hojnikb

Just a heads up; this board can now unlock it's graphics CUs; a lot of users are reporting full unlocks from 24 to 40CU. Up to +60% perf. boost after unlock.

I doubt the whole house runs on that... And having a gridtie system makes zero sense in that case, as usually gridtie systems don't work when there's no grid. So you need to have your loads going thru the inverter, just like an UPS. Gridtie systems makes sense when you have solar or load balancing /peak shaving batteries.

Yes, BMSes are software defined, but inverters/chargers are not. If you want to use lower cost solar inverter, chances are they won't support NMC charging profile, so you'll either have a hacky or unusual configuration of cells, or you won't be able to properly charge them. As for the cost; LFP cells nowadays are extremly cheap. I've seen LFP cells going for as low as 30-40$ a pop for 280Ah variants, which puts them down to below 50$/kWh. This is so cheap, it makes absolutly no sense to mess around with used or smaller format of cells, unless you got them for free. But even in that case, you likely have mismatched cells, which could be an issue of safety and a nightmare for balancing. I myself have built a 15kWh pack of grade A cell, BMS and a custom box for less than 1200€. And that was almost 6 months ago, sourced locally, so direct chinese source would be even cheaper. It's really not cost effective to build anything below 15kW (or ~7,5kWh in case of a 24V system) as the cost of cells will be marginally lower, while capacity being much lower. If you actually plan on building a serious system and not something to just play around, this is th way to go. Especially if it's combined with solar, as then batteries will get cycled a lot, so it's crucial that they're matched, well balanced and new.

Energy volumetric or specific density isn't really a benefit with house batteries as you both have storage to place them and (usually) no weight limits. And 9/10 LFP will be cheaper per kWh if you look at bigger 200Ah+ cells (as you woulnd't want smaller cells anyway for house). Also this is not true for LFPs. No lithium chemistry (including LFPs) doesn't like to be charged or even worse, kept at 100%. While it's true, that LFPs will suffer a lot less degradation if cycled 0-100%, they still will degrade more compared to being cycled between 20-80%. Just like with NMC, if they're kept at around 50% and cycled +-20%, they will last MANY MANY cycles. But in practise, additional cycle life isn't that important in a correctly sized system, as calander aging will be a bigger factor. The other reason, why a lot of manufacturers recommend charging LFPs to 100% (cars or storage batteries) is because that way, BMS is always correctly calibrated. LFPs are notoriously difficult to estimate charge level and without frequent 0-100% charges, you can have tons of BMS sway and as such incorrect SOC readings. Not so important with home batteries, but very important for EVs.

That's true. But odds of that happening with LFPS is much much lower. You can overcharge, pierce and smash LFPs and they won't easily explode or catch on fire. They're pretty tough and used in such applications for good reason. But of course, nothing beats simplicity and safety of good ol' lead acid.

Guys, if you plan on doing this at your house, don't use used NCM batteries. While they are great at capacity density and work great in an EV, they're not used in home storage for a reason. Their cycle life is much lower (LFPs can go for several 1000 cycles, some even claim 10000) and they're much less safe, if anything goes wrong. And there's also another thing; most cheaper inverters are designed with LFPs in mind and don't properly charge NMCs, as they require CC-CV charging. LFPs are more akin to lead acid. Another benefit of LFPs is also their lower sensitivity to frequent 100% charging. Because of their chemistry setup, they better handle high states of charge. So daily 0-100% cycling isn't a big deal at all, but will kill NMCs in only a couple of years. So the best way to go about this is to buy bare 280-320Ah cells and group them in a pack. If you're really cheap, you can do away with a simple plastic case, that would fit these cells or buy a premade case, designed especially for these cells. The latter option also looks more professional. Also there's BMS/balanacer. Most are again designed around LFPs. A good BMS will also have a passive (or even active) balancer so there nothing else you need. Cheap option here are JK-BMS or DALY. Both proven options with tons of features. For a 15kWh pack you can build everything for under 1500$ if you can source cells on cheap. Most of the times, alibaba is the way to go. So for under 100$/kWh you have a new battery, that you built yourself that will likely last at least 10-15 years if treated right. Another good option are used server rack batteries, but their pricing can vary and are not always a good price/performance option. And of course you'll need an inverter. 9/10 times a solar inverter (even if you don't use solar) is the best and most cost efficient option. And they can be as low as 200-300$ for cheaper lower power option. Inverter will act as a battery charger, UPS and of course solar MPPT.

2699v4 seems a pretty bad value right now, because CPU alone is pretty expensive and no kits are based around it. Might be a better idead to pick up one of those dual socket boards with 2680v4 combo cpus (for a total of 28 cores) but that's a pretty specific workload to use all of these cores and such setup comes with it's own challenges. But you're correct in the sense, that these are not comparable, given that aliexpress option is half used (cpus at least). But if you're chasing a budget, used should always be considered an option, as Linus himself hinted at. But all things said and done; aliexpress choice of used CPUs and motherboards is pretty amazing.

You guys should also check out these cheap Xeon 2011-3 kits from aliexpress. You can get amazing value for money for the mobo+cpu+ram combo. Some kits like this one can be had for as little as 75$. If you want something newer, faster and a little bit nicer, you can also pick one of the 14 core broadwell kits for around 100$. I think if you look past used cpus and non brand name motherboards, you can get an amazing value for money system (especially if gaming is not the primary focus) for quite literally dirt cheap. For 500$ you could easily fit a 3060/3060Ti, a decent PSU and a case in this system and you'd get a pretty good gaming setup, albeit with some cpu bottleneck, where single core performance matters more. In multithreaded applications, this 14 core cpu is very close to 8 core ryzen 5th gens (i measured ~13000 points in cinebench r23 with dual channel), which isn't so bad for 100$ all in.

I'm surprised there isn't a big outcry from the sex work community against Linus for "devaluing" their work, like it happend when bella thorne joined.

Don't forget, that you cannot combine current capability of all rails, as there are combine limits as well.

Nope, Sandisk had mass produced QLC flash for higher capacity microsd cards since 2009. So economy of scale was there. TLC and QLC was not viable for SSDs back then because controller technology was not there yet.

QLC is nothing new. Sandisk had QLC based microsd cards 12 years ago. It' really quite simple. QLC flash, because it is by design worse than TLC, needs more spare blocks on die and stronger ECC to combat more errors. Couple that with more complex flash controllers and firmware, it's obvious why ssds based around it are not cheaper.

Did you try disabling truetone? Mine was also damn awful with truetone enabled (way way too warm for my taste). Disabled that and it's fine.

Just to update this topic; It turns out, CPU was not seated properly to the socket. After trying _everything_ and removing cpu a couple of times, it magically started working.

Even if you partition the card at 16GB or thereabouts, there's no guarantee that controller won't just "write" data to the remaining 240GB of empty space. Throw it in the trash and make a dispute on ebay or where ever did you buy it.