Why Is AMD Selling Broken Playstations?

[AlexTheGreatish]

We bought a PC that uses the exact same silicon as the PS5.. and it's only a bit broken!

 

 

 

[CerealExperimentsLain]

This doesn't seem to be publicly listed on YouTube at the moment.

[RZF77X]

Damn this looks interesting yk

[mariushm]

Another clickbait title.

 

It's not broken, it does everything it claims/advertises to do.

 

You're not saying all Intel processors are broken because they can't reach 5 ghz on all cores and they're binned and sold under some sku number.

 

[GodOfROG]

52 minutes ago, AlexTheGreatish said:

We bought a PC that uses the exact same silicon as the PS5.. and it's only a bit broken!

 

Can i pay for someone to hack this into a PS5?

[BuckGup]

34 minutes ago, mariushm said:

Another clickbait title.

 

It's not broken, it does everything it claims/advertises to do.

 

You're not saying all Intel processors are broken because they can't reach 5 ghz on all cores and they're binned and sold under some sku number.

 

This is LTT where spreading misinformation to their community is dismissed because they want more money. Sorta dumb if your channel is based upon educating your community if you ask me 

[Kronoton]

"Clickbait" ? Well yes, but what isn't these days on UTube?

 

"Misinformation"? If you start from the assumption that this is the PS5 chip then yes it is "broken" just as one could call some i5 broken i7s (or whatever) but that would only confuse what it is, while calling this  a "broken PS5" makes it clearer.

 

I had seen these offered both locally and mentioned in discussions on this forum (pointing to Asian markets) but the pricing is pretty close to 5600G systems which should perform better (or not??) and offer a clear upgrade path (if you stay away from OEM HP boxes).

[mariushm]

Dude, the title is "Why is AMD selling broken Playstations"  ... which is completely wrong and clickbait.  I would not complain if they said "Why is AMD selling broken PS5 APU chips"   but the "Playstations" in title tricks people into thinking of actual consoles,

 

AMD is selling a kit with a processor clearly advertised as 8 cores, no integrated graphics, and with only 4 pci-e lanes, on a custom motherboard and with dedicated gddr5/gddr6 onboard. There's no mention anywhere about playstation in the AMD advertising for the product and only the knowledgeable person would know these are processors made for PS5 which have failed some validations.

 

The kit could be usable in other places, could be a lightweight office pc, could be the starting point for some educational pcs, could be used for example in kiosks, advertising display boards, just put a cheap video card inside for 2d rendering / movie playback etc ... but the price is not right, it should be under $200 for the kit.

 

Oh ... and it's also nothing new ... they did the same thing AMD A9-9920 ... I think it's Xbox One S cpu or whatever, here's a chinese motherboard with this APU chip soldered to the board :

 

 

 

[riklaunim]

Motherboards only or mobo  + RX550 are available from bigger PC wholesalers and they were quite early on more problematic markets like Brazil so at first it looked like they wanted something chip-ish on markets where logistics or import duties make it problematic. And with the availability it looks like something more than just rejected silicon (or they are burning a lot of dies on 7nm due to some absurd Sony requirements).

 

In Poland the board only is ~358USD, and 480 USD for board + RX550.

[ToboRobot]

Let's say there have been 10 million PS5 made. 

Let's assume that 0.5% of PS5 that were built were defective silicon but still had limited functionality.

 

That is 50,000 units per 10mil consoles. 

The economics on this seem to be bad.

[Helpful Tech Witch]

34 minutes ago, ToboRobot said:

Let's assume that 0.5% of PS5 that were built were defective silicon but still had limited functionality.

 

Thats a reallly conservative statement. 

The chances of a bit of a complex thing with several diffrent parts like a gpu is higher than you think. 

And apu with a broken gpu would still be fine as a cpu

So instead of throwing away some broken peices of silicon, they take it and disable it. Id have to guess at least 20% have some gpu issue, epsecially since the zen 2 cpu itself has a 70-80% yeilds 

[Nystemy]

Plugging in 100-120 Volt AC into a 200-240 Volt AC PSU will generally not end up with much more than disappointment.

 

In all my years of electronics engineering, I have never had a PSU blow up due to too low AC input voltages. Using a variac and slowly turning up the voltage is actually a fairly effective trouble shooting method when repairing power supplies. _{(Though, sometimes the PSU burns up when doing that test, but then it is due to a seriously failed component, like a shorted bridge rectifier.)}

 

So personally I wouldn't in the slightest worry about the 200-240 volt only PSUs. The one's I worry about is the one's that don't support 200 volt operation. Though, PSUs that have the good old switch on them can be a death trap if set to 110 volts when plugging them into 230... Thankfully modern power supplies tend to use a more universal design from the get go, especially when having PFC.

[Freakwise]

They selling it here for around $560 conversion for the same spec as lmg have but with Silverstone sg13 as a case (still crappy psu though), i guess it's good value but then again that pcie lane kinda of disappointment

[ironbound]

Would be interested in poking around in the registers and PSP, lots of undocumented things can be hidden on a CPU till someone digs in to find them 

[jSONBB]

The BIOS for this board CAN be edited with AMIBCP

Maybe someone at LTT can do some... poking around?

Version 5.02.0031 opens it.

 

P.S. there is "Memory Clock" controls among other various options to unlock;

Set "Access/Use" to "USER"

default RAM speed is 800MHz

Edited October 24, 2021 by jSONBB
RAM speed is 800MHz

[Nathanpete]

Can anyone explain the math they did at 13:06. How do u calculate bandwidth  with memory type, size, and requency?

[SteelSkin667]

1 hour ago, Nathanpete said:

Can anyone explain the math they did at 13:06. How do u calculate bandwidth  with memory type, size, and requency?

They didn't do any math, Linus is just reading benchmark results.

 

The way you would calculate memory bandwidth is by multiplying its data rate (usually expressed in MT/s or in Gbps per pin) by its bus width (64 bits per channel in the case of DDR type memory).

[mariushm]

8 hours ago, Nystemy said:

Plugging in 100-120 Volt AC into a 200-240 Volt AC PSU will generally not end up with much more than disappointment.

 

In all my years of electronics engineering, I have never had a PSU blow up due to too low AC input voltages. Using a variac and slowly turning up the voltage is actually a fairly effective trouble shooting method when repairing power supplies. _{(Though, sometimes the PSU burns up when doing that test, but then it is due to a seriously failed component, like a shorted bridge rectifier.)}

 

So personally I wouldn't in the slightest worry about the 200-240 volt only PSUs. The one's I worry about is the one's that don't support 200 volt operation. Though, PSUs that have the good old switch on them can be a death trap if set to 110 volts when plugging them into 230... Thankfully modern power supplies tend to use a more universal design from the get go, especially when having PFC.

 

You could damage the Active PFC circuit and potentially the bridge rectifier, if the power supply was 230v only to save costs and use undersized components in the Active PFC circuit

Boosting 110v to 400v-ish is harder on components compared to 230v to 400v-ish - there can be higher current spikes which could damage mosfets or diodes.

 

Sure, no damage would happen when you start a psu with no load or very low load, that's why starting with variac is relatively safe. But if the computer pulled 200w or more things may be different.... it may start but it may die eventually after a few hours or days due to overheating or due to components being used above their maximum limits and gradually being damaged internally.

 

So basically it's not safe to assume that any 230v only power supply won't work with 110v, just as it's not OK to assume a 230v only power supply is like that because it doesn't have a power factor correction circuit (as Linus said in video).

 

 

[SteelSkin667]

The PCIe situation is a bit puzzling. That M.2 slot in the PS5 is exposing 4x Gen 4 lanes, and then that integrated SSD is also connected through another 4. Presumably once the chip relocated to another board these eight lanes would still be available. Maybe some of the I/O had to be disabled at the same time as the GPU?

[Kronoton]

33 minutes ago, SteelSkin667 said:

Maybe some of the I/O had to be disabled at the same time as the GPU?

 

Maybe, I'd rather guess that 1 PCIe lane gets eaten up by SATA and doubt you could do a x7 slot.

[Granular]

Would have been interesting to see some breakdown of which use cases are more affected by the increased memory latency and which ones are not (in place of the irrelevant riffing on the PSU, for example).

[Nystemy]

2 hours ago, mariushm said:

 

You could damage the Active PFC circuit and potentially the bridge rectifier, if the power supply was 230v only to save costs and use undersized components in the Active PFC circuit

Boosting 110v to 400v-ish is harder on components compared to 230v to 400v-ish - there can be higher current spikes which could damage mosfets or diodes.

 

Sure, no damage would happen when you start a psu with no load or very low load, that's why starting with variac is relatively safe. But if the computer pulled 200w or more things may be different.... it may start but it may die eventually after a few hours or days due to overheating or due to components being used above their maximum limits and gradually being damaged internally.

 

So basically it's not safe to assume that any 230v only power supply won't work with 110v, just as it's not OK to assume a 230v only power supply is like that because it doesn't have a power factor correction circuit (as Linus said in video).

 

 

All PSUs have a mains side fuse.
If the PFC circuit draws too much current, the fuse will blow in short order. And most mosfets can safely operate at fairly high currents compared to their rated current, they however do generate a fair bit more heat but most PFC circuits have their mosfet on a heat sink so it will take a bit of time for it to get warm. Though, a lot of PFC controllers also have over current protection.

For PSUs without PFC, some switchmode controllers will just work even if the voltage is too low. However, they will quickly bump up against their max PWM duration, and at that point it can't deliver more power. Its output voltage will likely fall and the power good signal will go low potentially turning off the computer. For resonance mode supplies it is a bit different, but effectively a similar limit is what holes it back _{(all though, resonance mode supplies tends to be heavily input voltage dependent when it comes to regulating their output, so without a PFC circuit to boost the input voltage up to a nice somewhat stable voltage, one would have an infuriatingly hard time to support a wide range of input voltages while retaining the output power specs. In short, almost all resonance mode supplies have PFC.)}. A PSU can effectively only provide power at roughly the square of the voltage it has to work with as long as the output side isn't the limit.

Even under load I wouldn't greatly worry about it. Worst case one blows the usually soldered in mains side fuse. _{(and if the PSU doesn't have a mains side fuse, then it has bigger issues, like being illegal.)}

[Nystemy]

2 hours ago, Granular said:

Would have been interesting to see some breakdown of which use cases are more affected by the increased memory latency and which ones are not (in place of the irrelevant riffing on the PSU, for example).

Yes, it would be nice to see a slew of benchmarks and compare it to a similarly configured PC but with more "normal" memory.

 

I bet a game like CS go won't like the latency, considering how it is somewhat single threaded and already cares more about single core performance above most else. And larger memory latency eats into that performance quite quickly. _{(Cache memory can somewhat save one here, except on a cache miss, then one faces a large stall...)}

 

While a program like blender might have better overall performance. It might even like it since it could potentially make use of all that additional memory bandwidth. However, one might need to tell the rendered to use more threads than what is actually available in the CPU, so that when a thread inevitably stalls it can switch to another that has now got its data. In short, tuning blender rendering can be a fun thing to indulge in, since even small things like adjusting the tile size can make noticeable impacts on render times.

[Kronoton]

22 minutes ago, Nystemy said:

All PSUs have a mains side fuse.

 

Ask TechJesus how much those helped recently.

 

Running a crappy PSU within spec is a risk one might want to take, but out of spec your pretty much doomed.

[Nystemy]

14 minutes ago, Kronoton said:

 

Ask TechJesus how much those helped recently.

 

Running a crappy PSU within spec is a risk one might want to take, but out of spec your pretty much doomed.

That PSU however had a design flaw. since multiple units failed in the same fashion while operating within spec. The exact source of the fault is however a different question. _{(I suspect it being a simple gate breakdown due to a too high Vgs. Something I have seen a few times before.)}

 

But a supply that is designed correctly won't have that issue.

 

The main difference between a cheap power supply and a fancy "high end" one is mainly cheaper 85C rated capacitors from less known brands, fewer comfort features like modular cables _{(connectors are actually fairly expensive, not to mention terminating the cables on both ends...)} and a high end supply also tends to spend a bit extra on powder coating the case and putting in a more silent fan. The core design is fairly boring and don't tend to change. _{(Most of the time, the manufacturer just follows the SMPS controller manufacturer's guidance, usually too the letter without question.)}

 

If a PSU breaks or is dangerous in any way from being undervolted, then it is having a flawed design that wouldn't get regulatory approval _{(bigger manufacturers however knows how to navigate paper work and sometimes gets approved without actually having gone through the procedures. And some really small/no-name companies are just passing under the regulatory radar)}. Surviving a severe brownout is practically a legal requirement.

 

In short.
If the 220 Only PSU blows up or is otherwise dangerous if plugged into 100-120 volt, then it shouldn't be on the market... The other way around however is a completely different situation. Overvolting a 110 only PSU by plugging it into 230 can result in a fire _{(Though, usually the fuse blows in this case.)} and still be considered acceptable by the regulatory bodies.