Cumulative Ryzen 5000 series TPU review discussion thread

[MxPhenom 216]

G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin PC RAM DDR4 4000 (PC4 32000) Desktop Memory Model F4-4000C15D-16GVK - Newegg.com

Buy G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin PC RAM DDR4 4000 (PC4 32000) Desktop Memory Model F4-4000C15D-16GVK with fast shipping and top-rated customer service. Once you know, you Newegg!

www.newegg.com

it would be nice to see a review with this ram. 4000 cas 15... only 170 bucks...

I tried to get my cousin to get that RAM for his 5800x that he wants to get, but he was like "No i want 32GB for under $200" so he just went with Trident Z 3200 CAS15 (B-Die) and plans to take it to 3600mhz or higher.

He initially bought that Ripjaw ram but 3600 16-19-19-39 (Not B-Die from what i could find) timings and I said return it and get the Trident Z.

 

[TheLostSwede]

Exactly. It remains to be seen if the review samples are cherry picked, or how much BIOS can be improved.

It improved hugely on the Ryzen 3000 series, although this might have been down to the whole UEFI/AGESA being quite immature at launch.
From what I've been told, 3866MHz 1:1 is quite possible on more recent 3000-series parts. Not a huge improvement, but AMD hasn't actively done anything to improve it, it's mostly just down to TSMC fine tuning their process.

 

[SL2]

What people seem to be misunderstanding is that the referenced DDR4-3800 frequency AMD mentions in the slide isn't the sweet spot, it's what you could hope for.

Like I said, I NEVER referred to that slide. Thät Låsst Swede has a photographic memory of old slides..
(Or maybe I did, unconsciously. )

I guess you never saw the slide below before then? It's what's normally referred to when talking about Ryzen 3000 memory speeds and why a lot of people are getting hung up on the whole 3733MHz thing.

I clearly remember that latency graph, but since I never got one of those CPU's I never looked into it anymore.

From what I've been told, 3866MHz 1:1 is quite possible on more recent 3000-series parts. Not a huge improvement, but AMD hasn't actively done anything to improve it, it's mostly just down to TSMC fine tuning their process.

The IO chip is still the same, right?

 

[TheLostSwede]

Like I said, I NEVER referred to that slide. Thät Låsst Swede has a photographic memory of old slides..
(Or maybe I did, unconsciously. )


I clearly remember that latency graph, but since I never got one of those CPU's I never looked into it anymore.


The IO chip is still the same, right?

Fair enough, but please don't abuse the Swedish language like that.

Maybe not quite the same, as I just heard that the new X570 boards that are launching are going to use a new revision of the chipsets. That's how Asus did their passively cooled board.
Since the chipset and I/O die are the same, that suggests that there has also been some improvements done over at GloFo.

 

[SL2]

The found 1,3% performance difference (average on several use case)

Grazie.
Those three games tested.. 7 - 9 % in one game is a bit surprising, I was expecting less. Only one game tho.

Fair enough, but please don't abuse the Swedish language like that.

Sorry about that. I can see that you spend a lot of time reading English as you seem to forget that that the language I was abusing wasn't Swedish.

Maybe not quite the same, as I just heard that the new X570 boards that are launching are going to use a new revision of the chipsets. That's how Asus did their passively cooled board.
Since the chipset and I/O die are the same, that suggests that there has also been some improvements done over at GloFo.

I for one like that. There's no need for a new chipset, the old one got everything you need for mainstream desktop.

 

[milewski1015]

Like I said, I NEVER referred to that slide. Thät Låsst Swede has a photographic memory of old slides..
(Or maybe I did, unconsciously. )

I think LostSwede and I assumed you were referring to the slide W1zz and I both linked. That's the only official slide from AMD I can think of that mentions DDR4-4000, so we were confused (I was anyway) as to where you were getting "4000 MHz RAM, aka the sweet spot according to AMD?". As Swede mentioned, I think part of the issue is the different interpretations of "sweet spot": price to performance sweet spot, max 1:1:1 ratio sweet spot, etc.

 

[birdie]

The Linus Tech Tips channel which I used to watch occasionally, is now openly lying to cover Ryzen 5000 price increase:

Yeah, the Ryzen 7 3700X which was introduced at $330 and now costs on average around $310 (and occasionally is being sold for as low as $250), has suddenly become $90 more expensive according to Linus Tech Tips (with over 12 million subscribers) - the channel has openly become very pro-AMD recently to the point where they criticize or/and underplay anything released by NVIDIA/Intel: "I'm DONE covering for NVIDIA - RTX 3070 Review".

And Linus conveniently forgot about the existence of 3600 which was introduced at $200 (and now can be have for less that $160). Yeah, "just a measly $50 price increase" when reality it's starts from 50%. Oh, and the Ryzen 5 3600X costs $250, not $300! AMD does love fat margins!

 

[SL2]

I think LostSwede and I assumed you were referring to the slide W1zz and I both linked. That's the only official slide from AMD I can think of that mentions DDR4-4000, so we were confused (I was anyway) as to where you were getting "4000 MHz RAM, aka the sweet spot according to AMD?". As Swede mentioned, I think part of the issue is the different interpretations of "sweet spot": price to performance sweet spot, max 1:1:1 ratio sweet spot, etc.

Yeah it was clearly my own interpretation of what I actually read before in said slide. Now I get how "sweet spot" can get twisted after that older slide.

 

[W1zzard]

The IO chip is still the same, right?

It is exactly the same, we checked with AMD, also mentioned in all reviews near the core config page

 

[birdie]

It is exactly the same, we checked with AMD, also mentioned in all reviews near the core config page

It's exactly the same meaning it's still based on the 12nm node? I wonder how they've managed to shake off a few watts in idle.

 

[W1zzard]

It is 100% identical physically.. from the same assembly line and same production run. All power improvements come from the 7 nm CCDs, and of course software power management, which of course can affect IO die power, too.

 

[HD64G]

AMD Ryzen 9 5900X + Ryzen 9 5950X Dominate On Linux Review - Phoronix

www.phoronix.com

https://openbenchmarking.org/embed.php?i=2011040-FI-AMDRYZEN976&sha=c4b0d35106fe&p=2

Seeing that chart in the link above makes me think that AMD does offer much higher vfm in every one of their new Zen3 processors apart from the 5600XT ($250 would make it invinsible for gaming PCs) when compared to Zen2 MSRP pricing. And discounts for the latter will come round later for the new CPUs as usual.

 

[Rahnak]

The Linus Tech Tips channel which I used to watch occasionally, is now openly lying to cover Ryzen 5000 price increase:

View attachment 174538

Yeah, the Ryzen 7 3700X which was introduced at $330 and now costs on average around $310 (and occasionally is being sold for as low as $250), has suddenly become $90 more expensive according to Linus Tech Tips (with over 12 million subscribers) - the channel has openly become very pro-AMD recently to the point where they criticize or/and underplay anything released by NVIDIA/Intel: "I'm DONE covering for NVIDIA - RTX 3070 Review".

And Linus conveniently forgot about the existence of 3600 which was introduced at $200 (and now can be have for less that $160). Yeah, "just a measly $50 price increase" when reality it's starts from 50%. Oh, and the Ryzen 5 3600X costs $250, not $300! AMD does love fat margins!

Relax man, the 3700X in that screen is a typo, they meant the 3800X. If you look further ahead in the video, that's the CPU they used for comparisons.

 

[birdie]

Relax man, the 3700X in that screen is a typo, they meant the 3800X. If you look further ahead in the video, that's the CPU they used for comparisons.

3600X for $300 is also a typo? Too many convenient typos in a single slide, man. Way too many. Also, did you listen to what Linus said? If I'm not mistaken he talked about the 3700X.

 

[Deleted member 24505]

The Linus Tech Tips channel which I used to watch occasionally, is now openly lying to cover Ryzen 5000 price increase:

View attachment 174538

Yeah, the Ryzen 7 3700X which was introduced at $330 and now costs on average around $310 (and occasionally is being sold for as low as $250), has suddenly become $90 more expensive according to Linus Tech Tips (with over 12 million subscribers) - the channel has openly become very pro-AMD recently to the point where they criticize or/and underplay anything released by NVIDIA/Intel: "I'm DONE covering for NVIDIA - RTX 3070 Review".

And Linus conveniently forgot about the existence of 3600 which was introduced at $200 (and now can be have for less that $160). Yeah, "just a measly $50 price increase" when reality it's starts from 50%. Oh, and the Ryzen 5 3600X costs $250, not $300! AMD does love fat margins!

Imo linus is a tit, can't watch him

 

[Rahnak]

3600X for $300 is also a typo? Too many convenient typos in a single slide, man. Way too many. Also, did you listen to what Linus said? If I'm not mistaken he talked about the 3700X.

Well, it is a wrong value, the 3600X did launch at $249, I don't really believe it would be intentional. Doesn't make sense to lie about launch prices, everyone can google that. I gave the video another quick look and didn't hear any mention of the 3700X by name. In the offending slide he even says "these all coincide with their Ryzen 3000 series namesakes in terms of core count and market position". 3700 and 3800 do not coincide in name. And all the slides after use the 3800XT.

I'll agree with the AMD bias though.

 

[mouacyk]

I have found a review comparing ddr 3200 vs ddr 4000 for the 5900x.

fclk:uclk:mlck should be 1:1:1 for this test.

timings 16-15-15-15-36 for both 3200 and 4000

Recensione Ryzen 9 5950X e Ryzen 9 5900X: con Zen 3 AMD vince anche nel gaming

Test dei processori AMD Ryzen 9 5950X e Ryzen 9 5900X basati su architettura Zen 3. Le nuove CPU Ryzen, grazie a un'architettura totalmente rinnovata, compiono un ulteriore balzo in avanti per quanto riguarda le prestazioni, in particolare quelle single-thread, come i videogiochi.

www.hwupgrade.it

The reviw is in italian (but you can translate with google or bing translator).

The found 1,3% performance difference (average on several use case)

If the reviewer had bothered to tune the 4000 overclock, the performance uplift should be consistent. The fluctuating results are indicative of a lazy primary-timing overclock, that is also potentially unstable and doing error correction on the CPU.

 

[MaurizioC]

If the reviewer had bothered to tune the 4000 overclock, the performance uplift should be consistent. The fluctuating results are indicative of a lazy primary-timing overclock, that is also potentially unstable and doing error correction on the CPU.

tune what?

they did not perform any cpu overclock just set the memory speed and IF speed keeping same timings to perform a clean comparative having the memory speed as the only difference.

The results to me are indicative that some workload are sensitive to memory bandwith while other are not

 

[Taraquin]

I've noticed a difference in ram-timings testing between Ryzen CPUs. On the Zen 2 tRC is 48, on Zen 3 it's 73. This makes me wonder: What are other seconderies etc? Is tWR, tRFC, tFAW the same on all setups? This can impact performance quite a bit in many games and certain apps.

 

[TheoneandonlyMrK]

Good going IMHO, last year the Intel fans were talking up 1080p performance, now it's 720p, or price.
Oh the Comedy.

And of course it's a hard ask to jump from 3Xxx to 5Xxx, in price, 3Xxx are still on the shelf , should they sell em for nowt, pass em out?!, Good business.

But choose with your wallet , I am, As Ever.

 

[MaurizioC]

I've noticed a difference in ram-timings testing between Ryzen CPUs. On the Zen 2 tRC is 48, on Zen 3 it's 73. This makes me wonder: What are other seconderies etc? Is tWR, tRFC, tFAW the same on all setups? This can impact performance quite a bit in many games and certain apps.

where?

 

[Space Lynx]

I've noticed a difference in ram-timings testing between Ryzen CPUs. On the Zen 2 tRC is 48, on Zen 3 it's 73. This makes me wonder: What are other seconderies etc? Is tWR, tRFC, tFAW the same on all setups? This can impact performance quite a bit in many games and certain apps.

yes please tell us where you got this info. the only person I know capable of these timings is @1usmus and his calculator is not updated yet. also buildzoid knows a lot about ram, but those are only two i know of.

 

[DemonicRyzen666]

AMD explained that 2000 IF is achievable by some good silicon CPUs, just like 1900 IF was possible on many Zen 2 CPUs, but not all of them

Can someone tell me if there is an Infinity link between the two CCD's or not ?
If not then there is still a bottleneck for the damn IO die causing slow downs still from spill over from CCD to CCD communication.

 

[MxPhenom 216]

Can someone tell me if there is an Infinity link between the two CCD's or not ?
If not then there is still a bottleneck for the damn IO die causing slow downs still from spill over from CCD to CCD communication.

But the CCDs are unified now though I thought?

 

[DemonicRyzen666]

But the CCDs are unified now though I thought?

that's CCX (core complexes) inside the chiplet
CCD is the chiplet it self
what I'm asking is if the two chiplets are linked
like how the CCD is linked to the I/O die, with an Infinity link.