The No.1 Website for Pro Audio
The MONUMENTAL shift in CPU power
Old 23rd September 2019
  #1
Gear Addict
 
Hammer v2's Avatar
The MONUMENTAL shift in CPU power

Only a couple of years ago the best you could hope for was a highly strung 4 core chip. Which was fine to a point but still required some CPU watching in order to get the job done.

Then Ryzen happened.

After copping a bollocking, intel relented and started releasing 6 core, then 8 core mainstream CPUs. This is GREAT for us as even the puny i8400 was a cracker getting very close to the point where the CPU becomes irrelevant.

Then Ryzen 3 (Zen2) happened.

12 cores now, 16 cores in November all attainably priced using standard consumer parts. Intel hasn't countered yet, but they will and I reckon that once that happens the CPU will no longer be a limitation in audio. Period.

In three years the game has changed for DAWs in a very, very good way.


What do you reckon the outcomes of this monumental shift will be (other than really cheap hdx/uad cards!)?
Old 23rd September 2019
  #2
Gear Maniac
 

IIRC, I bought my 6 core Intel in 2012.

Many of my audio programs still do not take advantage of all that multicore action.

Until software is coded to properly leverage multicore CPUs, then to me, it's not terribly exciting to have 12 cores vs 6 cores.

I am more excited about M.2 drives and any single-threaded performance gains.
Old 24th September 2019
  #3
RiF
Lives for gear
 
RiF's Avatar
My 4-core/8-thread i7 in my Macbook maxes out a single core with ease (Ozone, Diva, Amplitube).
16+ cores do not solve this issue.

That said, many cores are nice but a decent single-core performance is still needed for audio in many cases.

PS: I have to admit that I could not bring the i9-9900K in my desktop PC to it's knees. Neither a single core and certainly not all of them.
Old 24th September 2019
  #4
I personally wonder if it would be worthwhile to upgrade my sandybridge (i2600k) , yes the pass mark tells me its 8000 vs 20000 score, but will omnisphere realize those score in any way? I am not so sure about that.

I wish that there would be a passmark database, which tells me - upgrade and then get 16 /32 tracks, with processing , running at that buffer size.
Old 24th September 2019
  #5
Gear Maniac
 

Quote:
Originally Posted by RiF View Post
16+ cores do not solve this issue.

That said, many cores are nice but a decent single-core performance is still needed for audio in many cases.
Agree. Though, I think Diva and Repro utilize multiple cores very well. IIRC, Serum can only use one core.

Quote:
Originally Posted by RiF View Post
PS: I have to admit that I could not bring the i9-9900K in my desktop PC to it's knees. Neither a single core and certainly not all of them.
Really?! Even Serum with max unison and morphing? Edit: Sorry, I thought you listed Serum. In any event, I'd be really curious to know how the 9900k handles Serum.
Old 25th September 2019
  #6
Lives for gear
 
ponzi's Avatar
Its true that the number of cores available has gone up quite a bit. I don't think I needed 8 cores, but it was only something like $100 more than 4-6 core models, so I just went with it.

Single core performance is not increasing very much in the last few years, and that is a significant aspect of daw performance. (Many others have expressed their strong disagreement with my view on this). I maintain there is a diminishing return aspect to increasing number of cores for a personal computer.

There is a narrative that amd is running circles around intel. From what I have seen so far, the intel products are price competitive, and some of those using the new amd products are running into cooling challenges. Anecdotal, but not seeing a tsunami of amd taking over daw market. I am waiting to get more information on this as more people try out amd chips.
Old 25th September 2019
  #7
Lives for gear
 

Maybe plugin and daw creators should finally start coding on that way so their software utilizes all cores and not a single core, so, we wouldnt have problem getting at 100% cpu so easily while mixing and producing...
Old 25th September 2019
  #8
RiF
Lives for gear
 
RiF's Avatar
Quote:
Originally Posted by FoxMulderFBI View Post
Maybe plugin and daw creators should finally start coding on that way so their software utilizes all cores and not a single core, so, we wouldnt have problem getting at 100% cpu so easily while mixing and producing...
It's not that easy.
Multiple cores/threads only help if the processing can be split up into parallel paths which then can be assigned to the different cores/threads.

The processing of audio on a single track is inherently sequential. Same with most if not all processing within a plugin.
An exception can be polyphonic/multitimbral synths where each voice can be processed individually.

In short, the amount of processing that can be done on a single track or in a single plugin is limited by the single core speed of the CPU. And as ponzi said, single core performance did not increase that much recently.
And I am guessing that we pretty much have reached kind of the vertical limit due toclock speed and stuff here, hence the horizontal "escape" of the CPU manufacturers into multi-core architectures.
Old 26th September 2019
  #9
Quote:
Originally Posted by RiF View Post
It's not that easy.
Multiple cores/threads only help if the processing can be split up into parallel paths which then can be assigned to the different cores/threads.

The processing of audio on a single track is inherently sequential. Same with most if not all processing within a plugin.
An exception can be polyphonic/multitimbral synths where each voice can be processed individually.

In short, the amount of processing that can be done on a single track or in a single plugin is limited by the single core speed of the CPU. And as ponzi said, single core performance did not increase that much recently.
And I am guessing that we pretty much have reached kind of the vertical limit due toclock speed and stuff here, hence the horizontal "escape" of the CPU manufacturers into multi-core architectures.
This is true, it's not so much about developers being "lazy".
if you do a lot of audio processing in a serial configuration (e.g. 10 tracks all going through a single bus with Effect A -> Effect B -> Effect C -> Effect D etc.) it's very hard or impossible to do those calculations in parallel. You simply don't know what to feed into Effect D until the 10 source tracks (which can be done in parallel) and the previous effects in the chain (which has to be done in series) have been calculated.

Of course there's still a *lot* of parallel processing in music production. Often you have dozens of buses/tracks that are completely independent until they go to the master bus. The DAW can then process those in parallel and make sure they're all synced up when they reach the mix bus/master output. Also, as you mentioned, a polyphonic synthesizer is essentially a bunch of independent monophonic synthesizers (apart from any master effects), which is why plugins like Diva work well with Multi-Core processing.

For me, it's *extremely* rare to run out of processing power with my 8-core Ryzen 1800X on Studio One 4.5. I almost have to do it deliberately, e.g. by inserting ridiculous number of plugins on buses, using 16X oversampling etc.
Old 26th September 2019
  #10
Gear Head
 

Quote:
Originally Posted by RiF View Post
It's not that easy.
Multiple cores/threads only help if the processing can be split up into parallel paths which then can be assigned to the different cores/threads.

The processing of audio on a single track is inherently sequential. Same with most if not all processing within a plugin.
An exception can be polyphonic/multitimbral synths where each voice can be processed individually.

In short, the amount of processing that can be done on a single track or in a single plugin is limited by the single core speed of the CPU. And as ponzi said, single core performance did not increase that much recently.
And I am guessing that we pretty much have reached kind of the vertical limit due toclock speed and stuff here, hence the horizontal "escape" of the CPU manufacturers into multi-core architectures.
Thankyou for this clear explanation.

It explains why after upgrading to a new PC (same speed, more cores) the performance is identical as before.

All my mastering sessions are using a single track, and Sequoia won't share this across cores.
Rendering is still a slow process - damn.
Old 26th September 2019
  #11
Multithreading itself is not the issue and is already done properly by most applications. The issue indeed is that many plugins have heavy algorithms which cannot be spread over cores.
Add the fact that send fx, groups/buses and master inserts all stack up the time domain for a single buffer to be processed, even if other cores are used for this, and you have a limitation of core speed.

So the only way this gets improved is:
-better software algorithms, performance wise (SSL Duende is exemplary although the new reverb algo is heavy)
-faster cores (this is going to be an issue as both Intel and AMD are hitting their limits. Intel chips already are cooking when reaching their TDP limit)
-better workflows (this is up to you: reduce master inserts, use lower algo settings for realtime playback, etc)
Old 26th September 2019
  #12
Motown legend
 
Bob Olhsson's Avatar
 

A developer told me the point of diminishing returns in real-time performance between core speed and thread management overhead is just six cores. He also said the most important spec is the number of floating-point calculations/sec and not the core speed.
Old 26th September 2019
  #13
Lives for gear
Maybe the shift seem to be massive but it is still nowhere near what could enable comfortable work on projects. My i9 9900k 5.1GHz barely copes with relatively simple sessions (with high quality plugins, e.g. from Acustica) and I doubt new Ryzen will make a huge difference. There is still room for improvements when it comes to single core processing (that more or less determines how many plugins you can run in a single channel without breaking up).
Old 26th September 2019
  #14
Here for the gear
 

Is it possible to use 2 computers for mixing

Hey Guys i have a mac Pro 2.1, 6 cores Xeon at 3ghz pretty old but still running and reliable, except that i bought Waves Scheps Omni Channel strip and waves Abbey Road plate reverb and i'd like to be able to use them extensively while mixing. I also have an 8 cores i7 3.2ghz windows 10 pc. Is there a way that i could have one running ableton live 10 and the other one computing the plugins??? I have two soundcards Liquid Saffire Pro for the MAC and a RME babyface for the PC....thank you in advance
Old 26th September 2019
  #15
Lives for gear
 
ponzi's Avatar
What I said on the subject is that in a daw running, there will be one thread that takes the longest to complete (before 'need more data' interrupts from the dac). Which thread it is, which cpu its on is not important. So, at some point when that thread cannot make the deadline, the daw is no longer functional. Adding more cores of the same speed will not help that situation.

We know that to some, regrettably unknown, degree, adding cores spreads the work around, but of course, its axiomatic that there will be one thread that takes the longest to get the job done even when all possible multi-threading is being done--and only speeding up the cpu will address that. (Or, of course changing the project to reduce its work, or lengthening the time between interrupts by increasing buffer size.)

Of course, the question is what and when, and there are so many unknown variables that only using the daw will determine its limits. For me, arguing about whether additional cores are worth it is kind of a moot point because my 9700k has 8 cores and it wasn't much more expensive than the 4-6 core cpus. So I have what I believe is more than I need, but I guess its there if I do need it.

I just looked it up and I can get a 9700k today for $300 and 9900k for $450, so $150 extra. If someone thinks their projects need lots of cpu, seems to me the 9900k is a no--brainer. Just get one of those monster sized cpu fans like the noctuia d15s, and liquid cooling if you plan on overclocking it.

PS. I know a guy who does video sfx for a living and he just got a home computer with an i8700--so that chip still commands plenty of respect. I believe the 9700 is slightly faster, or the same, but the cost was close to the same, and I think the 9700 supported some newer mobo features like 64 gig ram, so I went with it. Its a tough call on the 8700--it has 6 physical cores , but 12 'cores' with hyperthreading. The 9700k is 8 cores with no hyperthreading.
Old 26th September 2019
  #16
Gear Maniac
 

Quote:
Originally Posted by jbone1313 View Post
IIRC, I bought my 6 core Intel in 2012.

Many of my audio programs still do not take advantage of all that multicore action.

Until software is coded to properly leverage multicore CPUs, then to me, it's not terribly exciting to have 12 cores vs 6 cores.

I am more excited about M.2 drives and any single-threaded performance gains.
What is the advantage of M2 drives in your case ? For what kind of applications do you consider them especially useful ?
Old 26th September 2019
  #17
Lives for gear
Quote:
Originally Posted by ponzi View Post
What I said on the subject is that in a daw running, there will be one thread that takes the longest to complete (before 'need more data' interrupts from the dac). Which thread it is, which cpu its on is not important. So, at some point when that thread cannot make the deadline, the daw is no longer functional. Adding more cores of the same speed will not help that situation.

We know that to some, regrettably unknown, degree, adding cores spreads the work around, but of course, its axiomatic that there will be one thread that takes the longest to get the job done even when all possible multi-threading is being done--and only speeding up the cpu will address that. (Or, of course changing the project to reduce its work, or lengthening the time between interrupts by increasing buffer size.)
I think this is all true. However, you seem to disregard the fact that this "one thread that takes the longest to complete", who we all likely have met in our sessions , is indeed already competing with other processes, possibly a few that aren't related to the DAW.

Other processes, is something one needs to control to be able to make the best use of the hardware. Yet, it seems to me more people, than I would have thought, don't analyze these other processes and control them when needed.

In addition, as more cores could mean other threads is more likely to run on another core, chances would increase that core could complete it's work "on time". Provided of course that multi-cores can retain single core throughput. I think that is what looks better with these new gen CPUs, though I'd assume it varies.
Old 26th September 2019
  #18
Gear Maniac
 

Quote:
Originally Posted by Tom Barnaby View Post
What is the advantage of M2 drives in your case ? For what kind of applications do you consider them especially useful ?
In my case, I am hoping they speed up the loading of sample libraries a lot.

I use a lot of drum samples, especially Samples From Mars (which are fantastic).

I think a lot of the kits are 1-3GB. It takes 10-30 seconds to load them. If I can get that down to a few seconds, that would be really cool.

In addition to the time it takes to pull the samples off the disk (which the M.2 will hopefully reduce a by a lot), there is also the time it takes to load into memory and instantiate the Simpler instances in Ableton live (which takes CPU).

Until I actually have my new machine, I won't know how much it helps. But, as a kind of proof of concept, I am thinking about creating a RAMDisk on my current machine and loading from there.
Old 26th September 2019
  #19
Lives for gear
 
ponzi's Avatar
Not sure how you are connecting your m.2 drives, but they can mount on some mobos, and I use $20 adapters to put them in pci-e slots. Should be a lot faster loading samples. If you mount the m.2 in some kind of card that exposes a sata connector, the sata speed is way less than the pci-e direct.

IIRC, ramdisk is not persisted between boots. I would not spend time on it. If you have lots of ram, windows stores the stuff in ram anyway.
Old 26th September 2019
  #20
Gear Maniac
 

Quote:
Originally Posted by ponzi View Post
Not sure how you are connecting your m.2 drives, but they can mount on some mobos, and I use $20 adapters to put them in pci-e slots. Should be a lot faster loading samples. If you mount the m.2 in some kind of card that exposes a sata connector, the sata speed is way less than the pci-e direct.

IIRC, ramdisk is not persisted between boots. I would not spend time on it. If you have lots of ram, windows stores the stuff in ram anyway.
Would you say that 32 GB is more than enough when working with high quality samples libraries ?
Old 26th September 2019
  #21
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by Tom Barnaby View Post
Would you say that 32 GB is more than enough when working with high quality samples libraries ?
I have no idea --I am sure it depends on the libraries and number of tracks, but ramdisk just takes ram away from the other applications, so no free lunch on that.
Old 27th September 2019
  #22
Quote:
Originally Posted by Tom Barnaby View Post
Would you say that 32 GB is more than enough when working with high quality samples libraries ?
It goes fast.....
Old 27th September 2019
  #23
Lives for gear
Quote:
Originally Posted by ponzi View Post
...
IIRC, ramdisk is not persisted between boots. I would not spend time on it. If you have lots of ram, windows stores the stuff in ram anyway.
That's the reason I got interested in PrimoCache. It creates a RAM cache sized as you like, and configured to service the drive letters of your choice. That's the 'L1' cache. It also allows you to define an 'L2' cache that is an SSD whose job it is to maintain the current state of the L1 cache as it evaporates on a power down; restoring it when you reboot.

Cache can be configured as R/W, R-only, or W-only. It's pretty cool. Windows only. {...yes, of course I wrote a 3-part article series about it...why do you ask?}
Old 27th September 2019
  #24
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by MediaGary View Post
That's the reason I got interested in PrimoCache. It creates a RAM cache sized as you like, and configured to service the drive letters of your choice. That's the 'L1' cache. It also allows you to define an 'L2' cache that is an SSD whose job it is to maintain the current state of the L1 cache as it evaporates on a power down; restoring it when you reboot.

Cache can be configured as R/W, R-only, or W-only. It's pretty cool. Windows only. {...yes, of course I wrote a 3-part article series about it...why do you ask?}
Your call, like I said, that ram is taken away from your samplers and so on.
Old 27th September 2019
  #25
Quote:
Originally Posted by Mikael B View Post
I think this is all true. However, you seem to disregard the fact that this "one thread that takes the longest to complete", who we all likely have met in our sessions , is indeed already competing with other processes, possibly a few that aren't related to the DAW.

Other processes, is something one needs to control to be able to make the best use of the hardware. Yet, it seems to me more people, than I would have thought, don't analyze these other processes and control them when needed.
A good configuration does not have such issues. Some DAWs have/had issues with a strong load on the first core, no matter what the project looks like, but generally, at least on Windows, a good configuration causes excellent spreading.

Quote:
In addition, as more cores could mean other threads is more likely to run on another core, chances would increase that core could complete it's work "on time". Provided of course that multi-cores can retain single core throughput. I think that is what looks better with these new gen CPUs, though I'd assume it varies.
Correct. If you have many tracks, at some point that track with 8 inserts is going to share a core with one or more other tracks, which does not help. More cores certainly help in this regard. However, if that single track already kills the core without any other tracks present, consider freezing or getting a better source...


Quote:
Originally Posted by ponzi View Post
Not sure how you are connecting your m.2 drives, but they can mount on some mobos, and I use $20 adapters to put them in pci-e slots. Should be a lot faster loading samples.
If the M.2 and the M.2 slots are PCIe, it will not be faster by putting them in a PCIe slot. If the M.2 is SATA, they will not be faster either if you need to put them in a PCIe M.2 adapter. I guess you intend to say to choose PCIe M.2 SSDs insstead of SATA M.2 SSDs?
Old 27th September 2019
  #26
Quote:
Originally Posted by MediaGary View Post
That's the reason I got interested in PrimoCache. It creates a RAM cache sized as you like, and configured to service the drive letters of your choice. That's the 'L1' cache. It also allows you to define an 'L2' cache that is an SSD whose job it is to maintain the current state of the L1 cache as it evaporates on a power down; restoring it when you reboot.

Cache can be configured as R/W, R-only, or W-only. It's pretty cool. Windows only. {...yes, of course I wrote a 3-part article series about it...why do you ask?}
Not sure how much of a difference it makes in practice. Windows itself already maintains several layers of cache. You can see how much data Windows has cached from the Task Manager. On my work laptop with 16 GB of RAM, Windows has currently cached 8.4 GB (while 7.1 GB are being used by applications).
Old 27th September 2019
  #27
Lives for gear
Quote:
Originally Posted by AdvancedFollower View Post
Not sure how much of a difference it makes in practice. Windows itself already maintains several layers of cache. You can see how much data Windows has cached from the Task Manager. On my work laptop with 16 GB of RAM, Windows has currently cached 8.4 GB (while 7.1 GB are being used by applications).
My only point in mentioning PrimoCache is that RAM disks are clumsy to manage. They have a fixed size, and they're not persistent across reboots.

My reasons for fiddling with PrimoCache are entirely related to stretching the usefulness of my 5,1 Mac Pro, and dealing with video editing. Consider it a case of M.2 NVMe envy. (see what I did there?) Nothing in the audio that I do is a challenge to my machine.

However, in video the files are gigantic and zinging up and down the timeline of a 6-camera concert makes for lots of writing (color adjustments, effects) and repetitive reading. Having the disk I/O more frequently happen in RAM reduces the I/O latency and overall makes for a smoother editing experience. It cannot overcome bog slow HDD's, so I generally run at least a RAID-0 pair of primary media HDD drives to give me an underlying 440MBytes/sec read rate. The video caching drive is made of 3x 2TB SSD's and does about 1000MBytes/sec write and 1200MBytes/sec reading. It's so big because a 240GB concert in H.264 can immediately expand to be over 2TB in DNxHR.

I've defined a 20GB L1 RAM cache, but not to worry, the machine has 96GB of RAM. The RAM cache is now an agile sliding window along the most active parts of whatever I'm working on. The RAM cache benchmarks at about 11000MBytes/sec. In video, I'm always zooming in and out along various portions of the timeline, and the re-paint time is shorter and smoother with the PrimoCache engaged. Again, audio was not a consideration for doing any of this, and I never bothered to engage the L2 persistent cache option because things went well enough without it, and as always, disk ports are precious.

As a peek into the future, while I'm waiting for the AMD 3950X to land, I'm building a DIY Synology box to be connected via 10GbE links; possibly dual links. The idea is to have more elegant and stable housing for the media pool, a smoother backup process, and possibly even house the cache drives.

So, back to audio...
Old 27th September 2019
  #28
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by DAW PLUS View Post
...If the M.2 and the M.2 slots are PCIe, it will not be faster by putting them in a PCIe slot. If the M.2 is SATA, they will not be faster either if you need to put them in a PCIe M.2 adapter. I guess you intend to say to choose PCIe M.2 SSDs insstead of SATA M.2 SSDs?
Both the mobo slot and the pcie-slot adapter, if it does not expose a sata plug, are electrically the same--a pci-e connection.

PS. I have been reading the Windows Internals book for windows 10 of late. Just a reminder on threads:

1) Windows will launch multiple threads for a program (technically a process, as a program is static code that gets turned into a process when its executed) even if the developer takes no extra steps to do so. For instance, there is often a thread for gui tasks in addition to other things a process is doing.

2) At any given time, it is normal for there to be hundreds of threads existing on a windows 10 system. Most of these are in a waiting state and are not executing or ready to execute. Process explorer shows on my system right now over 100 processes and these have multiple threads. A quick check shows that cubase 10 process opened to a 4 track project has 94 threads.

3) The developer can add code to launch threads in addition to the threads created from the normal instantiation of the process. We hope that with newer plug ins, the developers have done this a lot to allow the daw process to take advantage of multiple cores.

4) Normally, neither the developer, nor the process itself, has control over which threads are executed on which core. This is done by the thread scheduler based on workload and other factors, and the process executing has no awareness of where its threads got scheduled.

5) It is possible for the developer to set thread affinity to a given processor, but there are few circumstances where constraining the thread scheduler is prudent, and looks like an approach that reduces multi-core load spreading. Thread affinity appears to seek re-use of cache.

6) Only the developers know whether each track has its own thread--same with plug-ins, and due to #4 , should have no control over which processors the threads are scheduled on.

My whole point here is that what cores tracks or plug-ins run on is opaque to the process in the normal course of things. While there may be exceptions to this, in general there is not a reliable way to look at a running process and identify which of its threads are running on which cpu.

For discussion of thread affinity: https://en.wikipedia.org/wiki/Processor_affinity

Last edited by ponzi; 27th September 2019 at 02:59 PM..
Old 27th September 2019
  #29
Quote:
Originally Posted by ponzi View Post
Both the mobo slot and the pcie-slot adapter, if it does not expose a sata plug, are electrically the same--a pci-e connection.
I still don't get your point regarding using M.2s on a PCIe adapter. That only makes sense if you have no M.2 slots onboard or if they are already ocupied.

Quote:
1) Windows will launch multiple threads for a program (technically a process, as a program is static code that gets turned into a process when its executed) even if the developer takes no extra steps to do so. For instance, there is often a thread for gui tasks in addition to other things a process is doing.

2) At any given time, it is normal for there to be hundreds of threads existing on a windows 10 system. Most of these are in a waiting state and are not executing or ready to execute. Process explorer shows on my system right now over 100 processes and these have multiple threads. A quick check shows that cubase 10 process opened to a 4 track project has 94 threads.
The problem with the term "threads" is that Microsoft uses it for processes which have split tasks (to say it in a simplistic way) while Intel uses it to describe logical cores, i.e. real cores and Hyperthreading cores. When people here discuss threads, they mean the Intel naming. Everything else is just confusing and irrelevant to most readers here, IMO.
Old 27th September 2019
  #30
Lives for gear
 
ponzi's Avatar
Quote:
Originally Posted by DAW PLUS View Post
...The problem with the term "threads" is that Microsoft uses it for processes which have split tasks (to say it in a simplistic way) while Intel uses it to describe logical cores, i.e. real cores and Hyperthreading cores. When people here discuss threads, they mean the Intel naming. Everything else is just confusing and irrelevant to most readers here, IMO.
Bottom line is that when folks talk about a track or plug in running on a core, that's not how it works. Since people hear all sorts of things on the internet, I lay out my reasoning. Those who try and follow it will learn something they did not know before. If they are not interested, they are free to not read it.

A core is not a thread. A thread runs on a core. I do not recall intel as calling a core a thread. Is my 9700k listed as a 8 thread cpu?

Hyperthreading is a marketing label for technology that allows two threads to run on a single core (its more complicated than that, but this is close enough). This simply means that a core can run two threads at the same time.

For a consumer or average daw user, this knowledge is not necessary. Folks can read reviews, look at benchmarks, or simply buy powerful computers. However, if folks are going to tell me that I am wrong when I say how things work, I think an accurate understanding of terminology and how things work is essential to convince me of their claims.

As to m.2, my point is that m.2 sockets or pcie m.2 adapters are equal in speed to each other and that m.2 adapters that expose a sata interface are slower as they are limited to sata speeds. I have seen this in some pci-e m.2 adapter cards that hold two m.2 drives where one of them is pci e-attached and the other goes to a sata socket on the card.

I agree that when one has a mobo m.2 socket, there is no reason to buy a pcie m.2 adapter. If one does not have a mobo with m.2 plug, the pci-e adapter is a great way to get m.2 speeds. I looked closely at my adapter and it appears that for the most part, the traces go straight from the m.2 socket to the 'pins' on the pci-2 plug, so it looks like the m.2 drive interface is essentially pci-2 protocol in a different pinout.

Last edited by ponzi; 27th September 2019 at 04:20 PM..
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump