Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[nicalandia]

He just had a modification of the Genoa list, claimed the clocks are conservative all-core turbo given by OEM materials, so....... it looks to be just speculation. And reputation of that guy seems doubtful......

The dude has full access to Sapphire Rapids, Milan and now has full access to Genoa. I trust that source.

 

[igor_kavinski]

makes it sound that more cache may indeed matter

AutoDesk 3ds Max 2017 CPU Performance

Processors often get a ton of reviews and benchmarks, but unfortunately for content creators (and other professionals), it is difficult to find benchmarks that focus on professional applications. In this article we will be looking at a number of Intel CPUs and how they perform in Autodesk 3ds Max.

www.pugetsystems.com

This seems to suggest that memory bandwidth plays a far greater role than number of cores. I suppose the cache would help with its higher bandwidth. Ideally the 5900X3D would be a better choice than the 5800X3D but you will have to wait and see or just go direct to Zen4X3D. Regarding PCIe 5.0, any new tech initially commands a premium.

 

[moinmoin]

Regarding storage - this would no doubt help, but IMO moreso with scene saving/loading times, rather than viewport performance during actual work, when the scene is entirely loaded inside RAM. At least i presume that.
This is actually one of the things that slow down my workflow the most, because i keep my work data on regular HDD (WD Caviar Black from 2011). I was thinking to keep the most recent stuff on the M2 disk, to improve on this, and only move it away to said HDD storage, once its done and not being worked on anymore, but there is issue with path to external files like textures, xrefs and whatnot - trying to load such scene from different disk is a pain, as its missing that stuff and it needs to be linked manually and whatnot...

I'd strongly suggest to replace your HDD with an NVMe of the same size right away. The jump from HDD to NVMe is night and day, even if it's "only" PCIe 3, nevermind 4. Waiting for PCIe 5 on top of that won't make such a huge difference anymore.

 

[Timmah!]

I'd strongly suggest to replace your HDD with an NVMe of the same size right away. The jump from HDD to NVMe is night and day, even if it's "only" PCIe 3, nevermind 4. Waiting for PCIe 5 on top of that won't make such a huge difference anymore.

I am aware of that, but i am worried about the safety of using SSD for storage. When HDD breaks, there is some possibility you can retrieve the data (i did so in the past myself) - i dont think thats possible when your NVMe goes down.
I realize i could do RAID 1 to alleviate this issue or set-up some automated back-up somewhere (another HDD, NAS, cloud) - but that would require additional patience to learn how to do it, and no doubt additional costs as well

This leads to a question:

say i buy zen4 cpu with 670(E) board - which has like 4 or 5 m2 slots.
I plug in 2x GPUs (which would be running in 8x mode obviously) and 2x PCI-E 4.0/5.0 m2 drives (depending on price/availability) - one for windows and apps, the other for games.
I assume at that point i am done with m2 disks, right? Cant plug any more of those, like 2 more for the aforementioned RAID solution?...since i already used up all 24 PCI-E lanes.

 

[igor_kavinski]

I assume at that point i am done with m2 disks, right?

OWC Express 4M2

Four M.2 SSD bays with pure adrenaline inducing Thunderbolt 3 performance.

www.owcdigital.com

I suppose you could use that for RAID 1+0 with four cheap NVMe SSDs, provided your chosen mobo has thunderbolt.

Regarding SSD reliability, if you choose a good brand like WD or Crucial and keep an eye on the remainng writes left (usually easy to do with manufacturer dashboard or Crystaldiskinfo), you will know beforehand when your SSD's life is going to end so you can take the appropriate action. Not sure if the life remaining status can be accessed behind that thunderbolt enclosure.

 

[maddie]

OWC Express 4M2

Four M.2 SSD bays with pure adrenaline inducing Thunderbolt 3 performance.

www.owcdigital.com

I suppose you could use that for RAID 1+0 with four cheap NVMe SSDs, provided your chosen mobo has thunderbolt.

Regarding SSD reliability, if you choose a good brand like WD or Crucial and keep an eye on the remainng writes left (usually easy to do with manufacturer dashboard or Crystaldiskinfo), you will know beforehand when your SSD's life is going to end so you can take the appropriate action. Not sure if the life remaining status can be accessed behind that thunderbolt enclosure.

I don't think he means that, but a sudden failure. The data is mostly still accessible on the magnetic disks, though you have to do some work to recover it.

When a SSD fails, can you still recover the data from the flash chips?

 

[igor_kavinski]

When a SSD fails, can you still recover the data from the flash chips?

Then he has no choice but to backup regularly to an external HDD for peace of mind. Or just keep suffering the slowness of an hdd.

 

[MadRat]

Or alternate work on a pair of SSD on their own buses if your board supports it. Work alternating on each SSD, and script data results to archive over to the mechanical drive. The mechanical drive running on the same bus as one of the SSDs will be relatively a mild hit in performance. Lastly, while the working SSD is busy you preload the next work on the idle SSD. Fill up your idle drive with any possible pre-staging to minimize transitions when its time to launch the next processing workload. Any time you can saturate your system with steady work you know you are getting your money's worth. If your system cannot keep up with the pace of workloads then its time to start looking at the next upgrade.

 

[Doug S]

When a SSD fails, can you still recover the data from the flash chips?

It depends on the type of failure. SSD failure scenarios such as exceeding write life / running out of spare blocks merely stop you from writing new data. They do not impact your ability to read data, so it is trivial to copy to a replacement device.

Controller failure is a different matter, but HDDs suffered from that as well. The difference was with a HDD you could swap out the controller from the same model drive and recover the data - I did this once back in the mid 2000s, whether this is still possible on modern shingled drives I have no idea.

To my knowledge controller swaps are impossible with an SSD - and even if it could be done would require pretty nifty microsoldering skills versus simply dealing with a couple small ribbon connectors on the HDD.

 

[eek2121]

I am aware of that, but i am worried about the safety of using SSD for storage. When HDD breaks, there is some possibility you can retrieve the data (i did so in the past myself) - i dont think thats possible when your NVMe goes down.
I realize i could do RAID 1 to alleviate this issue or set-up some automated back-up somewhere (another HDD, NAS, cloud) - but that would require additional patience to learn how to do it, and no doubt additional costs as well

This leads to a question:

say i buy zen4 cpu with 670(E) board - which has like 4 or 5 m2 slots.
I plug in 2x GPUs (which would be running in 8x mode obviously) and 2x PCI-E 4.0/5.0 m2 drives (depending on price/availability) - one for windows and apps, the other for games.
I assume at that point i am done with m2 disks, right? Cant plug any more of those, like 2 more for the aforementioned RAID solution?...since i already used up all 24 PCI-E lanes.

For what it’s worth, I have been using SSDs since the early 32/64gb SATA drives dropped, and I have never had one fail. During that same time period I have had multiple drives fail.

Regardless of your storage, you should be doing regular backups to multiple locations. My “Documents” folder is encrypted and synced to 2 cloud providers and also an external NAS in near real time.

 

[Zepp]

am aware of that, but i am worried about the safety of using SSD for storage. When HDD breaks, there is some possibility you can retrieve the data (i did so in the past myself) - i dont think thats possible when your NVMe goes down.

I feel the same. I've had 2 SSD's fail, admittedly they were cheap but not noname brands. There was zero warning signs and I couldn;t retrieve data on either. I've never had a HDD fail in 20 years. They end up being retired for larger ones. usually last for 5-7 years.

The only way I could justify using SSD's for big storage is spending double and setting up a RAID 1. I currently have 14TB of storage so what would 28TB of NVME cost?

 

[igor_kavinski]

I currently have 14TB of storage so what would 28TB of NVME cost?

That 14TB can't all be hot data that you always need to be working on. NVMe is for things you absolutely can't live without and must get access to as quickly as possible. Rest can be stored on spinners.

 

[LightningZ71]

Thanks for the links.

Checked the Techgage article, and despite the Pro level card being recommended, if you look here:

https://techgage.com/wp-content/uploads/2018/07/SPECviewperf-13-AMD-vs-NVIDIA-3ds-Max-Performance.png

Geforce has more or less the same performance as Quadro, at least in 3Ds max. Anyway, Quadro is outside of my budget and i need GPUs primarily for the rendering, so i am looking for the highest-end geforces, which provide the same performance, but for significantly less money.

On topic of more CPU cores being helpful, this i did not know, it surprises me. Though i doubt this would scale over 16 cores, which would be my baseline.

Regarding storage - this would no doubt help, but IMO moreso with scene saving/loading times, rather than viewport performance during actual work, when the scene is entirely loaded inside RAM. At least i presume that.
This is actually one of the things that slow down my workflow the most, because i keep my work data on regular HDD (WD Caviar Black from 2011). I was thinking to keep the most recent stuff on the M2 disk, to improve on this, and only move it away to said HDD storage, once its done and not being worked on anymore, but there is issue with path to external files like textures, xrefs and whatnot - trying to load such scene from different disk is a pain, as its missing that stuff and it needs to be linked manually and whatnot...

BTW do we know when the PCI-E 5.0 m2 drives are going to be up for purchase? And what pricing to expect? I looked at current offering and Samsing 980 Pro 1TB is 159 EUROs with VAT around here. Which is acceptable, i guess. But if the PCI-E 5.0 replacement of this wont be same price, but significantly more, that will be indeed very disappointing.

Last thing, i looked more into this CPU affecting viewport performance myself and found this article:

UI and Viewport performance lags or seems slow on fast multicore CPUs in 3ds Max | 3ds Max | Autodesk Knowledge Network

3ds Max lags or shows a delay in performance, even on fast CPUs. This includes, but is not limited to Intel i9 and AMD 'Ryzen Threadrippers'. This includes clicking any area of the user interface or moving objects, animation playback (FPS, or frames per second) in the viewport. When trying to...

knowledge.autodesk.com

This part:

• Under certain PC hardware configurations, the multicore CPU's cache memory access may become bottle-necked when performing certain calculations. The CPUs may run fast as long as they can hit the data they need directly (in the cache), but can become stalled when hitting a "cache miss.".

makes it sound that more cache may indeed matter, so perhaps i should be looking at the v-cache version, even if it has slower clocks.

timmah!, what's your current setup with respect to ram and hard drives and how the hard drives connect to your system? How large are the files that you are working with? How full are your drives? Are your tasks more read-heavy or write-heavy? What OS are you using? What processor and chipset do you have?

There are a lot of ways to arrange your storage and a surprising number of things can heavily influence your performance.

 

[Doug S]

The only way I could justify using SSD's for big storage is spending double and setting up a RAID 1. I currently have 14TB of storage so what would 28TB of NVME cost?

The system I'm running I have two smallish SATA SSDs for active data and two hard drives that may go days between times I access data on them, each mirrored.

In my next PC I build I'm going to use two m.2 SSDs and do away with hard drives, but if I had 14 TB to store I'd be sticking with the mirrored HDDs for stuff I rarely use like I'm doing now. If you're actively using that entire 14 TB then you'd have to ask yourself whether the price of that much mirrored SSD storage is worth it for the performance you gain.

There are alternate scenarios to protect that 14 TB of SSD storage with HDD storage. Let's say you have 1 TB of "very active" data, i.e. the OS, desktop/home directory, etc. and 13 TB of active but not quite so active stuff.

You could use a 1 TB SSD to mirror a 1 TB partition and 13-14 TB hard drive to backup the rest. Would a daily copy via script of 13 TB of your SSD to a HDD work for you? If not, another possibility which may exist (I'm not sure, and it would depend on you OS) is an asychronous mirroring solution.

An asynchronous mirror isn't true mirroring, basically it is a mirror where the SSD would be allowed to run at full speed and the HDD is behind while it is unable to keep up (i.e. it'll be mirrored when inactive, almost perfectly mirrored when active but not so active the HDD can't keep up, and behind when you write faster than the HDD can write)

This is something I have experience with in enterprise WAN mirrors (SRDF in asynchronous mode, for those with similar experience) I don't recall ever seeing an asynchronous mirroring option within a single system (though I haven't ever had reason to want one, so I haven't looked) but with the performance and cost difference of SSD vs HDD it would make a lot of sense to be supported. Even when HDDs are gone someday it would still be useful, as it would be cheaper to mirror a very expensive SSD designed for max performance with a lower performance one designed for lower cost.

 

[Timmah!]

timmah!, what's your current setup with respect to ram and hard drives and how the hard drives connect to your system? How large are the files that you are working with? How full are your drives? Are your tasks more read-heavy or write-heavy? What OS are you using? What processor and chipset do you have?

There are a lot of ways to arrange your storage and a surprising number of things can heavily influence your performance.

Currently i have X299/7940x/32GB of RAM - 3000MHz
I intend to replace that, perhaps this year already with Zen4 16core or alternatively, if Intel does release it, Fishhawk Falls "HEDT" platform.
Right now i have Win10, but will possibly jump to Win11 with new rig.

I have quite a lot of drives:

- M2 Samsung 950 Pro 256GB for Win/apps
- M2 Samsung 960 Evo 512 GB for few more work related apps, autoback folder for 3Dsmax
- Samsung 850 512 GB for games
- some kind of Intel 240GB SSD, where i keep rendering assets - models/textures/materials
- WD Caviar Black 2TB - here i keep my documents, work stuff (autocad/3dsmax/photoshop files) and some media
- WD Gold 6TB - here i keep more media, for example TVshows like all the Star Trek episodes, and recently copied here back-up of those work files - it was 890GB :-O
- WD Green 2TB - random stuff, installation folders for apps and old games and whatnot

Aside of the system and games drive, all the disks have reasonable amount of space left (at least 1/3, but generally more, up to 80 percent in case of that 960 Evo)

I intend to get 2x new 1TB M2s as part of the new build. Then obviously keep some of these. If you have idea, how to reorganize it for improved speed, i am curious.

EDIT: Thanks to all of you for your advices and fine discussion.

 

[Mopetar]

Regarding the v-cache chips, do you:

- think they solved the clocks issue with it and there wont be such a big difference in clocks between the regular and 3d version, as is the case of 5800x/58003d? Since the high clocks and resulting performance uplift seem to be the main point of improvement over Zen3, it would be rather disappointing to get 3d version and end up with zen3 clocks…

It wasn't really a clock issue so much as a voltage issue for the v-cache. That meant the CPU couldn't get as much either and the clocks weren't as high as a regular 5800X.

I don't know if the same issue will persist with the 5nm process being used for the cache chiplets with Zen 4, but for gaming we already saw that in a majority of cases the extra cache was more than worth the loss of clocks.

Zen 4 also has another upside in that there's more TDP room on AM5. Maybe AMD already intends to use all of that (in which case any v-cache CPU will need to trade some clock speed for watts necessary to drive the extra cache) but if not it does theoretically mean that the 3D CPUs will have the extra headroom to utilize.

- if that happens though, does anyone here now, between 5800x/58003d, which one provides better viewport performance in CAD apps (specifically autocad, 3dsmax)? I know v-cache suits better to games, in general, but my interest and the reason to upgrade is the performance in these apps. Especially 3dsmax can slow down significantly, when dealing with bigger models…

I haven't seen too many benchmarks that include AutoCAD so I have no idea. For something like 3DS Max or other rendering software you're going to be better off getting more cores. From other results the 5800X with the extra clock speed performs better. Unless the application needs to frequently access some recently used data that won't fit in the L2 cache, it doesn't do much. I don't think that rendering really fits that use case. It might make AutoCAD a bit snappier in some cases, but that can be hard to measure.

 

[Timmah!]

It wasn't really a clock issue so much as a voltage issue for the v-cache. That meant the CPU couldn't get as much either and the clocks weren't as high as a regular 5800X.

I don't know if the same issue will persist with the 5nm process being used for the cache chiplets with Zen 4, but for gaming we already saw that in a majority of cases the extra cache was more than worth the loss of clocks.

Zen 4 also has another upside in that there's more TDP room on AM5. Maybe AMD already intends to use all of that (in which case any v-cache CPU will need to trade some clock speed for watts necessary to drive the extra cache) but if not it does theoretically mean that the 3D CPUs will have the extra headroom to utilize.



I haven't seen too many benchmarks that include AutoCAD so I have no idea. For something like 3DS Max or other rendering software you're going to be better off getting more cores. From other results the 5800X with the extra clock speed performs better. Unless the application needs to frequently access some recently used data that won't fit in the L2 cache, it doesn't do much. I don't think that rendering really fits that use case. It might make AutoCAD a bit snappier in some cases, but that can be hard to measure.

Interesting thing about those voltages. Did not Intel chips have separate voltage for "core" and "uncore" parts of their chips? AMD does it differently?
Anyway you reckon 5nm process may perhaps fix that particular issue and allow v-cache models to be on par with vanilla ones frequency ones?

 

[moinmoin]

It wasn't really a clock issue so much as a voltage issue for the v-cache. That meant the CPU couldn't get as much either and the clocks weren't as high as a regular 5800X.

I don't know if the same issue will persist with the 5nm process being used for the cache chiplets with Zen 4, but for gaming we already saw that in a majority of cases the extra cache was more than worth the loss of clocks.

I'd hope AMD prepared AM5 with a dedicated voltage lane for v-cache.

 

[Timmah!]

I'd hope AMD prepared AM5 with a dedicated voltage lane for v-cache.

So its a socket thing? They could not do it for 5800X3D, cause there werent pins on the socket dedicated for that purpose?

 

[LightningZ71]

Currently i have X299/7940x/32GB of RAM - 3000MHz
I intend to replace that, perhaps this year already with Zen4 16core or alternatively, if Intel does release it, Fishhawk Falls "HEDT" platform.
Right now i have Win10, but will possibly jump to Win11 with new rig.

I have quite a lot of drives:

- M2 Samsung 950 Pro 256GB for Win/apps
- M2 Samsung 960 Evo 512 GB for few more work related apps, autoback folder for 3Dsmax
- Samsung 850 512 GB for games
- some kind of Intel 240GB SSD, where i keep rendering assets - models/textures/materials
- WD Caviar Black 2TB - here i keep my documents, work stuff (autocad/3dsmax/photoshop files) and some media
- WD Gold 6TB - here i keep more media, for example TVshows like all the Star Trek episodes, and recently copied here back-up of those work files - it was 890GB :-O
- WD Green 2TB - random stuff, installation folders for apps and old games and whatnot

Aside of the system and games drive, all the disks have reasonable amount of space left (at least 1/3, but generally more, up to 80 percent in case of that 960 Evo)

I intend to get 2x new 1TB M2s as part of the new build. Then obviously keep some of these. If you have idea, how to reorganize it for improved speed, i am curious.

EDIT: Thanks to all of you for your advices and fine discussion.

That's a mess of different drives. It looks like you are currently using 2 X M.2 slots and 5 SATA ports. Not knowing what your budget is, for my personal usage, I would probably save my pennies until I could afford the following:

Buy 4 of a decently rated 6TB Hard Drive. In the end, reliability over performance, avoid SMR.
Buy a single 12TB+ Hard Drive.

Keep the 950 Pro for boot.
Use Windows Storage Spaces (via powershell) to create a tiered storage array with the 4 X 6TB drives in a RAID 0 (no redundancy, stripped) as a HDD tier and the 960 EVO as the SSD Tier.
Configure the 12TB drive to automatically backup the storage array nightly.

That covers 2 M.2 NVME ports and 5 SATA ports. If you have a 6th SATA port, you can retain the 850 512GB for games and non-essentials.

The above configuration will give you effectively NVME speeds for any of your active tasks while keeping 24GB of storage available to you. You keep all of your programs and data on it, save your games if you have the SATA port for it. You have near-line disaster recovery with the backup drive. You should already have an offline backup, so I did not consider that.

I know the above works because I have almost that exact configuration in my son's computer and have set similar configurations for multiple others.

The rules, make sure that the SSD tier NVME drive is healthy. Always, ALWAYS, check that your backup drive is making good backups. Not having that drive is not an option for a work computer. Once you build the configuration, it will work almost seamlessly.

There are a lot of people that are not fans of MS Storage Spaces. You need to do your research on how to make the configuration as you shouldn't use that if you don't know how to build it. I have had excellent success with it and haven't yet had an array fail for anything other than a mechanical drive failure. That's what the backup is for.

 

[igor_kavinski]

Buy 4 of a decently rated 6TB Hard Drive. In the end, reliability over performance, avoid SMR.
Buy a single 12TB+ Hard Drive.

Just confused here. How is a 12TB drive supposed to backup a 24TB array? Is that only temporary until the array fills past 12TB and then Timmah has to get a bigger backup drive?

 

[LightningZ71]

Just confused here. How is a 12TB drive supposed to backup a 24TB array? Is that only temporary until the array fills past 12TB and then Timmah has to get a bigger backup drive?

Yes. I have found that most use cases can easily work with a nearline backup that is 50% of the volume size. Backups typically include some compression as well. But, drives larger than 12 TB start getting quite expensive quickly. 12 is also much larger than what he's currently fitting in.

 

[DrMrLordX]

So its a socket thing? They could not do it for 5800X3D, cause there werent pins on the socket dedicated for that purpose?

Um basically. If you want a component of a CPU to have an independent voltage plane, it needs . . . an independent voltage plane, either from a separate VRM phase on the motherboard or something like an IVR.

I hope I'm saying that right.

 

[eek2121]

The only way I could justify using SSD's for big storage is spending double and setting up a RAID 1. I currently have 14TB of storage so what would 28TB of NVME cost?

Listen to others obviously, but I had to comment. A quick glance at Amazon shows I can get 14tb of (SATA based) SSD storage for $1,160. 😉

 

[LightningZ71]

4 X 6TB HDD, SATA commodity price is ~$70 each. $280 for 24TB of space. He can use the existing SSD to cache that array with MS SS. 12TB drives start around $150. So, for $430, he can have 24TB of storage space, 50% covered with backup, with NVME speeds for most of his active files.