Corsair Vengeance RGB DDR5-6000 CL30 (AMD Expo) 2x 16 GB

[ir_cow]

Corsair's Vengeance RGB takes the AMD DDR5 platform to a sweet spot. Using the new EXPO 6000 MT/s profile that's optimized for Zen 4 and tight timings of 30-36-36-76, this kit is designed to fit right into the Corsair ecosystem to provide a great end-to-end user experience.

Show full review

 

[Garrus]

Got to get that 5600C36 stuff on the charts. The one ram that is in mass supply, and the best pricing, is not in the charts

5600C36 is what everyone is considering to buy.

 

[igralec84]

Will try 30-36-36-76-112 on my Kingstor Fury Beast 5600 CL40 since it's both SK Hynix, maybe it'll work as i'm running 6000 CL32 atm

 

[ir_cow]

Will try 30-36-36-76-112 on my Kingstor Fury Beast 5600 CL40 since it's both SK Hynix, maybe it'll work as i'm running 6000 CL32 atm

99% likely Hynix M-Die if it is SK Hynix. So 28-36-36-30-50 with 1.5V easy to do. Might be able to get away with 1.45v if the bin is good enough. Millage may vary and overclocking voids the warranty. So I do not recommend it. aka don't blame me.

The Corsair kit this review overclock was with 1.4v (same as EXPO). You can get lucky with nice ICs, but often its 1.5v I've found.

 

[Braegnok]

I have two G.SKILL Z5 Neo kits,.. the 6000 CL30, and the 5600 CL28, both run @ 1.35V and are Hynix M-Die.

The Corsair kit has tighter EXPO timings, they were out of stock when I ordered the Neo kits.

 

[ir_cow]

I have two G.SKILL Z5 Neo kits,.. the 6000 CL30, and the 5600 CL28, both run @ 1.35V and are Hynix M-Die.

The Corsair kit has tighter EXPO timings, they were out of stock when I ordered the Neo kits.

Hmmm. I didn't see any difference in the timings besides AUTO for EXPO ignoring some of the SPD data and doing whatever it wanted. I just manually typed in the values to keep it consistent. XMP didn't have that problem. This is the GSKILL Fare X5 kit. It has XMP and EXPO.

Edit: I don't think I've tried 1.35v for 6000 CL30 yet. Not with any of these newer kits. I should give it a go. At least with 1.5v and you can always lower that until it stops booting or produces errors.

 

[Braegnok]

Hmmm. I didn't see any difference in the timings besides AUTO for EXPO ignoring some of the SPD data and doing whatever it wanted. I just manually typed in the values to keep it consistent. XMP didn't have that problem. This is the GSKILL Fare X5 kit. It has XMP and EXPO.

Edit: I don't think I've tried 1.35v for 6000 CL30 yet. Not with any of these newer kits. I should give it a go. At least with 1.5v and you can always lower that until it stops booting or produces errors.

The Corsair EXPO profile is 6000 @ 30-36-36-76-112 1.4v.

My Neo Z5 EXPO profile is 6000 @ 30-38-38-96-134 1.35v.

 

[ADB1979]

Having admittedly only flicked through this review, reading important points, looking at setting and graphs, I was bewildered to see that the games performance charts are only in 1080p.!

What I (and I guess) most gamers want to know, is how does this RAM perform in this setup in 1440P and 2160p compared to alternatives such as the cheapest available DDR5, the cheapest "performance" DDR5 and so on.

 

[The King]

Yay! Zentimings

Will it be possible to state whether the RAM is M-die or A-die under the specification table?

IC Manufacturer:	SK Hynix (M-die)

 

[ir_cow]

Having admittedly only flicked through this review, reading important points, looking at setting and graphs, I was bewildered to see that the games performance charts are only in 1080p.!

What I (and I guess) most gamers want to know, is how does this RAM perform in this setup in 1440P and 2160p compared to alternatives such as the cheapest available DDR5, the cheapest "performance" DDR5 and so on.

I should probably dedicate a whole page in the reviews to why these higher resolutions are not testing. The answer is simply because when the game becomes GPU bound, memory speed has such a small impact on frame rate, it becomes hard to distinguish the between them all. . collect enough benchmark data and margin shrinks even more. This was only 2 runs for each averaged. The 3080 TI I use is the limiting factor. Different graphics card or CPU will give a different outcome. Though, the statement never changes. Memory have the most impact when the CPU is the limiting factor in frame generation.

4K benchmarks.

Will it be possible to state whether the RAM is M-die or A-die under the specification table?

IC Manufacturer:	SK Hynix (M-die)

I thought of this before and the only reason I haven't added it was because these ICs can change at any time. Certain freqs must be Hynix or Samsung. Like for example you will never see a Micron kit at 6000+. This is why I try not to weight the overclocking results in the overall review outcome. Its nice to have good performing memory, but the sold binned is what the review is based around. I might add this anyways and put a small print at the bottom saying to not count on the memory being the same as what I have.

 

[ADB1979]

I should probably dedicate a whole page in the reviews to why these higher resolutions are not testing. The answer is simply because when the game becomes GPU bound, memory speed has such a small impact on frame rate, it becomes hard to distinguish the between them all. . collect enough benchmark data and margin shrinks even more. This was only 2 runs for each averaged. The 3080 TI I use is the limiting factor. Different graphics card or CPU will give a different outcome. Though, the statement never changes. Memory have the most impact when the CPU is the limiting factor in frame generation.

Thank you, a great answer

 

[igralec84]

Set my Kingston Fury Beast to 6000 30-36-36-76-112 and got the following Aida64 scores with the 7600X:

Memory Read: 54452 MB/s
Memory Write: 81535 MB/s
Memory Copy: 58813 MB/s
Memory Latency: 71.2 ns

The read and copy are quite bad, but never checked what it does without the OC I did improve in TimeSpy and Port Royal though

 

[ir_cow]

@igralec84 that is normal. You only have one chiplet active on the CPU. Its just how AMD has designed the memory throughput. It seemly has no impact on anything outside of throughput workloads. The scientific stuff.

 

[igralec84]

@igralec84 that is normal. You only have one chiplet active on the CPU. Its just how AMD has designed the memory throughput. It seemly has no impact on anything outside of throughput workloads. The scientific stuff.

I have upgraded to the 7950x now and memory read is still very low, barely get 70GB/s.

Any ideas what to look at, i see 6000 CL30 kits getting 77-80GB/s. Read the GEIL review on Guru3D and it seems 6200 CL34 set to 6000 CL34 got 81GB/s on the same motherboard with a 7900X

Might be the lack of EXPO and some additional settings it sets, as i select XMP for the 5600 CL40 profile, then just change speed to 6000 and the timings to 30-36-36-76-112 along with voltage to 1.4v and that's that.

Not that it affects gaming or anything, could probably run the XMP and forget about it

 

[ir_cow]

Different

I have upgraded to the 7950x now and memory read is still very low, barely get 70GB/s.

Any ideas what to look at, i see 6000 CL30 kits getting 77-80GB/s. Read the GEIL review on Guru3D and it seems 6200 CL34 set to 6000 CL34 got 81GB/s on the same motherboard with a 7900X

Different AIDA64 version, Windows version or maybe stuff is running in the background for you. Not to throw shade at Guru3D, but its hard to compare apples to apples between sites. Only way I get 80 GB/s is either a healthy tuning of the timings or using dual rank memory. Some of those MBs have those extra sub-timings adjusted by default. That can have a large impact on read throughput.

 

[igralec84]

Different

Different AIDA64 version, Windows version or maybe stuff is running in the background for you. Not to throw shade at Guru3D, but its hard to compare apples to apples between sites. Only way I get 80 GB/s is either a healthy tuning of the timings or using dual rank memory. Some of those MBs have those extra sub-timings adjusted by default. That can have a large impact on read throughput.

View attachment 274470

I did have firefox running, but didn't affect much, maybe latency is 67 without it and 69 with it in the background. Ah, Guru is running win10, so that might explain it and it's basically a downclocked EXPO profile that they tested, so probably no use comparing.

Might be missing out on that 10% due to sub timings then as i only change the primary ones, would need to take a look at what that corsair kit is running and copy all of them if i can. All the other speeds seem inline with the ones from your review, copy and write maybe a little higher except for latency that is a bit worse.

 

[ir_cow]

Also that Giel kit on Guru3d is 6200 and the ASRock MB wasn't using the default timings, it had "Aggressive mode" enabled. Kinda defeats the purpose of the whole review.... Mystery solved.

 

[igralec84]

Actually i have figured out the "problem" and it's quite lame. If the PC goes to sleep and wakes up, the memory read score will be 69-70k. After a reboot it's 77-78k, with the competitive timings preset it's 80-81k but gets an error at 1100-1600%. Maybe if i could get zen timings screenshot of a 6000 CL30 EXPO profile, i could probably get up to 80k, but yeah, it's the sleep that drops the score