months it would be capable of DDR5 8000 assuming your board and CPU can handle it (my kit is a 6000 CL30 Flare X5 kit, it does 7800 CL36 easy, it can probably do 8000 CL38 if I put more effort into the CPU side of the equation, and it can run benchmarks at 8400 CL34-45-44-32). I will want to point out that overclocking DDR5 is very difficult, especially on the 13th gen memory controller, so unless you're prepared to do a ton of trial and error to make sure that high frequencies are stable you will want to aim for something a bit more realistic like 6800MT/s (depends on the board though, some have 7600 be super easy to get, others struggle to get much above 6400). Plus there's Intel things like half the secondary timings don't actually exist and you should set the tertiary timings for them instead. 

 

10 hours ago, kitnoman said:

So can anyone tell why did the guys in hardware unboxed have a lower T-ras, compared to the other settings?

Hynix RAM does very low tRAS. That simple. Hynix A die, the stuff you're more likely to end up getting nowadays, in my experience is does require it to be a bit looser, usually I just run it equal to CAS latency, but it still does that timing rather low compared to the XMP specs of it. 

 

10 hours ago, kitnoman said:

I think it''s a bit late, but what is the general rule to tighten timings again? I mean I know lower is better, but I've always just paired t-rcd,t-rp and tras at the same digit or t-rcd and t-rp the same, but Tras was never lower than the rest.

On 11th gen and earlier Intel platforms, tRCD and tRP had to be set identically because of a limitation with the memory controller. On Hynix DDR5 though, with the primaries the only real rule is that CAS latency has to be even on pretty much every board but the Z690/Z790 Dark from EVGA. tRP usually does one tick lower than tRCD though that's not really guaranteed, and you can often trade tRP to tRAS as long as their sum adds up to the same number, but if you're lazy just leave them to the same value. 

 

The reason why tRAS was never lower than the other timings on DDR4 is that DDR4 had tRAS limits of 21 on Ryzen and 28 on Intel. Very little DDR4 wasn't able to hit those bottom limits, and most of the other primary timings did lower settings than 21 at most frequencies you're were likely to run. 

[shaolin95]

On 3/17/2023 at 9:50 PM, RONOTHAN## said:

With 13th gen, frequency is everything, so with nearly every benchmark higher frequency is better. A 6400 CL40 kit should be better than a 6000 CL30 kit near exclusively, so you generally want to go for the highest rated kit you can find. 

 

And yet the video posted here and many others prove you wrong.