Originally Posted by Skwaidu ➡️

You would possibly get a benefit from going to CU-DIMM. Search for my reverb count comparison in this thread. Isn't most of Acustica convolution-based - so logic would follow that it would benefit from memory throughput like reverbs (in my test both algo and convolution-based gained a massive boost with just memory performance changed, though the latter slightly less so). I wonder if someone has compared memory perf with Acustica...?

(As an aside, reading this thread the amount of Acustica love is still baffling, I remain deeply uninterested in the subject and refuse to install them on my new machine.)

Originally Posted by Magnus_N ➡️

Thanks for bringing cudimm up again. In what way is cudimm different than regular dimm, and why would that impact on convolution-based performance? According to Pete Kaine above, differences were negligible when he tested it and compared it to regular dimm.

I will look for your reverb test.

You are correct that Acustica is all about convolution.

Best,

Magnus

Originally Posted by momomel ➡️

Last week, the 14700K was selling for $200 or so on Amazon, so I upgraded from a 12700K, but the performance only improved by about 10-15% in reality, lol. I think DAW bench test score is not realistic.

Originally Posted by daskeladden ➡️

CUDIMM is not only about speed according to Intel it offers improved stability and reliability as well...so what is not to like?

Originally Posted by uOpt ➡️

This implies that DIMMs are unstable, which is FUD.

If you want improved reliability you want ECC memory (and not the lame DDR5 built-in nonsense). And ECC is not available in CUDIMM format yet.

Originally Posted by demonkoryu ➡️

CUDIMMs are stable at *higher frequencies* because they have their own clock generator. Normal DIMMs are clocked from the mainboard and that's the primary reason why AMD's 9XXX CPUs can't run RAM at as high a frequency as Intel can, and with 4 RAM sticks it's even worse.

Originally Posted by Lesha ➡️

Single thread performance should be 10% faster, while multithread should be around 35%.

https://www.cpubenchmark.net/compare...ntel-i7-14700K

Originally Posted by norbury brook ➡️

it's completely realistic , it's a bench mark test, how can it not be?

10-15% is a really good upgrade, why are you unhappy?


M

Originally Posted by elegentdrum ➡️

Hello all. Long ago I was active, today I have a question. Hoping this is the right thread to post this.

I'm seeking suggestions for a Motherboard & CPU to upgrade my current situation. Budget is 500$ or less.

Currently using an 11 year old PC from Sweetwater. A creation station. System works ok, but desire to run at lower buffers. Currently I have to run at 256 buffers to be stable, looking to reduce that to 64. Few specifics: Cubase 14/96K

The reason I want to run buffers that low is E-drum use combined with the "all on" studio setup.

Here is the list of parts I desire to re-use on a new Motherboard. Ram being the most restrictive.
The current Motherboard is Z370 current CPU is 3.2 8600
RME MADI FX (A32/SSL AL/AX/RME QS8 B2in/out)
64 Gigs ram
7 port USB
2XUAD Octo cards

I recently took a 4 channel network card out and put in an older 1030 vid card. That helped a bit by moving the Vid ram away from shared.

Please suggest a Motherboard & CPU for me

Originally Posted by uOpt ➡️

That's still the wrong way to put it. Nobody forces a user to overclock non-CUDIMMs to where they are not stable. They are stable at the prescribed frequency (except for lacking ECC but let's leave that aside).

Originally Posted by demonkoryu ➡️

My 4x DDR5 are running at 3600 with AMD while CUDIMMs would be running 6000.

Originally Posted by daskeladden ➡️

CUDIMM is not only about speed according to Intel it offers improved stability and reliability as well...so what is not to like?

Originally Posted by demonkoryu ➡️

CUDIMMs are stable at *higher frequencies* because they have their own clock generator. Normal DIMMs are clocked from the mainboard and that's the primary reason why AMD's 9XXX CPUs can't run RAM at as high a frequency as Intel can, and with 4 RAM sticks it's even worse.

Originally Posted by uOpt ➡️

How do you know that?

Originally Posted by Pete Kaine ➡️

The price?



Doesn't the 9000 series only support CUDIMMs in "bypass" mode however, which essentially ignores the CKD entirely?

Interesting through, might finally be a solution to getting 192GB up and running on Ryzen at none donkey speeds. I've only got hands on with 24GB Dimms so far unfortunately, I'll look forward to getting ahold of larger kits in due course.

Originally Posted by demonkoryu ➡️

Indeed AMD doesn't benefit from CUDIMMs. When I built my 192GB AMD 9950X system a month or so ago, there was no supported way of clocking any higher than the 3600 MT/s I'm currently running. You might be able to reach higher speeds by manual configuration but that's outside the safe (stable) zone. It's a shame.

Originally Posted by Pete Kaine ➡️

The price?

Originally Posted by Lesha ➡️

Well, it is anybody's guess if this will perform at 8000 on your motherboard.

G.SKILL Announces World's First Large Capacity 128GB (64GBx2) DDR5 Overclocked Memory Kit at DDR5-8000

Originally Posted by ANAYV ➡️

Hi I am looking for some tips to set up my cheapo PC , an i5-13500, 32GB RAM, 1TB SSD( for OS and Software programs) and a 500GB SSD drive for storage . Integrated GPU.

Using Studio One 7 Pro.

I upgraded from a i7 8700.

That had 6 cores and 12 log.
This newer PC has 14 cores and 20 log.

I opened a song and thought the CPU meter in S1 would show lower percentage...but I think I need to tweak this a bit to get better performance out of it.

Thanks for any help.

Originally Posted by DEA ➡️

So, whatever happened to Pictus?

Seems like he left in a huff in the fall of 2024 (I think) and has not been seen in this neck of the forum since.

Perhaps he could not handle criticism that not everyone is one the same page regarding what are *useful* system optimizations?

Originally Posted by ANAYV ➡️

Hi I am looking for some tips to set up my cheapo PC , an i5-13500, 32GB RAM, 1TB SSD( for OS and Software programs) and a 500GB SSD drive for storage . Integrated GPU.

Using Studio One 7 Pro.

I upgraded from a i7 8700.

That had 6 cores and 12 log.
This newer PC has 14 cores and 20 log.

I opened a song and thought the CPU meter in S1 would show lower percentage...but I think I need to tweak this a bit to get better performance out of it.

Thanks for any help.

Originally Posted by DEA ➡️

So, whatever happened to Pictus?

Seems like he left in a huff in the fall of 2024 (I think) and has not been seen in this neck of the forum since.

Perhaps he could not handle criticism that not everyone is one the same page regarding what are *useful* system optimizations?

Originally Posted by daskeladden ➡️

Things I would try:
- Turn on ultimate performance or high performance in Power Plan
- Turn on XMP profile 1 in bios

Originally Posted by troggg ➡️

Two things are true about Pictus:

1) He freely gave of his time to help people in ways he believed were beneficial.

2) Part of that help involved recommending installing multiple small utilities, often developed by one-man run shops which could come and go, that required constant maintenance and updates, which was appealing to him as a tweaker, but was also out of mainstream recommendations.

Anyone reading this is free to also prioritize tweaking over production.

Personally, since we have a window into actual Microsoft development on this forum, I'm a lot more inclined to buy into Pscyhlist1972's philosophy of leaving Windows well enough alone most of the time—an opinion seconded by power builders, testers, and users here with unassailable track records.

Originally Posted by Kyle P. Gushue ➡️

Id stress test, use latency mon, and have a heavy daw session handy. Then repeat during and after tweaking.

(No particular order)

Stress test
Occt
https://www.ocbase.com/

Huge List Of Tweaks By Pictus

DPC latency better with AMD graphic cards (3 card comparison)

Phyclist Windows 10/11 Tweak Guide

https://devblogs.microsoft.com/windo...-guide-part-1/

Psychlist disabling "finish setting up PC" nag and other suggestions

Windows 11 is rolling out ...


----

Does anyone know which memtest is the better one? Looks like passmark owns memtest86, (memtest86.com) and there's the open source memtest+ (memtest.org)

Originally Posted by Kyle P. Gushue ➡️

Are the builders like Scan not doing any performance tweaks, just a stock windows install?

i have been leaving installs pretty much stock with this round of installs over here at home, but I also more power than I need atm.

Originally Posted by DEA ➡️

Perhaps he could not handle criticism that not everyone is on the same page regarding what are *useful* system optimizations?