Quantcast
The "today we build our studio pc" thread - Page 509 - Gearspace.com
The No.1 Website for Pro Audio
The "today we build our studio pc" thread
Old 3rd December 2021 | Show parent
  #15241
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
Quote:
Originally Posted by markusschloesser ➡️
The first link IMHO compares apples and oranges and is therefore not very good (some disk inbuilt the others thru usb, different OSes, paragon, ntfs on osx).
The second one with its analysis is fantastic! Thanks! ������
I skimmed through them - my comment was not for people who are using a template with just a measly 24GB of samples preloaded, (probably not even for one using "just" 16 instances of Kontakt), but more like 1000 VI-track templates with low RAM preloading settings. It's not everyone, but obviously people also start doing stuff like that if it become possible and if it helps them work faster.

Last edited by Skwaidu; 3rd December 2021 at 10:15 AM..
Old 3rd December 2021 | Show parent
  #15242
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Skwaidu ➡️
Like I mentioned before, apparently for *heavy* VI use there would be a definite advantage on the PCI 4.0 (980 Pro). For regular audio recording/editing/mixing even with lots of tracks, not really (even SATA SSD's do that pretty ok).
My experience and some tests are also inline with the test results Pictus posted here- I have different machines with a hodgepodge of SATA SSD all the way to WD Black g4.0 (500MB/s to 7500MB/s).

The difference across the board is pretty underwhelming. Session and sample lib load times are mostly driven by CPU speed- so is OS boot time.

For Kontakt voices, a SATA SSD is going to get to about 3000-4000 (as it shows in Kontakt where each mono ch counts as a voice). From what I've seen, 1st gen NVMe is fast enough for Kontakt that the bottleneck is the CPU- where modern CPUs are capping out around 6-8k voices depending on the lib.

Those may not be completely irrelevant voice numbers because on some of the modern libs one actual midi note is triggering 30+ voices- but I've never found a test scenario where there was a difference with any gen of NVMe with Kontakt. And the modern libs are using more CPU than the old ones.

Maybe your friend isn't using default kontakt settings? But, it also seems possible that he may not have approached this in an empirical way.
Old 3rd December 2021
  #15243
Gear Guru
 
🎧 10 years
New DAWbench data posted by Vin:

DAWBench DSP / VI Universal - Cross Platform DAW Benchmarks :
Old 3rd December 2021 | Show parent
  #15244
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
Quote:
Originally Posted by RyanC ➡️
My experience and some tests are also inline with the test results Pictus posted here- I have different machines with a hodgepodge of SATA SSD all the way to WD Black g4.0 (500MB/s to 7500MB/s).

The difference across the board is pretty underwhelming. Session and sample lib load times are mostly driven by CPU speed- so is OS boot time.

For Kontakt voices, a SATA SSD is going to get to about 3000-4000 (as it shows in Kontakt where each mono ch counts as a voice). From what I've seen, 1st gen NVMe is fast enough for Kontakt that the bottleneck is the CPU- where modern CPUs are capping out around 6-8k voices depending on the lib.

Those may not be completely irrelevant voice numbers because on some of the modern libs one actual midi note is triggering 30+ voices- but I've never found a test scenario where there was a difference with any gen of NVMe with Kontakt. And the modern libs are using more CPU than the old ones.

Maybe your friend isn't using default kontakt settings? But, it also seems possible that he may not have approached this in an empirical way.
Right, sounds reasonable. Need to revisit the chat at some point and ask for some references / explanations.
Old 3rd December 2021 | Show parent
  #15245
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
Quote:
Originally Posted by mattiasnyc ➡️
Thanks for the link. I'll crosspost my comment as it's relevant here as well:

I just wish there was a real-world mixing test included as well. "as many plugins as possible" isn't that (but creating long serial busses with CPU-intensive modern plugins would be).

(Less important here, but relevant to the overall comparison of the DAW world: I assume there is an explanation that I've missed why Apple M1's are not included?)
Old 3rd December 2021 | Show parent
  #15246
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Skwaidu ➡️
(Less important here, but relevant to the overall comparison of the DAW world: I assume there is an explanation that I've missed why Apple M1's are not included?)
Kontakt at least still doesn't have apple native. From my understanding, using it as is in rosetta 2 is causing obnoxiously long load times- especially for this amount of voices. From the M1 thread, it would probably take nearly an hour to load a template that used >5000 voices.

I don't know about the other plugins Vin uses- but I think they all need to be native for it to be relevent.


Also agreed- at this point, multithreaded performance is generally good enough for most users across the platforms. The two walls that are most relevent at this point (in my workflow at least) are low-buffer single-core performance for tracking/monitoring, and single-core power for loading the big plugins in series.

I also think these high of Kontakt voices is starting to get irrelevant. Newer Kontakt libs are more CPU intensive, I can't imagine many users needing 8000 voices of a 15 year old Kontakt library.

For big plugins in series- when extra latency isn't a concern (mixing), AudioGridder can be used to split inserts in series across cores- so at least there is an option there.
Old 3rd December 2021 | Show parent
  #15247
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
Quote:
Originally Posted by RyanC ➡️
Kontakt at least still doesn't have apple native. From my understanding, using it as is in rosetta 2 is causing obnoxiously long load times- especially for this amount of voices. From the M1 thread, it would probably take nearly an hour to load a template that used >5000 voices.

I don't know about the other plugins Vin uses- but I think they all need to be native for it to be relevent.


Also agreed- at this point, multithreaded performance is generally good enough for most users across the platforms. The two walls that are most relevent at this point (in my workflow at least) are low-buffer single-core performance for tracking/monitoring, and single-core power for loading the big plugins in series.

I also think these high of Kontakt voices is starting to get irrelevant. Newer Kontakt libs are more CPU intensive, I can't imagine many users needing 8000 voices of a 15 year old Kontakt library.

For big plugins in series- when extra latency isn't a concern (mixing), AudioGridder can be used to split inserts in series across cores- so at least there is an option there.
Agreed 100%, my 10 year old 12-core Mac Pro is powerful enough for multithreaded, and I still cannot mix on it any more.
Old 3rd December 2021 | Show parent
  #15248
Gear Guru
 
🎧 10 years
How long are your plugin chains?
Which plugins are in those chains?
Old 3rd December 2021 | Show parent
  #15249
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Skwaidu ➡️
Agreed 100%, my 10 year old 12-core Mac Pro is powerful enough for multithreaded, and I still cannot mix on it any more.
Did you check out AudioGridder?

It's a little clumsy, but you can split cores and add buffers per instance in series. So you can multi-thread inserts in series with it- at the cost of more latency (which is reported to the host so it gets compensated).

Reaper can multithread inserts in series, you only gain performance at higher buffers (and actually loose performance at lower buffers), but it IS possible for high-buffer mixing. It's the option "use multiprocessing for live inputs". Hopefully we start to see more DAWs implement this feature in some way at least for playback tracks.
Old 3rd December 2021 | Show parent
  #15250
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by mattiasnyc ➡️
How long are your plugin chains?
Which plugins are in those chains?
I rarely run into that problem in playback/mixing- at least not an issue for me on the 5950x. But it's also busses and nested busses etc. But also here a heavy VI that can't multithread plus some inserts at 96k can push one core over. Also in S1 the console/tape options in the mixer are one-core and I generally can't use them for that reason.

But, for sure, it's not too hard to hit a single core wall with AA/N4 plugins in series- I've only tinkered with them though a bit.
Old 4th December 2021 | Show parent
  #15251
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
Quote:
Originally Posted by mattiasnyc ➡️
How long are your plugin chains?
Which plugins are in those chains?
Like RyanC said, it's just all of the submixing and nested chains.

I'd have to look at tons of sessions and start calculating, but off the top of my head I guess

2-4 plugins on the master busses (Pro Tools), doing just room correction, delays, 5.1/7.1 home theatrical etc (on surround/ cinematic/Atmos stuff) - some of these will be migrated to the MTRX Studio I'm about to install

2-8 or so plugins on a master buss (stereo and surround work, usually more in stereo but I run most of my stereo stuff at 96k anyway)

After that it gets impossible to gauge. Some tracks have 0-2 plugins, some have 10 or more i.e. already bussed (rarely)

Submixes/ nested submixes, again super situational, but a film score thing might have for example a 7.1 routing folder (submix), that has another 7.1 routing folder inside of it, 1-4 plugins on both busses plus a plugin or a few on the tracks. Stereo mixes (or mixes where I don't need to deliver stems) have somewhat less bussing.

As far as plugins, just modern stuff, but some quite CPU intensive ones as well. in 5.1/7.1/7.1.2 it could be a 1-2 Fab EQ's, possibly a Fab multiband, maybe a multi-mono Soothe 2 and Fab's L 2 on all stem subgroups, for example. A lot of Nugen Halo 3D's, Penteos and Liquidsonics / Exponential multichannel / 3D verbs etc.

I guess I tend to start using the new tools when I have them, and more of them when I get power, until I need to back off, figure CPU efficient strategies and do commits if I run out of power. It's bound to happen with the i9 as well...

(I don't use the crazy "modeling" plugins much at all, but still things like Soothe2, Halo, Unveil, Gullfoss etc do use quite a bit more than your regular "DAW compressor" like what they are using in that test.

Last edited by Skwaidu; 4th December 2021 at 12:40 AM..
Old 4th December 2021
  #15252
Lives for gear
 
ponzi's Avatar
 
🎧 15 years
So, I have been on the sidelines of pc building for a while here. I just looked for ddr4 32 gig of ram and its now lower than $150. That seems quite a bit higher than 32 gig used to be, but maybe ddr3 is less expensive. So, I am going to be getting an amd setup wtih x570 aorus elite wifi. I am not sure I will be getting the same ram I have now when I upgrade (crucial), has anybody had any issues mixing ram brands? I do install it such that the two are working as a pair together.
Old 5th December 2021 | Show parent
  #15253
Gear Head
 
🎧 10 years
Quote:
Originally Posted by ponzi ➡️
So, I have been on the sidelines of pc building for a while here. I just looked for ddr4 32 gig of ram and its now lower than $150. That seems quite a bit higher than 32 gig used to be, but maybe ddr3 is less expensive. So, I am going to be getting an amd setup wtih x570 aorus elite wifi. I am not sure I will be getting the same ram I have now when I upgrade (crucial), has anybody had any issues mixing ram brands? I do install it such that the two are working as a pair together.
Do not mix RAM brands, they have different timings called CAS timings and will blue screen your system faster than you can say ""!
Old 5th December 2021 | Show parent
  #15254
Gear Head
 
🎧 10 years
Intel 12900k faster than 5950x

Uh oh, some bad news for all you Intel haters or AMD fan boys.

The 12900k beats your beloved AMD 5950x, and beats it with a 10nm wafer (where the 5950x is 7nm). (for anyone interested in what this means its the size of the lithographic wafer that cpus are engraved/etched into). This isnt even the flagship/performance Intel chipset that will be out next year (699x or 799x). It simply beats the AMD 5950x with with the Z690 chipset. Intel changed architecture on these chips, with insane L2 Cache and two types of modified cores, the biggest architectural shift in years (I want to say decades?)

I have a box of tissues for all of you.

Results are here
https://hothardware.com/reviews/inte...-review?page=1
Old 5th December 2021 | Show parent
  #15255
Gear Guru
 
🎧 10 years
Quote:
Originally Posted by psionic ➡️
Uh oh, some bad news for all you Intel haters or AMD fan boys.

The 12900k beats your beloved AMD 5950x, and beats it with a 10nm wafer (where the 5950x is 7nm). (for anyone interested in what this means its the size of the lithographic wafer that cpus are engraved/etched into).


The wafer is 10 NANO meters?.. "nm" refers to something else, and it's measured differently by each company. Jesus...

If you're going to try to look smarter than "fan boys" then what you teach should probably be at least correct. Otherwise you look like a tool.

Quote:
Originally Posted by psionic ➡️
This isnt even the flagship/performance Intel chipset that will be out next year (699x or 799x). It simply beats the AMD 5950x with with the Z690 chipset.
The cores aren't going to perform better because of the chipset. What the HEDT line will bring is more PCIe lanes, i/o, memory channels etc.

Secondly, the relevant tests that have been done that actually use a DAW can be found here, and they look as follows:









Without any other DAW-use-specific data it seems the reasonable conclusions are:

A: If you are going to run mainly DSP and not virtual instruments the best CPU is the 5950X.

B: If you are considering the 12900K with DDR4 and are using a lot of VIs at low buffer sizes the 5950X is still better, and so is the 5900X.

C: If you are using a lot of VIs at low buffer sizes the 12900K with DDR5 is better.

As for the future it's looking really great for everyone on 'both sides' of the fence. Because AMD will release their 3D cache on updated Zen 3 chips soon (announces in January), and later in the year they'll move to Zen 4 on a new platform with PCIe 5.0 and more importantly DDR5.

Now, if you look at the charts above you'll see the very large gains made by using DDR5 (or at least so it seems). If those gains carry over to Zen 4 then the pendulum will have swung back again. Because at that point you're likely looking at about 10-15% uplift from 3D cache, another 15% from the new process node (that's the "NM" again) and architecture, and then the DDR5.

Quote:
Originally Posted by psionic ➡️
Intel changed architecture on these chips, with insane L2 Cache and two types of modified cores, the biggest architectural shift in years (I want to say decades?)
That's as impressive as AMD's Zen architecture for exactly the same reason; just like AMD made processors that were behind the curve and didn't innovate Intel was left behind and didn't really innovate. Now this architecture looks impressive, but that's just because they were stagnant before due to a lack of serious competition.

Rather than speaking about this the way you do (see below) you should be ecstatic that AMD did so well and that Intel now is doing so well. We finally have competition that's driving prices down for great performance for everyone on either platform.

No need for the "fan boy" nonsense.

Quote:
Originally Posted by psionic ➡️
I have a box of tissues for all of you.
I'm not surprised you have a box of tissues given the immature language. Don't overdo it though or you'll go blind.
Old 5th December 2021 | Show parent
  #15256
Here for the gear
 
🎧 10 years
Well said mattiasnyc. I'm sure most of us make a decision based on what platform works best price wise and performance at the time we need to upgrade? I have no allegiance to Intel or AMD and had always used Intel up until my last upgrade.
Old 5th December 2021 | Show parent
  #15257
Gear Nut
 
Skwaidu's Avatar
 
🎧 15 years
A) still wrong. Wouldn't get the 5950 right now for mixing (well, I didn't but I'm pretty confident about this - obv some more and better revealing tests still needed). And also, price.
Old 5th December 2021 | Show parent
  #15258
Gear Guru
 
🎧 10 years
Quote:
Originally Posted by Skwaidu ➡️
A) still wrong. Wouldn't get the 5950 right now for mixing (well, I didn't but I'm pretty confident about this - obv some more and better revealing tests still needed).
That last part is why the point isn't "still wrong". Because I prefaced it all by stating quite clearly "Without any other DAW-use-specific data it seems the reasonable conclusions are:"

That you may have some experience with both the 12900K and 5950X - or equivalent, relevant and comparable setups - that shows the conclusions are just plain wrong is basically anecdotal. I don't discount that experience, but in lieu of any other data set this is the best set we have.

Quote:
Originally Posted by Skwaidu ➡️
And also, price.
I don't really agree with you here. I think this is highly debatable. First we'd have to spec out the systems in order to have a fair comparison and so the first question is if the user is going to get a DDR4 or DDR5 setup. Most users that I've seen in this thread have gone for DDR4 - if I remember correctly. So comparing DDR4 results on the DSP test and excluding the issue of what specific combination of motherboard and memory the user chooses then at a generous (and Dolby Atmos mandated) buffer of 512 samples it looks as follows:

12900K - 385 instances for $620
5950X - 432 instances for $709

So technically you're correct: the 5950X should perform 2% better than it does for the price.

But if I wanted to be.. 'annoying'.. then I'd simply bring up all the times the counter-argument to "AMD has better price:performance ratio" was "At that price point the extra money doesn't matter". In other words:

1. how many people care about saving $89 on the CPU when they're blowing a lot more in total?

2. wasn't the argument before that we'd want to pick the best performing CPU? Seems to me we'd pick one or the other.
Old 5th December 2021 | Show parent
  #15259
Gear Guru
 
🎧 10 years
Quote:
Originally Posted by muso36 ➡️
Well said mattiasnyc. I'm sure most of us make a decision based on what platform works best price wise and performance at the time we need to upgrade? I have no allegiance to Intel or AMD and had always used Intel up until my last upgrade.
I agree. And it's obviously very difficult to compare platforms when we are looking single-user setups that are upgraded. I mean, the Ryzen 1700 was a great CPU for the money - for me - at the time I upgraded. I'm pretty sure it would have been a poor choice for someone using VSTi however, or for someone wanting 'ultimate' performance.

Likewise today getting a 12900K is going to be the by far best choice for some people, but arguably not for all. Either way competition is great and I'm really happy that Intel "is back" after AMD "came back".

It's a great time to be a content creator.
Old 5th December 2021 | Show parent
  #15260
Lives for gear
 
daskeladden's Avatar
 
Quote:
Originally Posted by psionic ➡️
Uh oh, some bad news for all you Intel haters or AMD fan boys.
Most AMD fanboys are on my ignore list. Cannot stand lying fanboys
Fanboy: Has an evangelical approach to the subject. Has a sick and perverted need to feel like an underdog even though it has no basis in reality (AMD is a gigantic multinational company). Trying to hide, deceive and flat out lie to convince others to convert to their product.
Old 5th December 2021
  #15261
Gear Guru
 
🎧 10 years
How incredibly childish.
Old 5th December 2021 | Show parent
  #15262
Gear Head
 
🎧 10 years
Quote:
Originally Posted by daskeladden ➡️
Most AMD fanboys are on my ignore list. Cannot stand lying fanboys
Fanboy: Has an evangelical approach to the subject. Has a sick and perverted need to feel like an underdog even though it has no basis in reality (AMD is a gigantic multinational company). Trying to hide, deceive and flat out lie to convince others to convert to their product.
So that means you didnt see mattiasnyc's reply? I certainly struck a nerve. Apparently this guy is able to call people "tools" and "childish", total indication that that is a projection of himself, but also some how entitled to feel that he is the point person on being able to dictate his A, B , C , D usage from a ReaXComp benchmark.

Last edited by psionic; 5th December 2021 at 11:25 PM..
Old 5th December 2021 | Show parent
  #15263
Gear Head
 
🎧 10 years
Quote:
Originally Posted by mattiasnyc ➡️


The wafer is 10 NANO meters?.. "nm" refers to something else, and it's measured differently by each company. Jesus...

If you're going to try to look smarter than "fan boys" then what you teach should probably be at least correct. Otherwise you look like a tool.



The cores aren't going to perform better because of the chipset. What the HEDT line will bring is more PCIe lanes, i/o, memory channels etc.

Secondly, the relevant tests that have been done that actually use a DAW can be found here, and they look as follows:









Without any other DAW-use-specific data it seems the reasonable conclusions are:

A: If you are going to run mainly DSP and not virtual instruments the best CPU is the 5950X.

B: If you are considering the 12900K with DDR4 and are using a lot of VIs at low buffer sizes the 5950X is still better, and so is the 5900X.

C: If you are using a lot of VIs at low buffer sizes the 12900K with DDR5 is better.

As for the future it's looking really great for everyone on 'both sides' of the fence. Because AMD will release their 3D cache on updated Zen 3 chips soon (announces in January), and later in the year they'll move to Zen 4 on a new platform with PCIe 5.0 and more importantly DDR5.

Now, if you look at the charts above you'll see the very large gains made by using DDR5 (or at least so it seems). If those gains carry over to Zen 4 then the pendulum will have swung back again. Because at that point you're likely looking at about 10-15% uplift from 3D cache, another 15% from the new process node (that's the "NM" again) and architecture, and then the DDR5.



That's as impressive as AMD's Zen architecture for exactly the same reason; just like AMD made processors that were behind the curve and didn't innovate Intel was left behind and didn't really innovate. Now this architecture looks impressive, but that's just because they were stagnant before due to a lack of serious competition.

Rather than speaking about this the way you do (see below) you should be ecstatic that AMD did so well and that Intel now is doing so well. We finally have competition that's driving prices down for great performance for everyone on either platform.

No need for the "fan boy" nonsense.



I'm not surprised you have a box of tissues given the immature language. Don't overdo it though or you'll go blind.
Unfortunately I dont have time to reply to this at the moment (Im working), but I certainly will. Also nm is certainly nanometer, it applies to the wafer and the distance between transistors on that wafer. The lower the NM the more transistors it can accommodate with less power and with less heat.

"Quote regarding NM": NP Resolution is defined to be the minimum feature dimension that can be transferred with high fidelity to a resist film on a semiconductor wafer. ... Throughput is the number of wafers that can be exposed per hour for a given mask level and is thus a measure of the efficiency of the lithographic process.
Old 6th December 2021 | Show parent
  #15264
Gear Guru
 
🎧 10 years
Quote:
Originally Posted by psionic ➡️
Unfortunately I dont have time to reply to this at the moment (Im working), but I certainly will. Also nm is certainly nanometer, it applies to the wafer and the distance between transistors on that wafer. The lower the NM the more transistors it can accommodate with less power and with less heat.
AMD and Intel aren't using the term that denotes an absolute measurement, it's more about the process. Intel's previous 14nm was actually smaller than that name would imply relative to AMD's 7nm.

Quote:
Originally Posted by psionic ➡️
"Quote regarding NM": NP Resolution is defined to be the minimum feature dimension that can be transferred with high fidelity to a resist film on a semiconductor wafer. ...
I'm not sure where you pulled that quote from, but my googling says it's from a document that actually doesn't say what you imply.

derBauer actually shaved off a part of some CPUs and took some images of it and it's pretty interesting;

https://youtu.be/1kQUXpZpLXI?t=720

And the more pop-culture talk about "nm" regarding CPUs by Linus if that's your fancy ("Should You Believe CPU Marketing? - Process Nodes Explained")

So again, the numbers aren't really telling us much at all. Only that TSMC 7nm is different and "better" than 10nm from the same company with the same design... for example.
Old 6th December 2021 | Show parent
  #15265
Lives for gear
 
🎧 10 years
Quote:
Originally Posted by mattiasnyc ➡️
AMD and Intel aren't using the term that denotes an absolute measurement, it's more about the process. Intel's previous 14nm was actually smaller than that name would imply relative to AMD's 7nm.



I'm not sure where you pulled that quote from, but my googling says it's from a document that actually doesn't say what you imply.

derBauer actually shaved off a part of some CPUs and took some images of it and it's pretty interesting;

https://youtu.be/1kQUXpZpLXI?t=720

And the more pop-culture talk about "nm" regarding CPUs by Linus if that's your fancy ("Should You Believe CPU Marketing? - Process Nodes Explained")

So again, the numbers aren't really telling us much at all. Only that TSMC 7nm is different and "better" than 10nm from the same company with the same design... for example.
I'm afraid facts and reason won't get you far here Mattias
Old 6th December 2021 | Show parent
  #15266
Gear Head
 
🎧 10 years
Quote:
Originally Posted by mattiasnyc ➡️
AMD and Intel aren't using the term that denotes an absolute measurement, it's more about the process. Intel's previous 14nm was actually smaller than that name would imply relative to AMD's 7nm.



I'm not sure where you pulled that quote from, but my googling says it's from a document that actually doesn't say what you imply.

derBauer actually shaved off a part of some CPUs and took some images of it and it's pretty interesting;

https://youtu.be/1kQUXpZpLXI?t=720

And the more pop-culture talk about "nm" regarding CPUs by Linus if that's your fancy ("Should You Believe CPU Marketing? - Process Nodes Explained")

So again, the numbers aren't really telling us much at all. Only that TSMC 7nm is different and "better" than 10nm from the same company with the same design... for example.
Why is smaller NM processes better?: Quoted from the article below:
CPUs are made up of billions of transistors and are housed in a single chip. The smaller the distance between transistors in the processor (in nm), the more transistors can fit in a given space. As a result, the distance traveled by electrons to perform useful work is reduced. This ultimately results in faster computing power, less energy consumption and heat dissipation, less thermal output around the board, smaller die size, which ultimately reduces costs and increases transistor density of the same size, resulting in more cores per chip. Intel currently employs 10nm or 14nm technology, while TSMC employs 7nm technology. These are the processor's lithography.


https://ourtechroom.com/tech/nm-in-p...20is%20reduced.
Old 6th December 2021 | Show parent
  #15267
Gear Head
 
🎧 10 years
Quote:
Originally Posted by psionic ➡️
Why is smaller NM processes better?: Quoted from the article below:
CPUs are made up of billions of transistors and are housed in a single chip. The smaller the distance between transistors in the processor (in nm), the more transistors can fit in a given space. As a result, the distance traveled by electrons to perform useful work is reduced. This ultimately results in faster computing power, less energy consumption and heat dissipation, less thermal output around the board, smaller die size, which ultimately reduces costs and increases transistor density of the same size, resulting in more cores per chip. Intel currently employs 10nm or 14nm technology, while TSMC employs 7nm technology. These are the processor's lithography.


https://ourtechroom.com/tech/nm-in-p...20is%20reduced.
I think what youre getting at is the few companies who can actually make these wafers have different architectural designs, where as the 10nm TSMC process is no where close to the Intel 10nm process, due to architecture of the cpu, so were not comparing those directly, what we are doing... is saying that the smaller the nm process the better for any of these companies.
Old 6th December 2021 | Show parent
  #15268
Gear Guru
 
🎧 10 years
Quote:
Originally Posted by psionic ➡️
I think what youre getting at is the few companies who can actually make these wafers have different architectural designs, where as the 10nm TSMC process is no where close to the Intel 10nm process, due to architecture of the cpu, so were not comparing those directly, what we are doing... is saying that the smaller the nm process the better for any of these companies.
Yes.
Old 6th December 2021 | Show parent
  #15269
AOS
Here for the gear
 
Quote:
Originally Posted by AOS ➡️
Thanks Pictus for the prompt response. I will follow all your guidance when building the PC. Regarding the GFX card I have managed to purchase the Gigabyte GeForce GTX 1650 4GB OC GPU. Please confirm I will have no issues with this one. I will look into a case with front USB-C. You have been very helpful buddy thank you. As you can imagine I am a little nervous coming from Intel
Hi Pictus,

Had to send the Gigabyte 1650 back as it was too noisy (Fan constantly on) Can you recommend a quiet 1660 Super which will be efficient/sufficient.

Thx buddy
Old 6th December 2021 | Show parent
  #15270
Lives for gear
 
Pictus's Avatar
 
Quote:
Originally Posted by AOS ➡️
Hi Pictus,

Had to send the Gigabyte 1650 back as it was too noisy (Fan constantly on) Can you recommend a quiet 1660 Super which will be efficient/sufficient.

Thx buddy
They are all very expensive :(
https://pcpartpicker.com/products/vi...t=price&page=1
The first in the list is the https://pcpartpicker.com/product/GzP...06g-p4-1068-kr
As you can see in the review of the sister card, EVGA implemented Fan Stop
https://www.techpowerup.com/review/e...-ultra/34.html

I think all Gigabyte 1650 with 2 fans have fan stop, but if the
case temperature is high they will not stop.

What I do with my GPUs is to change to 120mm fans



And use this fan control to make the case fans react to both GPU/CPU temp.
https://linustechtips.com/topic/1099...n-replacement/
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 4619 views: 627916
Avatar for Ryppa
Ryppa 1 week ago
replies: 55 views: 37964
Avatar for IM WHO YOU THINK
IM WHO YOU THINK 13th October 2020
replies: 98 views: 43882
Avatar for dfghdhr
dfghdhr 5th June 2021
replies: 1296 views: 193980
Avatar for heraldo_jones
heraldo_jones 1st February 2016
Post Reply

Welcome to the Gearspace 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…

Forum Jump
Forum Jump