I'd have loved to see 3600 but with very tight timings tested, that's where these will shine. I imagine these can achieve much better timings at 3600 than average DIMMs. Nobody is going to run these at 3600 C18 like you tested, I run my regular $80 set at 3600 C16. These will definitely do 3600 C14 or better.
This was my first thought when I read the test setup. No one is going to run these at 18-26-26-46 @3200 or 3600MT/sec. They are going to find the best stable timings at 3600MT/sec or run at 5000.
The CL14 kit of the Trident Z Neo is much much much more expensive than the regular C18 kit lol. And also virtually out of stock everywhere when I checked a few weeks ago, was just curious.
Yeah, I'd like to see Anandtech test extremely tight timings on this kit at 3600mhz.
However, review time is limited and expensive for the company, and I know firsthand that tweaking one timing down by 1, sitting through 2hr of memtest86, going back into bios, tweaking it down again, etc... Led to 2 ~ 3 weeks of recreational memory tuning before I dialed in the tightest possible settings for my CPU + MoBo + Memory kit that didn't result in errors in memtest86 for all 4 passes.
Also, I had dual-rank ram, and the ryzen memory timing calculator wasn't giving me stable timings (and yes, I did follow the correct process, it's just the utility likely isn't optimized for fairly rare dual-rank 16GBx2 memory kits), so I didn't have a good starting point for most timings to start tweaking down further than the already sparsely populated XMP profile that doesn't list a number for the majority of the smaller subtimings.
I don't think dual rank 16x2 memory kits are "fairly rare". In fact I've been running them for several years for a variety of purposes. Also, the x570 motherboards are claiming support for 128 GB RAM which would require 32x4 (havent seen any 64 GB desktop modules yet). NewEgg has 15 32x4 kits. If you are springing for a 3950X you may as well go all the way and max out the RAM too. :)
But no, the Ryzen Calc (at least at the time) wasn't providing me known good timings. I didn't even have Samsung B-Die, just Hynix CJR dual-rank memory. See the image here: https://i.imgur.com/iEb4Ctj.png The Ryzen calc was saying that a tCAS Latency of 14 at 3600mhz was stable. Absolutely not. Anything lower than 16 either wouldn't post or if it did post would reset itself back to JEDEC default timings or even more rarely, boot into memtest86 and spit out hundreds of errors. Same the withe tRFC values.
Because of the bad tCL and tRFC values (which I didn't know those specific timings were the ones that were causing me issues), the Ryzen calc's values weren't useful as a starting point to tuning my memory. I had to start from the 16-19-19-39 XMP values, and unfilled out subtimings to try to get where I got to in the end, and that takes a long-ass time of memtest86 reboots and saving BIOS profiles to quickly get back to the last working memory tuning profile, along with keeping an excel spreadsheet on another computer marking the last working setting for that subtiming and the last not-working subtiming value and what error I saw when I attempted to set that.
And yes, I even did multiple passes through the dozens of individual timings to see if i could tweak down previously tweaked timings lower after other timings were tweaked. My last pass through (before that image) was unsuccessful in lowering any single timing down by even 1 ns lower. So this is absolutely as low as it goes without going even more overkill on DRAM/SoC voltages, which were already overvolted.
If we assume 200MHz per tick of tCL, tRCD and tRP (worst case scenario), then at 3600MHz these sticks could do 11-19-19 at 1.50v. The high tRCD at 5000MHz is an IMC limitation so it could definitely go lower at more realistic frequencies.
Tighter timings lower mhz performs about the same as looser timings higher mhz. It's barely worth benchmarking, especially when there was such little variation in all the benchmarks.
I generally ignore these sort of halo products because most of the time they are just for willy wavers who want to own the very best..but normally even halo products have a tangible improvement in SOME scenario. These are just high score pieces, and would definitely judge the crap out of anyone who bought them.
Most of the time, these halo products are the door for newer cheaper and faster products for the masses in the next generation. As much as flamboyant and glamorous these products are, they have their uses.
Would be great memory for Renoir for Desktop once it becomes available. But then the price is certainly not in the same ballpark as the rest of the components. No one would spend this much on an APU based system for memory alone.
Would be interesting to see this test with a 3950X, specially in workstation scenarios where the 16 cores start to become limited by the dual-channel controller.
I don't understand. This is literally a performance memory preview, yet the memory heavy benchmark mode was declined in favor of a CPU-heavy benchmark?
"We report the results as the ability to simulate the data as a fraction of real-time, so anything above a ‘one’ is suitable for real-time work. Out of the two modes, a ‘non-firing’ mode which is DRAM heavy and a ‘firing’ mode which has CPU work, we choose the latter. Despite this, the benchmark is still affected by DRAM speed a fair amount."
While funny as a "what if" scenario, realistically nobody is going to pair a $150 APU with $1200 memory. AMD APUs likely couldnt push this stuff to 5000mhz anyway, hell most intel chips would struggle with 5000mhz.
The only thing this memory would benefit is an APU, it would be nice just to see what it would do. Nobody is going to buy this memory for any reason other than bragging rights so they might as well had tested it with a APU.
Is there any way you could run Prime95 on this? It is spectacularly memory limited (on an i9-7940X for the largest FFT sizes it is barely 50% faster on 14 cores than on 4), so a best-case test.
Never seen on any website comparison of dual channel versus quad channel versus 6-channel versus 8-channel memory architectures. Instead all testers pump and pump and pump for years everyone know barely useful MHz because memory companies pay for the ad. You forgot 10% and even 20% difference in PC computing (and 2x in supercomputing) is essentially equal to ZERO difference?
Take simple Gauss elimination Ax=B test from Intel XML library with AVX512 and compare different Intel processors, other tests and clear this question finally. May be even AMD processors which do not support ACX512 will shine here due to memory bandwidth.
Really puzzled by who would buy this outrageously expensive memory? The only use scenario I can imagine is in high-speed trading, where companies pay millions to be a few milliseconds ahead (for example, by getting their own fiber optic links to the CBE), but not sure if anyone in that field uses Ryzen systems.
Do these high speed RAM kits work at the advertised speeds when used with a Ryzen 3xxx series but on a B350 board? I know the memory controller is on the die, but i assume the traces on the board may also be a factor.
he wont.. he hates amd like no other.. will bash them till his fingers bleed.. but when called out on his fud and lies.. he just turns to insults and posts more bs.. he is and intel shill and still believes intels lies and bs.. he's all talk an no action.
It says in the intro that Intel CPUs can't use RAM this fast, and using a more powerful GPU would not alter the gaming benchmarks *because they are not sensitive to memory bandwidth*.
I figure it's a demonstration of what the hardware is capable of that doesn't really have a (practical) use case...yet. Not a new technology, just a stupid-fast result of binning higher-volume parts to show what they can do. I'll be interested to see where it leads.
Yeah the iGPU in the Ryzen 3700X would have screamed I tell you. Oh wait, my someone is telling me they don't have iGPU's. If you are going to complain at least read the article.
Sloppy, borderline laughable review. If you test lower clocks, lower the darn timings as well. THAT should be the whole point of this exercise not the infinity fabric.
Can't come to a conclusion when Zen2 itself will bottleneck fast GPUs and the 5700XT itself can't achieve high FPS to begin with. 5000MHz RAM won't be able to show any difference in these underwhelming conditions.
I think an update to this article with optimized timings at ALL frequencies would be a more real world test of the infinity fabric. I would include 4 stick testing just to show the performance increase, if any, and costs for quality memory and their configurations. A discussion on Samsung, Hynix, and Micron would also be useful IMO.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
54 Comments
Back to Article
LuckyX2 - Monday, January 27, 2020 - link
I'd have loved to see 3600 but with very tight timings tested, that's where these will shine. I imagine these can achieve much better timings at 3600 than average DIMMs. Nobody is going to run these at 3600 C18 like you tested, I run my regular $80 set at 3600 C16. These will definitely do 3600 C14 or better.Hulk - Monday, January 27, 2020 - link
This is exactly what I was thinking as I was reading the review.Sivar - Monday, January 27, 2020 - link
This was my first thought when I read the test setup. No one is going to run these at 18-26-26-46 @3200 or 3600MT/sec. They are going to find the best stable timings at 3600MT/sec or run at 5000.AnarchoPrimitiv - Monday, January 27, 2020 - link
Yeah, I've been saying for a while now that my dream kit would be 3600Mhz 14-14-14-34 timingsMakaveli - Monday, January 27, 2020 - link
I believe this is already doable with Samsung B-dieJoeyJoJo123 - Tuesday, January 28, 2020 - link
Samsung B-Die has been discontinued for a while.integralfx - Tuesday, January 28, 2020 - link
Only the OEM sticks have been discontinued. The IC themselves are still in production.FreckledTrout - Tuesday, January 28, 2020 - link
You can pick up something pretty close to that already set with XMP settings.Trident Z Neo - DDR4-3600MHz CL14-15-15-35 1.40V
BoemlauweBas - Thursday, January 30, 2020 - link
DIMM1: G Skill TridentZ RGB F4-3600C17-16GTZR @ 1801 Mhz (XMP) 17-19-19-39DIMM3: G Skill TridentZ RGB F4-3600C17-16GTZR @ 1801 Mhz (XMP) 17-19-19-39
..... I made a mistake didn't I ? Keep reading I should have gone for 4x8G
The RGB variant at 3600 will not go above 3800 :/
ArmedandDangerous - Friday, January 31, 2020 - link
The CL14 kit of the Trident Z Neo is much much much more expensive than the regular C18 kit lol. And also virtually out of stock everywhere when I checked a few weeks ago, was just curious.JoeyJoJo123 - Tuesday, January 28, 2020 - link
Yeah, I'd like to see Anandtech test extremely tight timings on this kit at 3600mhz.However, review time is limited and expensive for the company, and I know firsthand that tweaking one timing down by 1, sitting through 2hr of memtest86, going back into bios, tweaking it down again, etc... Led to 2 ~ 3 weeks of recreational memory tuning before I dialed in the tightest possible settings for my CPU + MoBo + Memory kit that didn't result in errors in memtest86 for all 4 passes.
JoeyJoJo123 - Tuesday, January 28, 2020 - link
Also, I had dual-rank ram, and the ryzen memory timing calculator wasn't giving me stable timings (and yes, I did follow the correct process, it's just the utility likely isn't optimized for fairly rare dual-rank 16GBx2 memory kits), so I didn't have a good starting point for most timings to start tweaking down further than the already sparsely populated XMP profile that doesn't list a number for the majority of the smaller subtimings.Ratman6161 - Tuesday, January 28, 2020 - link
I don't think dual rank 16x2 memory kits are "fairly rare". In fact I've been running them for several years for a variety of purposes. Also, the x570 motherboards are claiming support for 128 GB RAM which would require 32x4 (havent seen any 64 GB desktop modules yet). NewEgg has 15 32x4 kits. If you are springing for a 3950X you may as well go all the way and max out the RAM too. :)JoeyJoJo123 - Wednesday, January 29, 2020 - link
I went for a 3700x. Also, take a look here:https://www.reddit.com/r/Amd/comments/cw7ysm/32gb_...
But no, the Ryzen Calc (at least at the time) wasn't providing me known good timings. I didn't even have Samsung B-Die, just Hynix CJR dual-rank memory. See the image here: https://i.imgur.com/iEb4Ctj.png
The Ryzen calc was saying that a tCAS Latency of 14 at 3600mhz was stable. Absolutely not. Anything lower than 16 either wouldn't post or if it did post would reset itself back to JEDEC default timings or even more rarely, boot into memtest86 and spit out hundreds of errors. Same the withe tRFC values.
Because of the bad tCL and tRFC values (which I didn't know those specific timings were the ones that were causing me issues), the Ryzen calc's values weren't useful as a starting point to tuning my memory. I had to start from the 16-19-19-39 XMP values, and unfilled out subtimings to try to get where I got to in the end, and that takes a long-ass time of memtest86 reboots and saving BIOS profiles to quickly get back to the last working memory tuning profile, along with keeping an excel spreadsheet on another computer marking the last working setting for that subtiming and the last not-working subtiming value and what error I saw when I attempted to set that.
And yes, I even did multiple passes through the dozens of individual timings to see if i could tweak down previously tweaked timings lower after other timings were tweaked. My last pass through (before that image) was unsuccessful in lowering any single timing down by even 1 ns lower. So this is absolutely as low as it goes without going even more overkill on DRAM/SoC voltages, which were already overvolted.
integralfx - Tuesday, January 28, 2020 - link
If we assume 200MHz per tick of tCL, tRCD and tRP (worst case scenario), then at 3600MHz these sticks could do 11-19-19 at 1.50v. The high tRCD at 5000MHz is an IMC limitation so it could definitely go lower at more realistic frequencies.ses1984 - Wednesday, January 29, 2020 - link
Tighter timings lower mhz performs about the same as looser timings higher mhz. It's barely worth benchmarking, especially when there was such little variation in all the benchmarks.JlHADJOE - Thursday, January 30, 2020 - link
5000MHz @ CAS 18 = 3600MHz @ CAS 13bug77 - Monday, January 27, 2020 - link
It would be so cool seeing one cheaping out on a dGPU, only to spend $1,200 on RAM to make the iGPU faster.Hectandan - Monday, January 27, 2020 - link
What if they are assembling a sub-3L desktop? Can't fit a GPU thereTheinsanegamerN - Monday, January 27, 2020 - link
In such a small machine, you wouldnt be able to use any iGPU that could use 5000mhz RAM without overheating either. What's your point?PeachNCream - Monday, January 27, 2020 - link
Orly?https://www.anandtech.com/show/15414/ces-2020-zota...
Tunnah - Monday, January 27, 2020 - link
I generally ignore these sort of halo products because most of the time they are just for willy wavers who want to own the very best..but normally even halo products have a tangible improvement in SOME scenario. These are just high score pieces, and would definitely judge the crap out of anyone who bought them.tamalero - Monday, January 27, 2020 - link
Most of the time, these halo products are the door for newer cheaper and faster products for the masses in the next generation. As much as flamboyant and glamorous these products are, they have their uses.Spunjji - Wednesday, January 29, 2020 - link
Agreed - this is just silly.5080 - Monday, January 27, 2020 - link
Would be great memory for Renoir for Desktop once it becomes available. But then the price is certainly not in the same ballpark as the rest of the components. No one would spend this much on an APU based system for memory alone.PeachNCream - Monday, January 27, 2020 - link
That's a whole lot of coin for a barely measurable improvement in a few benchmarks.ahtoh - Monday, January 27, 2020 - link
The most important graph: derivative of performance/$$ is missingPVG - Monday, January 27, 2020 - link
Would be interesting to see this test with a 3950X, specially in workstation scenarios where the 16 cores start to become limited by the dual-channel controller.Sivar - Monday, January 27, 2020 - link
I don't understand. This is literally a performance memory preview, yet the memory heavy benchmark mode was declined in favor of a CPU-heavy benchmark?"We report the results as the ability to simulate the data as a fraction of real-time, so anything above a ‘one’ is suitable for real-time work. Out of the two modes, a ‘non-firing’ mode which is DRAM heavy and a ‘firing’ mode which has CPU work, we choose the latter. Despite this, the benchmark is still affected by DRAM speed a fair amount."
29a - Tuesday, January 28, 2020 - link
This article is full of all kinds of stupid. They should have test iGPU performance too.TheinsanegamerN - Tuesday, January 28, 2020 - link
While funny as a "what if" scenario, realistically nobody is going to pair a $150 APU with $1200 memory. AMD APUs likely couldnt push this stuff to 5000mhz anyway, hell most intel chips would struggle with 5000mhz.29a - Wednesday, January 29, 2020 - link
The only thing this memory would benefit is an APU, it would be nice just to see what it would do. Nobody is going to buy this memory for any reason other than bragging rights so they might as well had tested it with a APU.TomWomack - Monday, January 27, 2020 - link
Is there any way you could run Prime95 on this? It is spectacularly memory limited (on an i9-7940X for the largest FFT sizes it is barely 50% faster on 14 cores than on 4), so a best-case test.Alistair - Monday, January 27, 2020 - link
Seems like there is no point to this when buying a Threadripper and getting more memory channels gets you better results for less money.SanX - Monday, January 27, 2020 - link
Never seen on any website comparison of dual channel versus quad channel versus 6-channel versus 8-channel memory architectures. Instead all testers pump and pump and pump for years everyone know barely useful MHz because memory companies pay for the ad. You forgot 10% and even 20% difference in PC computing (and 2x in supercomputing) is essentially equal to ZERO difference?Take simple Gauss elimination Ax=B test from Intel XML library with AVX512 and compare different Intel processors, other tests and clear this question finally. May be even AMD processors which do not support ACX512 will shine here due to memory bandwidth.
eastcoast_pete - Monday, January 27, 2020 - link
Really puzzled by who would buy this outrageously expensive memory? The only use scenario I can imagine is in high-speed trading, where companies pay millions to be a few milliseconds ahead (for example, by getting their own fiber optic links to the CBE), but not sure if anyone in that field uses Ryzen systems.deneb - Tuesday, January 28, 2020 - link
This test would be great with threadrippers! Especially the 64 core.ManuelDiego - Tuesday, January 28, 2020 - link
Do these high speed RAM kits work at the advertised speeds when used with a Ryzen 3xxx series but on a B350 board? I know the memory controller is on the die, but i assume the traces on the board may also be a factor.Maxiking - Tuesday, January 28, 2020 - link
Could you please tell what sane person test impact of RAMS with AMDslow cpus and slow gpus? 5700xt? really?Did you outsource reviews to India?
catavalon21 - Tuesday, January 28, 2020 - link
Feel free to jump right in and write a better one.Qasar - Tuesday, January 28, 2020 - link
he wont.. he hates amd like no other.. will bash them till his fingers bleed.. but when called out on his fud and lies.. he just turns to insults and posts more bs.. he is and intel shill and still believes intels lies and bs.. he's all talk an no action.Spunjji - Wednesday, January 29, 2020 - link
It says in the intro that Intel CPUs can't use RAM this fast, and using a more powerful GPU would not alter the gaming benchmarks *because they are not sensitive to memory bandwidth*.Did you outsource your brain to a herring?
HikariWS - Tuesday, January 28, 2020 - link
Awesome article, congrats!Indeed, this RAM isn't worth it. It just won't be the bottleneck for the rig of ppl who would be willing to pay for it.
catavalon21 - Tuesday, January 28, 2020 - link
I figure it's a demonstration of what the hardware is capable of that doesn't really have a (practical) use case...yet. Not a new technology, just a stupid-fast result of binning higher-volume parts to show what they can do. I'll be interested to see where it leads.29a - Tuesday, January 28, 2020 - link
I can't believe you didn't test the one thing that would show a big difference, gaming on the iGPU.FreckledTrout - Tuesday, January 28, 2020 - link
Yeah the iGPU in the Ryzen 3700X would have screamed I tell you. Oh wait, my someone is telling me they don't have iGPU's. If you are going to complain at least read the article.29a - Wednesday, January 29, 2020 - link
You really don't think I was talking about a different CPU? Are your really that stupid or just intellectually dishonest, it's one or the other.Korguz - Wednesday, January 29, 2020 - link
instead of being rude, why not mention which igpu you are referring to ?? instead... you resort to insults...Iketh - Saturday, February 1, 2020 - link
lol why must he refer to one? ANY IGPUKorguz - Sunday, February 2, 2020 - link
maybe he has a specific one in mind..and there is this " You really don't think I was talking about a different CPU ?"
RobJoy - Friday, January 31, 2020 - link
Sloppy, borderline laughable review.If you test lower clocks, lower the darn timings as well.
THAT should be the whole point of this exercise not the infinity fabric.
dwade123 - Friday, January 31, 2020 - link
Can't come to a conclusion when Zen2 itself will bottleneck fast GPUs and the 5700XT itself can't achieve high FPS to begin with. 5000MHz RAM won't be able to show any difference in these underwhelming conditions.NOTELLN - Wednesday, June 24, 2020 - link
I think an update to this article with optimized timings at ALL frequencies would be a more real world test of the infinity fabric. I would include 4 stick testing just to show the performance increase, if any, and costs for quality memory and their configurations. A discussion on Samsung, Hynix, and Micron would also be useful IMO.