The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Threadripper 2 Speculation Recorders, Players & Tape Machines
Old 3 weeks ago
  #1
Here for the gear
 

Thread Starter
Threadripper 2 Speculation

Hi guys,

I'm interested in building a new audio rig, currently using a 7700k / 32 GB Ram / 2 TB SSD laptop running Live 10.

While my laptop will remain my live machine, I'd like to build up a studio desktop. I'm not in a hurry though, planned in doing so in August / September, so there's some more time to think everything through.

So I was looking into CPU's and there's Threadripper and the i9 7900X which pulled my attention. So I've been looking for reviews, benchmarks etc. and Threadripper seems to lack in low buffer size performance (< 256) while the i9 seems to be hardly manageable regarding temps.

With the Threadripper update scheduled for August this year, do you think they'll be able to fix the problems for Real Time Audio use?

I understand that this is heavy speculation, and we'll have to wait for the benchmarks. Although I'm not an expert on those topics and I'm pretty sure there are some around, that could maybe predict whether or not the Zen+ architecture will solve the problems or not...

Will I be an AMD user by the end of the year? Or is there an Intel response ready to engage in battle as soon as Threadripper 2 drops? Is it worth waiting even longer?

Thanks a lot!

Cheers!
Old 3 weeks ago
  #2
Lives for gear
 

Quote:
Originally Posted by roadjack View Post
So I was looking into CPU's and there's Threadripper and the i9 7900X which pulled my attention. So I've been looking for reviews, benchmarks etc. and Threadripper seems to lack in low buffer size performance (< 256) while the i9 seems to be hardly manageable regarding temps.

With the Threadripper update scheduled for August this year, do you think they'll be able to fix the problems for Real Time Audio use?
You could just as easily ask if the heat problems will be fixed for the i9.

I think the thing to do is to see how much you want to spend as well as figure out just how many Virtual Instrument voices you need. As long as you're below budget and the CPU can do what you need pick the cheaper option.

For someone like me who does barely any VI work the older Ryzen 1700x with included cooler and a reasonably priced motherboard was a combo hard to beat in terms of price/performance. I could have gotten a lot more performance with a Threadripper, and even more than that with a high end i9, but I just didn't need it at this point.

Quote:
Originally Posted by roadjack View Post
I understand that this is heavy speculation, and we'll have to wait for the benchmarks. Although I'm not an expert on those topics and I'm pretty sure there are some around, that could maybe predict whether or not the Zen+ architecture will solve the problems or not...
I still think the word "problem" is silly. If an AMD chip performs better for the same money at the non-VI part of the DAWbench test, do people go and say the Intel chips have a "problem"? No, they don't. Only AMD chips have "problems".

Now, as far as low latency performance with VI there has already been a couple of tests; one by Techreport and one by Scan. The improvement of Zen+ over Zen is about 30-40%, which in my book is a lot.

One would have to look at the Threadripper more in detail to make sure the current models don't already have those improvements. If they don't then they will have the same improvement.
Old 3 weeks ago
  #4
Gear Addict
I'm eyeing the 7900x or maybe even one of the higher models. Low latency VI performance is critical to me so the i9 seems like the natural choice.

I wonder how significant the difference in low latency performance is between team blue and team red.

I'm coming from an FX9590 so either one would probably be light-years ahead
Old 3 weeks ago
  #5
Lives for gear
 

Quote:
Originally Posted by EMMST View Post
I'm eyeing the 7900x or maybe even one of the higher models. Low latency VI performance is critical to me so the i9 seems like the natural choice.
I would again say that as long as your needs are met anything above that is money out the window. ("needs" also including future needs)

Quote:
Originally Posted by EMMST View Post
I wonder how significant the difference in low latency performance is between team blue and team red.
Click links. Read charts.

Quote:
Originally Posted by EMMST View Post
I'm coming from an FX9590 so either one would probably be light-years ahead
Most likely.
Old 3 weeks ago
  #6
Gear Addict
Quote:
Originally Posted by mattiasnyc View Post
I would again say that as long as your needs are met anything above that is money out the window. ("needs" also including future needs)



Click links. Read charts.



Most likely.
Agreed, over and above is money out the window, but hey...this is gearslutz...

I've read the DAWbench results and poured over benchmarks, reviews and charts. I guess I'm wondering how significant the difference will feel in real world use as opposed to numbers and such.

When I bought my 9590, all the bench results made it look like a monster, and it was for 2014 when I got it, but not as great for audio as I had hoped. If I could do it again, I would have gone with Intel.
Old 3 weeks ago
  #7
Lives for gear
 

As far as I recall the last time AMD had a CPU that held its own in terms of price / performance was a good 4-5 years before you bought that FX CPU. I looked at it briefly but got the impression Intel had better deals. I think I got my previous CPU back in '09 or so. (Yes, I squeezed close to a decade out of it).
Old 3 weeks ago
  #8
Gear Addict
Nice! That's why I'm thinking of getting a beefier i9. I figure if I keep it as a DAW and leave all other computing needs to a different machine, I won't need to upgrade for a long long time.

The FX is fine with a good water cooler, but does not play nicely with RAM, throttling it for some reason. It's still fine as a tracking and mixing rig, but these days I am doing much more track writing with dozens of not hundreds of VIs and I neeeeeeed low latency for that.
Old 3 weeks ago
  #9
Gear Head
The funny thing is.. I got my AMD CPU in 2012. Top of the line FX8350. Dirt cheap too.
Past this six years the performance of it has just... RISEN. WTF. Because of updates and how software have developed.
In the meantime an 2012 Intel has only seen huge drops in it's performance.

I'm in no need of an update as this machine can run anything i throw at it.
Old 3 weeks ago
  #10
Gear Head
Quote:
Originally Posted by EMMST View Post
The FX is fine with a good water cooler, but does not play nicely with RAM, throttling it for some reason.
What do you mean by "RAM throttling"? Do you mean that your RAM keeps changing it's operating frequency?

Quote:
Originally Posted by EMMST View Post
It's still fine as a tracking and mixing rig, but these days I am doing much more track writing with dozens of not hundreds of VIs and I neeeeeeed low latency for that.
Sir. I have to correct you here. For VI's you don't need lots of low latency performance. You need ASIO Guard, or similar tech.

If you have hundreds of VI's i'm sure you are not "actually playing" them all at the same time, right? Just one. Rest are playing back what you played earlier. These can be offloaded to ASIO Guard tracks which run at higher latency in the background.

And the thing you need most is more Cores. That's exactly why I went for an 8-core AMD instead of 2- or 4-core Intels.
Old 3 weeks ago
  #11
Lives for gear
 

Quote:
Originally Posted by mkdr View Post
Sir. I have to correct you here. For VI's you don't need lots of low latency performance. You need ASIO Guard, or similar tech.

If you have hundreds of VI's i'm sure you are not "actually playing" them all at the same time, right? Just one. Rest are playing back what you played earlier. These can be offloaded to ASIO Guard tracks which run at higher latency in the background.
So just so I understand:

You're saying that turning ASIO guard ON will result in far more VI running without any problems, because it automatically adjusts them, all while using very low latency (i.e. 32 or 64 samples) which still allows for "live" feeling playing of a VI?
Old 3 weeks ago
  #12
Gear Head
Quote:
Originally Posted by mattiasnyc View Post
So just so I understand:

You're saying that turning ASIO guard ON will result in far more VI running without any problems, because it automatically adjusts them, all while using very low latency (i.e. 32 or 64 samples) which still allows for "live" feeling playing of a VI?
Yes. Exactly.
Old 3 weeks ago
  #13
Lives for gear
 

hmmmm... I wish there were comparisons between ASIO GUARD on/off in Cubase/Nuendo when using DAWbench then, because that's highly relevant. When I did my DAWbench test without VIs I literally just stopped after a massive amount of plugins using ASIO guard. There simply was no point continuing. I didn't do the VI test because it's not what I do, but I'd think it's very relevant for anyone IF that workflow basically renders at least some concerns moot.

Maybe I'll re-run the test to see how it behaves on my Ryzen 1700.
Old 3 weeks ago
  #14
Gear Addict
Question

Quote:
Originally Posted by mkdr View Post
What do you mean by "RAM throttling"? Do you mean that your RAM keeps changing it's operating frequency?


Sir. I have to correct you here. For VI's you don't need lots of low latency performance. You need ASIO Guard, or similar tech.

If you have hundreds of VI's i'm sure you are not "actually playing" them all at the same time, right? Just one. Rest are playing back what you played earlier. These can be offloaded to ASIO Guard tracks which run at higher latency in the background.

And the thing you need most is more Cores. That's exactly why I went for an 8-core AMD instead of 2- or 4-core Intels.
RE RAM it's fairly well documented that the FX processors throttle memory frequencies.
Here is one such thread Ausu 990fx r3.0 fx 8350,$gskill ram issues - G.SKILL TECH FORUM

As far as VIs, allow me to elaborate.

As I stated I am typically using dozens (sometimes hundreds) of VIs along with several outboard processors and Synths. While bouncing is ideal, it's a hassle in my situation due to the way Ableton handles sidechain routing. Since my clients may ask for any number of changes at a moment's notice, I need to keep the midi data and VI settings, while also using sidechain. I really do need to have low buffer setting to work fast. 128 or lower is ideal. With the FX rig, I'm up to 1024 after about 30 mins of work.

I will check out asio guard for sure!!
Old 3 weeks ago
  #15
Lives for gear
 

Quote:
Originally Posted by mattiasnyc View Post
hmmmm... I wish there were comparisons between ASIO GUARD on/off in Cubase/Nuendo when using DAWbench then, because that's highly relevant. When I did my DAWbench test without VIs I literally just stopped after a massive amount of plugins using ASIO guard. There simply was no point continuing. I didn't do the VI test because it's not what I do, but I'd think it's very relevant for anyone IF that workflow basically renders at least some concerns moot.

Maybe I'll re-run the test to see how it behaves on my Ryzen 1700.

Ultimate Outsider: Comparing DAW Performance of Recent Cubase Versions on Windows
Old 3 weeks ago
  #16
Here for the gear
 

Thread Starter
Quote:
Originally Posted by mattiasnyc View Post
You could just as easily ask if the heat problems will be fixed for the i9.
True, I might have made an unlucky linguistic choice here. The Intels are by no mean less problematic I guess and I wish they'd fix the cooling issue.


Quote:
Originally Posted by mattiasnyc View Post
I still think the word "problem" is silly. If an AMD chip performs better for the same money at the non-VI part of the DAWbench test, do people go and say the Intel chips have a "problem"? No, they don't. Only AMD chips have "problems".
I guess for me and many others, Intel chips have been the standard for the last years and I'm happy to see AMD coming back. I didn't want to say, that AMD chips are problematic by any means. But when I'm heavily dependent on low latency computing power which isn't that strong on the current TR generation, that might just be a counterargument for going the AMD road in my specific case and thus "problematic" for me.

I'm very happy that the second gen Ryzen improved that much and I'm optimistic and looking forward to TR2. Thanks for the links, I was digging on Scan Pro, with no luck though.

Thanks for your input guys. I also agree about budget and going for less, if it suits your needs. This is my working machine though and it pays my bills, so I'm willing to spend some cash. I'm also planning to do more video content in the future, maybe even some streaming, I just want to be ready for the future and have a stable working horse, not worrying about computing resources.
Old 3 weeks ago
  #17
Lives for gear
 

Thanks. That's very interesting. When I did my Ryzen 1700 DAWbench (non-VI) test turning on AG made a huge difference.

Perhaps it's worth while re-evaluating recommendations or the tools used when testing (i.e. DAW+plugins) seeing that we now have huge differences with/without AG as well as depending on what plugin is used.
Old 3 weeks ago
  #18
Gear Maniac
 
Pictus's Avatar
 

Quote:
Originally Posted by roadjack View Post
But when I'm heavily dependent on low latency computing power
which isn't that strong on the current TR generation, that might just be a
counterargument for going the AMD road in my specific case and thus
"problematic" for me.

I'm very happy that the second gen Ryzen improved that much and I'm
optimistic and looking forward to TR2. Thanks for the links, I was
digging on Scan Pro, with no luck though.
Ryzen/Threadripper seems to benefit from faster RAM like G.Skill F4-3400C16D-16GSXW
with tighter settings as you can see here and here.
Old 2 weeks ago
  #19
Lives for gear
 

AMD Reveals Threadripper 2 : Up to 32 Cores, 250W, X399 Refresh

Two interesting factors will be what frequency we'll be looking at; core, turbo and maximum all-core overclock; as well as how memory access will be handled, since there are now two more dies but the same amount of memory channels (with memory controllers on-die it implies two dies having to get to RAM through the other two die's memory controllers, which is a potential increase in latency)....
Old 1 week ago
  #20
Lives for gear
 

Quote:
Originally Posted by mattiasnyc View Post
Perhaps it's worth while re-evaluating recommendations or the tools used when testing (i.e. DAW+plugins) seeing that we now have huge differences with/without AG as well as depending on what plugin is used.
It's why we test with AG off to stop it being skewed, after all, we're testing the CPU, not the sequencer. It's also the reason I test with more than one sequencer too.
Old 1 week ago
  #21
Lives for gear
 

From:https://www.anandtech.com/show/12906...w-x399-refresh
Quote:
For the first generation this meant that each of the two active die would have two memory channels attached – in the second generation Threadripper this is still the case: the two now ‘active’ parts of the chip do not have direct memory access.
Yeah, I was hopeful they might rework that after the feedback on the first generation.
Old 1 week ago
  #22
Gear Maniac
 
Pictus's Avatar
 

In the GIGABYTE's new X399 refresh motherboard there is a Thunderbolt header
near the SATA ports, like in the X399 DESIGNARE EX.
Old 1 week ago
  #23
Lives for gear
 

Quote:
Originally Posted by Pictus View Post
In the GIGABYTE's new X399 refresh motherboard there is a Thunderbolt header
near the SATA ports, like in the X399 DESIGNARE EX.
If TB wasn't enabled for the first version I would honestly not count on it happening for the refresh either... :-(
Old 1 week ago
  #24
Lives for gear
 

Quote:
Originally Posted by Pete Kaine View Post
It's why we test with AG off to stop it being skewed, after all, we're testing the CPU, not the sequencer. It's also the reason I test with more than one sequencer too.
That's understandable.

Reviewers however do tend to make conclusions and recommendations that go beyond the result of the CPU in and by itself, and if - hypothetically - a user won't max out an R7 1700 with included stock cooler because she's using AG, then it's worth mentioning that that's the case rather than that the 8700k is a better buy (which, it arguably wouldn't be since it's more expensive and possibly brings nothing of additional practically usable value to the table).

Not saying you're doing this btw, but having gone through the process of researching CPUs and then building a new rig, plus the shift in plugins and sample rates used, it sort of occurred to me that perhaps there's a better way measuring all of this moving forward. I'm not 100% sure what would be comprehensive enough or the best targeted approach, but it seems to me there's a pretty wide range now between testing at 24/48k regular DAWbench previously chosen plugin versus 24/96k new plugin+Vi or whatever.....

Perhaps a more exhaustive exploration of just how the used plugins work would be good so that everyone understand to whom the various results are relevant....

Quote:
Originally Posted by Pete Kaine View Post
From:https://www.anandtech.com/show/12906...w-x399-refresh


Yeah, I was hopeful they might rework that after the feedback on the first generation.
I'm not sure it was possible while retaining x399 compatibility. I think the x399 boards are limited to quad channel memory, and that the access comes out of certain pins and thus certain dies. So it wouldn't have been possible to send/receive signals from the other two dies.

Even if that would have been possible with an x399 board it would still have been a quad channel setup, meaning that either the memory controllers on die would have to pass data along anyway since there would have been more memory channels in the CPU compared to what's possible to get from the motherboard, or each die would have to drop down to single channel access in order to still total four channels - which in turn would limit the bandwidth for each controller... and I'd think that arguably be way worse for latency etc.
Old 1 week ago
  #25
Lives for gear
 

Quote:
Originally Posted by mattiasnyc View Post
Even if that would have been possible with an x399 board it would still have been a quad channel setup, meaning that either the memory controllers on die would have to pass data along anyway since there would have been more memory channels in the CPU compared to what's possible to get from the motherboard, or each die would have to drop down to single channel access in order to still total four channels - which in turn would limit the bandwidth for each controller... and I'd think that arguably be way worse for latency etc.
It's just the nature of the design they've implemented. X79/X99 etc... have been quad channel designs as well, but without dividing up the memory addressing the way AMD has.

Arguably for scalability, the AMD design makes a lot more sense when dealing with none real-time handling so I didn't expect a redesign anyway as it's working as intended for most market segments. Also as you say, doing so would ruin any chance at X399 compatibility and they've already made it clear they had no desire to do that at this point.
Old 1 week ago
  #26
Gear Head
Old 1 week ago
  #27
Lives for gear
 

Quote:
Originally Posted by mkdr View Post
Not a very good article to base any conclusions on as the testsystem seems to have some issues as it can't use all cpu power.
I recall maybe Pete (Scan) finding limits in all CPUs as far as how much of them end up being used. I don't think it necessarily invalidates the test.
Old 1 week ago
  #28
Lives for gear
You already have a 7700k, why upgrade so quickly? Are you really tapping out your cpu? 7700k is still a new cpu. Upgrading every generation is a waste and not worth the time it takes to get all the software reinstalled imo. Better to wait longer when you will get more of an upgrade.
Mentioned Products
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