|
BurritoJustice posted:1.43v is totally fine, Intel 7 can handle high voltages. The published Intel maximum is 1.72v. This is completely insane to me, I trust that the chips can do it but realize that electromigration must not be a concern at all. I remember people talking about 1.45 being unwise to pass on Sandy Bridge, 32nm!! My own 2500K did start to need a little more voltage to hit the same clocks it always had after 6 or 7 years, what would running these 10nm or 5nm parts at 1.5V for 24/7 computing look like a couple of years from now?
|
#
?
Nov 3, 2022 01:40
|
|
|
|
| # ? May 18, 2024 02:51 |
|
|
Those 3 and 5 volt rails are finally going to have a purpose again.
|
|
#
?
Nov 3, 2022 09:01
|
|
|
forbidden dialectics posted:Pssh, back in the day I ran my Celeron 300A at 2.5V to break the 450MHz barrier. IIRC I got it to boot at 500Mhz
|
|
#
?
Nov 3, 2022 13:40
|
|
|
ijyt posted:Those 3 and 5 volt rails are finally going to have a purpose again. hahaha, legit made me laugh
|
|
#
?
Nov 4, 2022 00:21
|
|
|
Speaking of voltage, gonna try again in case some Intel understander would know what's up. A while ago I bought a Chuwi Minibook with the N5100, understanding that I get what I pay for, which is a 28Wh battery. It turned out that the battery life is even worse than I thought, since it doesn't seem to clock down or enter whatever state that would let it save power. Basically it idles with package power at 1.5W, which seems silly for a 6W TDP CPU. My old Core M3 tablet could go down to 0.4W for package power when idle, and that's a 6 year old 14nm part. It also seems to be too high for an N5100, Chris in his review of another Chuwi N5100 laptop shows that it does go down to 0.66W, which is almost a third of this.  https://www.youtube.com/watch?v=wdvnkD8f2U4&t=968s Here's the best I've managed to do (core clocks down to like 400 Mhz, but a) I can't reproduce it again and b) this didn't really change the total power much, obviously:  As far as I can tell these laptops use the exact same CPU and even the same motherboard: https://i.imgur.com/YvC7OcX.mp4 It's not just the issue with the package power measurement either, at the wall I can't get it below 5W with the screen off while my Core M tablet goes down to 2.8, same charger and everything. So I'm thinking this could be a configuration issue that they just didn't tune properly at the factory because  . The bios is wide open, I just have no idea what to look for. Any suggestions what I could check and tweak?
|
|
#
?
Nov 4, 2022 22:21
|
|
|
I have no idea about what I'm talking about, BUT I suggest pressing buttons in throttlestop https://www.techpowerup.com/download/techpowerup-throttlestop/ as it will be a faster fiddling loop than rebooting to BIOS.
|
|
#
?
Nov 4, 2022 23:20
|
|
|
karoshi posted:I have no idea about what I'm talking about, BUT I suggest pressing buttons in throttlestop https://www.techpowerup.com/download/techpowerup-throttlestop/ as it will be a faster fiddling loop than rebooting to BIOS. Anyway, had some time to run more experiments, this time with the Jetstream javascript benchmark.   Excel seems to start charts at 0 by default, that's a new thing, isn't it? Anyway, seems like you trade off energy for time pretty linearly until 2GHz, at which point you don't gain gain anything by going slower. So if you don't mind poo poo taking slower to load (just open in background tab), clocking down is an option. Pretty similar to Cinebench.
|
|
#
?
Nov 7, 2022 14:33
|
|
|
mobby_6kl posted:
That's an improvement as far as defaults go. It's not necessarily true that starting your Y-axis at zero will always be the best choice, but it will definitely be the correct choice more often than setting it arbitrarily just below the lowest y-value. The x-axis is a little more flexible, but similar. If you dropped the Y-axis just below 140 it would make the relative differences in your y-values look way bigger than they actually are.
|
|
#
?
Nov 7, 2022 15:37
|
|
|
https://twitter.com/BIOSTAR_Global/status/1589596025349554177 a ton of serial headers and a ton of PCI(non-e) slots, now thats what peak performance looks like
|
|
#
?
Nov 7, 2022 16:58
|
|
|
Cygni posted:https://twitter.com/BIOSTAR_Global/status/1589596025349554177 It's not like it needs to be high-volume to keep the cost down; a company that needs one of those is gonna pay what it costs.
|
|
#
?
Nov 7, 2022 17:02
|
|
|
Love the PCB color on that
|
|
#
?
Nov 7, 2022 17:03
|
|
|
Cygni posted:https://twitter.com/BIOSTAR_Global/status/1589596025349554177 If you’ve got a factory floor that needs legacy PCI controller cards and serial connectors, this very much would be peak performance. I’d be interested to know how they implemented the legacy PCI bus here; from what I understood playing around with an industrial DFI SB600-C (itself a Sandy/Ivy board replete with PCI slots), everybody more or less stopped bothering with legacy PCI and simply used bridge chips for consumer market PCI slot backwards compatibility starting around Haswell on the Intel side and Bulldozer on AMD.
|
|
#
?
Nov 7, 2022 17:55
|
|
|
A bunch of companies make those kinds of legacy support boards and they rule. Here's one of MSI's with a bunch of serial headers. LGR did a great video using an ASUS one to build a new retro PC.
|
|
#
?
Nov 7, 2022 18:07
|
|
|
Hasturtium posted:If you’ve got a factory floor that needs legacy PCI controller cards and serial connectors, this very much would be peak performance. I’d be interested to know how they implemented the legacy PCI bus here; from what I understood playing around with an industrial DFI SB600-C (itself a Sandy/Ivy board replete with PCI slots), everybody more or less stopped bothering with legacy PCI and simply used bridge chips for consumer market PCI slot backwards compatibility starting around Haswell on the Intel side and Bulldozer on AMD. it has to be a bridge chip because Intel stopped putting native PCI into the chipset some time ago Hopefully the bridge chip isn't garbage though. Last time I used one, it had some read-around-write pcie bug and the only solution was to hard strap it to pcie gen1/x1 (i.e. please don't flood our bridge chip with much traffic)
|
|
#
?
Nov 7, 2022 18:12
|
|
|
just to be clear, i wasn't making fun of it. it rocks and i totally get why it exists. i briefly worked in a high end CNC shop and you better believe everything was RS-232. greenboards can also be a great tool for all sorts of various retro/niche projects.
|
|
#
?
Nov 7, 2022 18:48
|
|
|
WhyteRyce posted:it has to be a bridge chip because Intel stopped putting native PCI into the chipset some time ago I figured it had to be a bridge chip, yeah. It would be nice to think some enterprising firm built a reliable bridge that could handle the peak 666MB/second five vanilla PCI slots could drive. That’d amount to, what, PCIe 3.0 x1 with some generous headroom? Bigger question is whether it would accurately handle some PCI weirdnesses like port addressing or DMA behavior; there was some lamenting over on VOGONS that more recent PCIe bridge solutions weren’t allowing access to Yamaha OPL3 chips on elderly sound cards and the like. It makes me wonder about the reality of the economics of this sector and how much any of that comes into play for industrial purposes. Lord knows something like this would have been welcome at a lab where I used to work - virtualization and device passthrough for electron microscope controller boards that never got driver support past Windows 2000 was a concern all the way back in 2009… Hasturtium fucked around with this message at 19:30 on Nov 7, 2022 |
|
#
?
Nov 7, 2022 19:24
|
|
|
A company called NIXSYS sells brand new PCs with Windows 98 and ISA slots. Which aren't cheap but you're mostly paying for support for something that otherwise wouldn't have any. If a business or government agency needs something they will pay.
|
|
#
?
Nov 7, 2022 19:42
|
|
|
Aren't there also entire modern OSes built to maintain OS2 compatibility because it ended up being really good for specific use hardware?
|
|
#
?
Nov 7, 2022 21:32
|
|
|
no 1000w VRMs, and loving green PCBs? supreme failshit
|
|
#
?
Nov 7, 2022 21:51
|
|
|
Cygni posted:https://twitter.com/BIOSTAR_Global/status/1589596025349554177 Is....is it warm in here? 
|
|
#
?
Nov 7, 2022 21:52
|
|
|
Gonna keep these SCSI cards going forever in PCI slots. HP ScanJet 4c will never die.
|
|
#
?
Nov 8, 2022 08:16
|
|
|
Someone producing a new Skylake motherboard in 2022 is just 
|
|
#
?
Nov 8, 2022 09:17
|
|
|
Is green solder mask actually cheaper or is it just a market signal thing?
|
|
#
?
Nov 8, 2022 12:06
|
|
|
TheFluff posted:Is green solder mask actually cheaper or is it just a market signal thing? It makes the electrons go faster
|
|
#
?
Nov 8, 2022 16:52
|
|
|
TheFluff posted:Is green solder mask actually cheaper or is it just a market signal thing? It used to be the standard and other colors would have an upcharge, these days it's basically free to get other colors except at the lowest price points.
|
|
#
?
Nov 8, 2022 16:58
|
|
|
Kazinsal posted:Someone producing a new Skylake motherboard in 2022 is just Skylake to Coffee Lake are still very common for embedded and extended availability applications. Jokes aside, Intel has a lot of experience with and fab space for 14nm.
|
|
#
?
Nov 8, 2022 17:43
|
|
|
hobbesmaster posted:Skylake to Coffee Lake are still very common for embedded and extended availability applications. Jokes aside, Intel has a lot of experience with and fab space for 14nm. This is 100% correct. That’s not a board intended for high CPU performance in absolute terms anyway - looks like a 4+2 VRM setup. Cutting edge is a lot less important for these roles than “setup and configure once, then run without incident indefinitely.”
|
|
#
?
Nov 8, 2022 18:02
|
|
|
hobbesmaster posted:Skylake to Coffee Lake are still very common for embedded and extended availability applications. Jokes aside, Intel has a lot of experience with and fab space for 14nm. Yeah Skylake (really they are all just skylake) is the end of an era in a lot of ways and there is going to be a long long tail of embedded stuff etc, or just general desktop support (since it sold a billionty units because it was around for 5 generations). the other one that is popular in the embedded space is Apollo Lake/Gemini Lake... I actually have seen both a ton of mATX and mITX form-factor industrial motherboards (superIO focused with a couple serial headers, some have parallel ports, etc) from the likes of Asrock and Gigabyte, as well as fully integrated OEM systems (rugged/industrial PCs, switches, thin clients, etc). I think that is where embedded will go after this... after all a gracemont core performs about the same as a skylake in per-thread perf, so, if skylake is fine tremont or gracemont will be too. The Atlas Canyon NUCs (tremont quad-core for ~$230) finally are shipping in decent quantity btw, if anyone has been keeping an eye on that. This is, unfortunately, the gen before they added AVX support, and still no usb4/thunderbolt, but they have a gen11 media block and pretty solid CPU performance (bearing in mind it's a low-power mini PC). https://www.anandtech.com/show/17488/intel-atlas-canyon-nuc11atkpe-and-geekom-miniair-11-ucff-pcs-review-desktop-jasper-lake-impresses https://www.bhphotovideo.com/c/product/1720772-REG/intel_bnuc11atkpe0001_atlas_canyon_pentium_n6005.html https://www.connection.com/product/intel-intel-nuc-11-essential-kit-nuc/bnuc11atkpe0001/41376943
|
|
#
?
Nov 8, 2022 20:26
|
|
|
TheFluff posted:Is green solder mask actually cheaper or is it just a market signal thing? Flip that on its head. It's the absence of a signal. Market signalling is things ranging from non-green soldermask to RGBzzzz to tacky painted plastic shrouds which look like bad game art armor, all of which are important (to varying degrees) in capturing gamerbro money. If you're designing for a market which doesn't care about any of that noise, you just go with the default option, which is still green. Green just works well. It's nicely contrasty with white silkscreened legend while also being translucent enough to make it easy to see where surface layer traces are. These things matter when debugging and reworking prototypes; the more opaque colors are kinda unpleasant to work with. By the time you're ready for mass production, if there's no reason to change away from green, you don't.
|
|
#
?
Nov 8, 2022 23:23
|
|
|
Paul MaudDib posted:Yeah Skylake (really they are all just skylake) is the end of an era in a lot of ways and there is going to be a long long tail of embedded stuff etc, or just general desktop support (since it sold a billionty units because it was around for 5 generations). I’m curious to see how Elkhart Lake does, especially with the FuSa stuff. I guess some people got through with “real time is no longer possible on x86”
|
|
#
?
Nov 9, 2022 00:24
|
|
|
Holy poo poo I'm getting really lost in Intel codenames and branding now lol. Atlas Canyon using Tremont cores which is also Jasper Lake which are all Pentiums or Celerons, but so are Elkhart Lake just with lan and UART  Anyway wile I've been screwing around trying to get idle power consumption down, the performance on the N5100 is really nice for the price/power, pretty close to Ivy Bridge at almost 1/10th power. I can't imagine most industrial applications needing more than this. That you can get a fully functional, ready to use PC for like $120 is pretty amazing. https://www.aliexpress.com/item/1005001565518382.html
|
|
#
?
Nov 9, 2022 02:07
|
|
|
Tangent: Is there a reason why Intel for some godforsaken reason doesn't want to stick dual NICs on their NUCs? I would be all over that poo poo as a router device if they did.
|
|
#
?
Nov 9, 2022 06:08
|
|
|
mobby_6kl posted:Holy poo poo I'm getting really lost in Intel codenames and branding now lol. Atlas Canyon using Tremont cores which is also Jasper Lake which are all Pentiums or Celerons, but so are Elkhart Lake just with lan and UART atlas canyon and jasper lake (and skylake etc) are product families - like summit ridge or whitehaven or matisse. canyons are NUCs and lakes (post-skylake) are cpu packages. Jasper Lake is an atom cpu family that gets used in nucs and other chromebook type stuff, atlas canyon (like skull canyon, frost canyon, etc) is a NUC that is based on Jasper Lake. an architecture is a cove or a mont now (post-skylake) depending on the architecture series. this is like zen1/2/3/4 vs zen4c. summit ridge uses zen1 uarch, for example pentiums/celerons (ahem, intel processors now) are a product branding tier, like athlon/r3/r5/r7. intel does use codenames pretty freely, every chipset and cpu package and end-user product family (ssds, nucs, maker boards, etc) and all architectures, etc will all have their own codenames. This also exists for AMD (again, matisse, whitehaven, etc) but AMD is better about not using their codenames ("7900X" instead of "plum bonito", "hotpink bonefish", etc") in public and just referring to it by chip codenames (Navi 33), and a lot of their series naming is numeric (Zen1/2/3 and RDNA1/2/3) which is also helpful. But they do kinda the same thing with painter codenames and city names... which is better, milan or rome or genoa? vermeer or matisse? or the fish names too. with skylake and before, you had (eg) skylake-S desktop chips vs skylake-u laptop chips... those would be different lakes now I think. so I guess it basically aligns with die config or a family of die configs. Paul MaudDib fucked around with this message at 06:23 on Nov 9, 2022 |
|
#
?
Nov 9, 2022 06:17
|
|
|
SwissArmyDruid posted:Tangent: Is there a reason why Intel for some godforsaken reason doesn't want to stick dual NICs on their NUCs? I would be all over that poo poo as a router device if they did. I don't know if there's an actual reason but there's so many mini PCs with dual nics it's probably not a market segment they feel the need to get into.
|
|
#
?
Nov 9, 2022 08:56
|
|
|
SwissArmyDruid posted:Tangent: Is there a reason why Intel for some godforsaken reason doesn't want to stick dual NICs on their NUCs? I would be all over that poo poo as a router device if they did. They have 2.5G ports, right? You could do a router on a stick setup with VLANs and a trunk link going to the NUC? Just have the WAN port on it's own VLAN that goes only to the trunk.
|
|
#
?
Nov 9, 2022 14:29
|
|
|
Twerk from Home posted:They have 2.5G ports, right? You could do a router on a stick setup with VLANs and a trunk link going to the NUC? Just have the WAN port on it's own VLAN that goes only to the trunk. Are there any VLAN-capable switches with 2.5G ports that aren't silly expensive yet?
|
|
#
?
Nov 9, 2022 15:16
|
|
|
Also, ASRock makes an intel nuc with dual 2.5s: https://www.asrockind.com/en-gb/NUC%20BOX-1260P tweaktown liked it And there are usb-c adapters available to add 4 gigabit ports as well, like so .
|
|
#
?
Nov 9, 2022 15:25
|
|
|
SwissArmyDruid posted:Tangent: Is there a reason why Intel for some godforsaken reason doesn't want to stick dual NICs on their NUCs? I would be all over that poo poo as a router device if they did. If you're looking for an opnsense/pfsense/virtualization box, there are 4x lan boxes based on these socs all over AliExpress. Ex: https://m.aliexpress.us/item/3256803804266682.html?gatewayAdapt=glo2usa4itemAdapt&_randl_shipto=US Edit: fanless, too.
|
|
#
?
Nov 9, 2022 15:46
|
|
|
I've found the best keyword to search for on AE is "soft router", and possibly include the keyword "SIM" if you would like something with a backup LTE(A)-based modem (since 5G isn't notably faster because of a lack of airtime).
|
|
#
?
Nov 9, 2022 17:11
|
|
|
|
| # ? May 18, 2024 02:51 |
|
|
BobHoward posted:Flip that on its head. It's the absence of a signal. Market signalling is things ranging from non-green soldermask to RGBzzzz to tacky painted plastic shrouds which look like bad game art armor, all of which are important (to varying degrees) in capturing gamerbro money. If you're designing for a market which doesn't care about any of that noise, you just go with the default option, which is still green. Fun fact: my co we would usually do green or black solder masks and we thought hey we want to have another colour in the mix so we can tell what project it is just from sight, let's make this one black. Hooo boy. The fab house had all sorts of problems with a black solder mask, it was totally bizarre. If they're not familiar with it it's a big hassle. We went back to just green/black after that, nothing fancy. Also the usual thing at some companies outside of AMD is red boards are for early like rev A prototypes
|
|
#
?
Nov 9, 2022 19:12
|
|
