NowYou'rePlayingWithPower
Well-known member
Thanks for the followup!
japetus
Moderator
Always a good day when there is an Olaf sighting!
Exo-metal
Well-known member
NowYou'rePlayingWithPower
Well-known member
I already see a mod for these.
What's interesting is that he talks about delay around 5 minutes in which another tester states is 2ms, but according to the datasheet of the ir dsp it can be as low as 0.8 to as high as 6ms. Which is going to be dependent on both temps and the ir selected. They should have added a couple dollar copper heatsink to that processor, imo. If someone buys one of these, I recommend that to keep response times as low as possible or at least more consistent. It doesn't need it for reliability or stability imo, but it would help drastically lengthen use time before closing in on those higher delay times.
Elric
Active member
This is a bunch of nonsense. The DSP is clocked where it’s clocked and the latency of the convolution is going to depend on the length of the IRs, which is fixed for the system, and how many clock cycles the algorithm/software takes. If it measured at 2ms that is how long it’s gonna take.
NowYou'rePlayingWithPower
Well-known member
Hence why I stated dependent on the ir selected. The rest is absolutely true.
JTyson
Well-known member
Strange
New member
The X88-IR is manufactured by Boutique Amplification who, by the way, builds Soldano, Friedman, Egnater, Diezel, Synergy, Morgan to name a few. The construction is similar across platforms. I don't see the problem.
NowYou'rePlayingWithPower
Well-known member
To add source material outside of my opinion the manufacturer shows heat dissipation of the dsp package on it's own as lower than 40% with a maximum still under 50% dependent upon voltage applied. They also signify the importance of using a cooling stack solution including a heatsink and proper bottom plate (which is why I asked earlier for pictures of the pcb under the dsp) Meaning if the latter is not there, the dsp manufacture would not honor any warranty claims to the builder. Not that I expect these to fail, because...
Whether or not the dsp will run in spec is not the point. Thermal throttling is the eventual point that will 100% affect top available clock speeds and consistent efficiency. You cannot expect it to not happen when it's only capable of dissipating 40-50% of it's own heat. Add ambient temps from inside the case as well as in hotter climates, and I would expect to happen nearly immediately. It's an ARM 5v processor, so personally if I bought one, not only would I add a heatsink, but I would add a PWM fan to run off of the 5v rail. In fact depending on the VRM solution, it may even be worth swapping the DSP(not for beginners, since BGA) for a better variant since they tend to range a whopping $5-10.
Given the measurements of 2ms and the instructions required for each process, if anything I would say that either thermal throttling is already occurring in these tests and/or poor coding. I notice 2ms on a mouse or monitor, and we're looking at as high as +5.2ms from best performance to worst. Really these things need to be designed around parallel processing in the future, not necessarily for latency, but for maximum bandwidth saturation to ensure that as many nuances are present. Couple that with 3d stacked memory and we may see chips 99.999999999999etc...% on point with "feel under the fingers"
https://www.st.com/resource/en/appl...for-stm32-applications-stmicroelectronics.pdf
Here one can find the total time required for each instruction pathway of the architecture if you download the datasheet on the link
https://www.st.com/en/microcontrollers-microprocessors/stm32f429ng.html
Anyway you look at it, I'm not talking about some magical fix to zero latency or something that suddenly increases processor speed, but something that should have been there in the first place to give both the maximum level of it's processing speed for longer periods as well as a longer lifespan for the product itself. It's a $1 heatsink for crying out loud and is required to be used for a warranty from the chipmaker. I would still expect the failure rate to be very low without one. Far lower than cost of adding a heatsink to each device (say $1 heatsink and 1/500 units fail, so spend $500 per 500 units to prevent or $5-10 on the one that failed), and I'm sure Soldano came to the same conclusion. That doesn't make it right or optimal. I stand by my recommendation 100%.
Whether or not the dsp will run in spec is not the point. Thermal throttling is the eventual point that will 100% affect top available clock speeds and consistent efficiency. You cannot expect it to not happen when it's only capable of dissipating 40-50% of it's own heat. Add ambient temps from inside the case as well as in hotter climates, and I would expect to happen nearly immediately. It's an ARM 5v processor, so personally if I bought one, not only would I add a heatsink, but I would add a PWM fan to run off of the 5v rail. In fact depending on the VRM solution, it may even be worth swapping the DSP(not for beginners, since BGA) for a better variant since they tend to range a whopping $5-10.
Given the measurements of 2ms and the instructions required for each process, if anything I would say that either thermal throttling is already occurring in these tests and/or poor coding. I notice 2ms on a mouse or monitor, and we're looking at as high as +5.2ms from best performance to worst. Really these things need to be designed around parallel processing in the future, not necessarily for latency, but for maximum bandwidth saturation to ensure that as many nuances are present. Couple that with 3d stacked memory and we may see chips 99.999999999999etc...% on point with "feel under the fingers"
https://www.st.com/resource/en/appl...for-stm32-applications-stmicroelectronics.pdf
Here one can find the total time required for each instruction pathway of the architecture if you download the datasheet on the link
https://www.st.com/en/microcontrollers-microprocessors/stm32f429ng.html
Anyway you look at it, I'm not talking about some magical fix to zero latency or something that suddenly increases processor speed, but something that should have been there in the first place to give both the maximum level of it's processing speed for longer periods as well as a longer lifespan for the product itself. It's a $1 heatsink for crying out loud and is required to be used for a warranty from the chipmaker. I would still expect the failure rate to be very low without one. Far lower than cost of adding a heatsink to each device (say $1 heatsink and 1/500 units fail, so spend $500 per 500 units to prevent or $5-10 on the one that failed), and I'm sure Soldano came to the same conclusion. That doesn't make it right or optimal. I stand by my recommendation 100%.
JohnnyGtar
Well-known member
Yer killin' me! LOL
Exo-metal
Well-known member
Love having you and your stunning expertise here on the forum.
NowYou'rePlayingWithPower
Well-known member
I'm just a techy and computer nerd moonlighting as a guitar player
Really don't feel that the lack of cooling solution is going to cause immediate harm. So, I wasn't trying to be an alarmist or imply that the processor would somehow overclock itself. Just feel that someone would get longer playtime time with maximum/optimal clock speeds (in turn, shorter delay time) as well as not looking on ebay in ten, twenty, or fifty years for a replacement board with the same dead/dying processor issue.I honestly wouldn't question the lack of one in most other applications, it's just inside of a box with hot vacuum tubes, not far from a transformer, and in the hottest area of the case while trying to perform tasks that are time dependent for maximum enjoyability. It's also possible thermal throttling would only drop the clocks no more than a few percent and no one would be the wiser. Worst case it's just a cheap mod that doesn't do anything but extend the product life and is reversible in minutes. I wasn't trying to argue with anyone or imply magic, I was just thinking outloud on an inexpensive mod that at worst would cause no harm. Those who are only going to only ever plug this into a power amp aren't ever going to need to concern themselves with it, either way.
Anyways, I'm going to engrave my name on some matching purple anodized heatsinks and offer them up for $30, so please don't let people know you can buy one for 2-5% of that.
psychodave
Well-known member
Hope you're doing well Olaf.
JerEvil
Well-known member
Mine gets here Wednesday
MrWoodchuck
Well-known member
All Diezel amps are still built in Germany. Boutique Amps just builds the Diezel cabs and pedals in the US. This is the same with Bogner where all of the Ecstasy, Helios, Pandora, Uber Ultra are still built by Bogner and the pedals are built by Boutique Amps.
Exo-metal
Well-known member
Jer, bomb cyclone bro, bring it!!
Last edited:
JerEvil
Well-known member
It’s sitting at my front door as we speak! Lol
Just got this guy yesterday… you see where I’m going with this?
Hope to do my X88 vs Mezza Nirvana shootout this weekend.
Attachments
Exo-metal
Well-known member
Just left my front door
Strange
New member
I mis-spoke, yes the Bogners are still built by Reinhold and Diezels are built in Germany. But Friedman's, Morgans, Egnaters, Synergy and Soldano most likely share manufacturing.
Elric
Active member
Hoping you do caps of both, would love them for QC and ToneX.