So i got myself thinking and was wondering, is the DECR-1000A (And any derivatives) more reliable than your typical retail size PS3 (phat, slim and super slim) the reasons i could think it may be more reliable is the obviously larger heat sinks, more breathing room and the possibility this system is simply made better than retail systems as almost all debugs are made in Japan and all tools in Japan meaning better components overall are probably used and made by SONY in-house (Instead of mass production in China) as developers can't afford downtime and hindering productivity which could affect a dev companies bottom line if enough died (Or went partially working) quickly enough. The only thing that i could be the DECR-1000A's achilles heel in it's reliability is it's 90nm CELL and RSX (More heat) and based early PS3 components (Sony fixing YLOD problems in newer PS3 models) In fact, Is it possible all SONY dev hardware (Debug/TEST, Tools Made by SONY in Japan) PS2 onwards is made better than their retail counterparts (Direct comparison only applicable for TEST/Debug units of course as they're based largely on retail hardware) (Foxconn and others, China) and as a consequence more reliable meaning better longevity for those systems over a long period of time or are some equivalent reliability (Debugs) and only TOOLs made better or whatever happens the be the combination, if you have any knowledge if development hardware is made differently, let me know what you guys think. TL;DR Is SONY Playstation development hardware made better/more reliable than their retail counterparts?
You're correct. For the most part they do not have as many issues like the retail side (also DECR has 2 fans) Now that doesn't mean they're perfect as like you mentioned they use the larger RSX versions which, have proven themselves multiple times, to have issues if not treated well. As got why they last longer - my assumption is higher quality assurance and higher quality solder for the solder joints (seeing as this usually isn't an issue except for the RSX in a DECR if you're unlucky) since this is meant to go through slightly abnormal use during development stages. Ever notice why so many DECR usb ports are damaged? It was just part of the process of things they had to do while DECH systems had no such problem (for the most part, same for DECR1400)
In fact, the DECR1400 is the only ps3 unit that's purposefully designed to mitigate RSX issues, not even the 4K model was entirely RSX issue free.
For a moment I thought you meant a PS3 capable of running 4k, and I was like "wtf, when did that ever come out and why didn't I hear about it", LOL!
Huh, who would've thought, so you're implying a 1400A is a better daily driver for (single player) PS3 games (For RSX failure rates) than any other PS3 be it a 1000A or any CEX/DEX, and would last longer before failing, is it because of the material used in the solder joints or faster fan speeds or something else?
I believe it's the same issue that nvidia had with thermal shock on laptop gpu's around the same time. When things get hot they expand & when they get cold they contract. The problem comes when you have mix materials with different rates of expansion and contraction. It was common for laptops to leave fans turned off as long as possible before cooling down, which caused extreme changes. The problem was made worse by the switch over to lead free solder. It's unlikely that the solder was any better, the manufacturing may have been better but it's more likely that the cooling was better & they weren't trying hard to minimise fan noise.
They used better solder materials and an improved overall cooling design. The DECR-1000 is still supposed to be fairly robust as they implemented a huge heatsync and redundant cooling system to mitigate the unreliable early RSX chips. Unfortunately, quite a lot of these still failed and had to be refurbished, as such there are a few refurbished units in the wild which had the whole motherboard replaced with newer RSX revision chip (with rom b07) which incidentally do not support firmware earlier than 1.60 (even though the unit still allows to downgrade it lower! Leading to further bricks), in fact the RSX has been updated fairly often during the DECR-1000 lifespan, the release day RSX sported rom b01 (which as far as I know, was the most prone to failures), it eventually (in 2007) was replaced by a RSX chip sporting rom b03 which remained the de facto version for non-refurbished units until January 2009 (where units only supporting 2.30+ showed up, even though I myself never got my hands on such a unit as they were replaced by the DECR-1400 a few months later). Obviously if your DECR-1000 is well maintained and used in environment that has proper ventilation, failure is less likely to occur, sadly few development houses had such concerns back then.
Well, i have 2 fully functioning units DECR-1000As with no issues (Bulk deal) and one 1400A also fully functioning with no issues. So for now, I'm quite content, and I've heard the TOOLS use retail RSX chips so if you reball a respective model RSX revision from another PS3 system should a unit fail (Though unlikely as the system is very well ventilated) i should be able to repair it using said method. But i will keep in mind my DECR-1000A units may be newer revisions that do not support older firmware's, is it just a motherboard model outlining revisions?
That's technically incorrect, the RSX chips used on DECR-1000A units are not the same as their retail counterparts (the Rom revision is older), you can however reball it with a retail RSX and it should still work.
I didn't know for sure if they were identical or not but it's still nice to know a retail RSX (90nm) should work if the need arises.
I can't speak specifically to the DECR1000A (though I do have a few) but I can tell you my experience bringing similar hardware up from layout to NFF, FF/DK and to FF/retail. Usually our first run after layout will be a NFF (non form-factor) board that is physically much larger than the final version and has a ton of test points brought out so we can debug the firmware as we are de-risking and finalizing the hardware choices. These we do in runs of 10-50 depending on what we are making and who the client is (internal vs. external, etc.). These are very much bespoke units that we build in house with PCBs that we get fabricated with some local shops that we have experience with, and parts from Digikey or whoever. This is alpha hardware and where probably the bulk of development work is done; usually about 60%-80% depending on what we are building and what the lead time on our FF (form-factor) hardware is. The NFF hardware is much easier to develop on since we bring out most of our GPIOs, power, etc to test points that we can get at easily with an oscilloscope or logic analyzer. We also have versions with different FPGAs, NVRAM, radios, whatever hanging off them to test different design options and characterize the power consumption profiles for wireless devices. The next step is usually the FF for the DK if we are making one. If we aren't (it depends on the product) we go straight to the FF for the retail units. In any case these get done in runs of between 500-1000 with a domestic CM that is both physically close to us and that we have a lot of past experience with. This is because we spend a lot of time on-site at the CM teaching them how to run our automated test firmware, flash production firmware on the devices, taking notes, etc. We also have the EEs, MEs and FW guys there to diagnose any problems that pop up. This can be anything from solder cavitation to incorrectly mapped GPIOs, radio failures, cracked IMUs, cracked housings, whatever. The take away is that very few marginal boards will make it past the CM's and our QA teams since we are very closely examining the production line to find any potential problems; this allows us to de-risk our large scale production later on. While this hardware is intended to go to "external customers", they are other developers so we don't have to be 100% on point from a firmware point of view, but it is very important that we pick up any potential hardware issues at this stage. We aren't super concerned with our yield rates at this point but we always want to be above 90-95%. This increases the closer we get to the first rev for large-scale production. Failed units are returned to us for diagnosis if we can't figure out what happened on the line at the CM. If we didn't do a dev kit, we will do a similar process as above for the FF/retail units at the same domestic CM for what usually ends up on eBay as "engineering samples". The only difference is that we try to get the firmware much closer to final since most of consumer products don't have an easy upgrade path. These also get used for UL and FCC certifications and other regulatory things that we need to get done. In any case, we take the lessons learned and possibly iterate on whatever issues have popped up, and assuming management gives us the green light we stand up a large-scale CM in China for the retail FF. This is typically somebody like Flex who fluctuates around 30k-60k employees depending on the season and can dedicate multiple manufacturing lines and personnel and give us tens of thousands of units (or more, we don't have that much market share though) per month. Once we hit this stage there really isn't any going back, we can rev the hardware in the future of course but now economies of scale take over. We are only in China for a few weeks to get them started with the QA and validation procedures for this specific piece of hardware. Needless to say, with that may units and that many people getting rotated through the factory on a day-to-day basis it is much easier for units to slip through the QA cracks at this stage. These employees make dollars per day and aren't super motivated... They don't care too much if a few units for [insert American company name] don't work, since it doesn't really affect them personally. Flex does have guaranteed yields however IIRC, so we take this in to account and determine and acceptable an yield target before we head over. Sony does a lot more volume than we do, and is probably vertically integrated to a degree but I imagine the process is the same. I would say the 1000A's probably aren't "built better" but the units that went out were looked at very closely by Sony since it was designed early on in the PS3's lifecycle and were run at a much lower volume. If I'm not mistaken the 1000A's had kernel debug capability which means Sony probably used them internally for the PS3 software DK development as well. Once they went to the 1400A's they were clearly very comfortable with the overall PS3 design and likely made those at the same CM as the retail units, which had a lower bar for QA and weren't used internally as much. It also worth noting that production quality varies from CM to CM. If you are standing up more than one you can see weird issues on the same design because one has a pick-and-place machine that is different and uses too much force, or a derpy wave-solder machine, or whatever. This is the same process for most consumer electronic type hardware that gets built at scale in my experience. Hopefully that helps-
Great info mathieulh and Trident6, thats why i like so much being here. From my point of view i also think that lead free solder has also evolved since its begining. Much of the infamous (ylod, rlod ) has now stabilized a bit more becuse new board revisions do have smaller cpus and gpus and they generate less heat so boards dont warp so soon with heat cicles. Still bad refrigeration is big problem. I dont think sony or xbox nintendo (last one has better quality control in my oppinion) realy want durable products... consumers are guinea pigs on release hardware. Thats what i do feel as an early day adopter of some stuff i bought... And if you buy them 2 instead of one... thats OK for them right?
