TeamGroup Xtreem DDR4-4000 CL15 2x 16 GB Engineering Sample

[Mussels]

I don't tend to buy much ASUS. Their high-end stuff is great but overpriced and their entry-level stuff is typically worse than Gigabyte/MSI/Asrock.

For MSI and Aorus I usually see the auto divider doing what AMD claimed was standard behaviour at Zen2's launch:

View attachment 252212

It probably comes down to the tiny differences in how the boards set up the RAM dividers and such internally, as much as anything. I'll agree to just say dont trust auto past 3600

 

[zlobby]

I need to work on my memory timings.

View attachment 252182

Well, d'oh! Timings are way more crucial than bandwidth when it comes to real-time applications like games.

 

[mechtech]

Hmmmm interesting

so bdie nice
Dual rank nice

jedec 2400 instead of 3200
8 layer pcb instead of 10
And that 1.50v!!

Maybe should lower voltage and make 3600. Would be easier to get 1:1 then and 1T cmd??

 

[Snoota]

Me looking with my 32GB Ripjaws V 4000c16 kits that do 1.5v vDIMM 14-15-14-28:

 

[ir_cow]

I don't see 3800 CL15/16 in the AMD chart, only the Intel chart has 3800 CL15?

Its not updated yet. Still need to benchmark it. Might be a few days depending because I'm already onto the next thing.

 

[Braegnok]

Personally my best DDR4 memory kit for overclocking is Hynix DJR kit.

5866MHz CL21 daily settings, 6133MHz for benching.

 

[Chrispy_]

Well, d'oh! Timings are way more crucial than bandwidth when it comes to real-time applications like games.

Absolute timings in nanoseconds yes, and that's secondary to FCLK.
His problem IIRC is that Corsair Vengeance 3200 is garbage-tier quality that most motherboards struggle to memory train much faster than CL18. So 3800 CL22 is just as fast, with a 300MHz FCLK boost which is huge.

Getting Corsair Vengeance stable at even 3200 CL16 with manual timing override is hard on some boards, because it's the cheapest RAM you can buy in the UK for a reason. The quantity of it I've RMA'd because it's faulty at JEDEC 2133 timings is disgraceful and I refuse to buy Corsair RAM now, based solely on a few dozen kits of Vengance LPX. Never has any other brand failed so much and so hard.

 

[dalekdukesboy]

So question... Why didn't they try overclocking memory titled "over clockers" dream?

 

[ir_cow]

So question... Why didn't they try overclocking memory titled "over clockers" dream?

Short answer it's just a title. You clicked on it right?

Long answer for the title is a bit more complicated. Basically it is hand-picked Samsung-B die ICs binned to 4000 MT/s, so this memory kit already has a head start in the tweaking process. I pretty much do not have to validate that this memory will run at 4000 MT/s (even though I did anyways). When you have already well pre-binned memory its a "overclockers dream". But I wouldn't put so my thought into the title. It's just a title to get people to read the article.

If your asking why no overclocking was done, well that is worth an entire separate article. It really is. The article already touched on some of the subjects like Gear Ratio and Command Rates and overall timings. It also would be overwhelming to the reader to put it all in one article I think. Given that overclocking is never guaranteed so it will be a one-off and not directly transferable to another kit. The principles applies, but a copy-paste of timings, settings and voltages would not work out at all.

 

[ThrashZone]

Hi,
So what makes you think you need water on these sticks
You say nothing about how hot they get from testing.

 

[ir_cow]

Hi,
So what makes you think you need water on these sticks
You say nothing about how hot they get from testing.

I don't need water per say, but I won't go above 1.55v without it. Given that these are dual-rank and one side is just plastic, It's a waste of time to lower the timings if it will just error out from temps anyways. If I'm spending time on tweaking, might as well put a waterblock on and go all out.

 

[nexxusty]

When you lot have finished messing about with my ram, send it over please. Thanks

My Corsair scrap ram is currently running at 3800mhz, but CL22. Is that alright or should I start a new thread on ram ocing?

That's beyond terrible. LOL.

Well, d'oh! Timings are way more crucial than bandwidth when it comes to real-time applications like games.

If you're above 45gb/s, yes.

Personally my best DDR4 memory kit for overclocking is Hynix DJR kit.

View attachment 252265

5866MHz CL21 daily settings, 6133MHz for benching.

View attachment 252266

I really need to put the waterblock and heatsinks on my 4x8GB B-Die kit. They do 3866mhz 1:1 14-14-14-30, with a TRFC of 220 in a 4x8GB configuration @ 1.54v on my 11900k. They're BEGGING to be let loose.

I won't get much more in 4x8GB configuration (I'd love 4000mhz 1:1 though, NGL), however 2x8GB.... I bet I can get to 4800mhz at least.

I really want a Hynix DJR kit though after I max these out.

 

[The King]

Its not updated yet. Still need to benchmark it. Might be a few days depending because I'm already onto the next thing.

Thank you for taking the time to rebench. If possible or allowed by TeamGroup, Zentimings screenshot would be nice.

Personally my best DDR4 memory kit for overclocking is Hynix DJR kit.

5866MHz CL21 daily settings, 6133MHz for benching.

Im curious has to what you get in AIDA64 compared to DDR5 running at similar speeds. Could you post in this thread your AIDA64 numbers?

Share your AIDA 64 cache and memory benchmark here

Hey folks, how about sharing some cache and memory benchmark you get using AIDA 64? Rules: Use the latest stable(non alpha/beta) version of AIDA(currently 4.50.3000) So here goes mine. All the best. :) Table as requested. In order of entry.... Only the best scores of an individual are...

www.techpowerup.com

 

[dalekdukesboy]

Thank you for taking the time to rebench. If possible or allowed by TeamGroup, Zentimings screenshot would be nice.


Im curious has to what you get in AIDA64 compared to DDR5 running at similar speeds. Could you post in this thread your AIDA64 numbers?

Share your AIDA 64 cache and memory benchmark here

Hey folks, how about sharing some cache and memory benchmark you get using AIDA 64? Rules: Use the latest stable(non alpha/beta) version of AIDA(currently 4.50.3000) So here goes mine. All the best. :) Table as requested. In order of entry.... Only the best scores of an individual are...

www.techpowerup.com

This! I have not seem many places where a high frequency, tight timing kit was bench marked with aida for bandwidth.

 

[Braegnok]

Thank you for taking the time to rebench. If possible or allowed by TeamGroup, Zentimings screenshot would be nice.


Im curious has to what you get in AIDA64 compared to DDR5 running at similar speeds. Could you post in this thread your AIDA64 numbers?

Share your AIDA 64 cache and memory benchmark here

Hey folks, how about sharing some cache and memory benchmark you get using AIDA 64? Rules: Use the latest stable(non alpha/beta) version of AIDA(currently 4.50.3000) So here goes mine. All the best. :) Table as requested. In order of entry.... Only the best scores of an individual are...

www.techpowerup.com

The 5866MHz profile @ CR2, 1.74V with Latency 52.0 ns.

This is an old screenshot from my photo's cloud storage, thus AIDA64 v6.32.5649 is from back in 2021. I don't remember if AIDA64 was compatible with the 5333MHz kit or was accurate.

The XIII Apex system has been replaced with Z690 Apex for about 6-months,.. so only screenshots I have are from the cloud.

Found an old Port Royal run from November 18, 2021: https://www.3dmark.com/pr/1315847

 

[Mussels]

Absolute timings in nanoseconds yes, and that's secondary to FCLK.
His problem IIRC is that Corsair Vengeance 3200 is garbage-tier quality that most motherboards struggle to memory train much faster than CL18. So 3800 CL22 is just as fast, with a 300MHz FCLK boost which is huge.

Getting Corsair Vengeance stable at even 3200 CL16 with manual timing override is hard on some boards, because it's the cheapest RAM you can buy in the UK for a reason. The quantity of it I've RMA'd because it's faulty at JEDEC 2133 timings is disgraceful and I refuse to buy Corsair RAM now, based solely on a few dozen kits of Vengance LPX. Never has any other brand failed so much and so hard.

Just to be clear: you forgot "LPX" a lot in there. The Regular corsair vengeance stuff was often single rank and/or even timings, and had no issues.

The reason corsair LPX was a problem back then was a combination of dual rank 8GB sticks, and the even timings which AMD would gear-down to even latencies awkwardly, which could throw the below chart a little off if the RAM didn't like the change

Zen 1 was officially rated to 2400Mhz with four ranks
Because the RAM was cheap a lot of people slapped in four of them, and wondered why it didn't run at 2667+
(Intel got around this by just locking users to 2133 and 2400 on 6th and 7th gen, AMD tried let you overclock but forgot how non-tech savvy users are)

Look how far ryzen has come. 2400Mhz to 4000Mhz, technically possible on the same motherboard (x370)

 

[ACE76]

This is just insane...DDR4 4000 at CL15. Here I was so happy to have 3600mhz at 14-14-14-34.

 

[dalekdukesboy]

The 5866MHz profile @ CR2, 1.74V with Latency 52.0 ns.

View attachment 252370

This is an old screenshot from my photo's cloud storage, thus AIDA64 v6.32.5649 is from back in 2021. I don't remember if AIDA64 was compatible with the 5333MHz kit or was accurate.

The XIII Apex system has been replaced with Z690 Apex for about 6-months,.. so only screenshots I have are from the cloud.

Found an old Port Royal run from November 18, 2021: https://www.3dmark.com/pr/1315847

Thanks! However why is the bandwidth only 20k in dual channel?

 

[Mussels]

Thanks! However why is the bandwidth only 20k in dual channel?

i was wondering that too, they seem so low

 

[Braegnok]

Thanks! However why is the bandwidth only 20k in dual channel?

I remember most hardware monitoring software would show inaccurate results with that kit when running above 5333MHz. The HWiNFO64 screenshot shows Command Rate as 4T, AIDI64 shows CR2 and the crazy 20K bandwidth,.. I think it was Gear-2 when the inaccurate results happened, months later the next version fixed the issue with Gear-2 showing crazy results.

Running in Gear-1 at 5333MHz XMP 2.0 the software worked fine.

 

[The King]

Its not updated yet. Still need to benchmark it. Might be a few days depending because I'm already onto the next thing.

Has there been any progress on this? Thanks

 

[ir_cow]

Has there been any progress on this? Thanks

Sorry no. I've havent had time to go back and revisit this. Once a article is published, thats usually final.