yeah and amd are debating whether to have the 3000 series work on current boards. why? because after you put 2 generations of cpus worth of bios info on a bios chip theres not much room left. so then you have to either have mulitple bioses per board or you have to just break it off after 2 gen and start new. which is what intel does. ive heard the mobo makers are actually getting a bit pissy about it and might put their foot down. intel gives you 2 generations interchanging. amd has given 2 so far and maybe 3 but maybe not. there wont 4 for sure. wow big difference. its almost like intel had good reasons for some of the things it does. wow what a concept
@@bobhumplick4213 Not that big of an issue, if you view the entire microcode for CPUs that Intel have released, it all fits into a package of 1.3Mb of size. This goes from old socket 775 and up until todays CPUs, and it is easy cramming 1.3Mb onto a BIOS flash chip if you want to.
@Max Pain who ain't greedy? How much money do you leave on the table at the end of the day? Money that you know you could have walked away with scot-free.
@Max Pain Gamer's chipsets, they don't make big profits on them, needing to pay the R&D people too! shame on nerdy noobs, not understanding , less and less need PC's now, the smaller the market, the more money they need for to produce them! Need Pro parts, for gamer's only.... The cloud is NOT needing this, that's were they sell!
@Knut Gerber really you're better off getting a new motherboard anyways. While the performance improvements of a CPU are important additional features usually don't hurt either.
Thank you for bringing some science into this! I've been using a Biosmodded z170 for 8400 and now 8700k (OC'd to 5GHz) since March and July 2018 with no Issues whatsoever. I've done a few biosmodding videos, now there is actually an automated tool for it, which does SKU Swap, VBios, GOP and Microcode all from a menu. As for the VRM upgrades from Z170->Z370, there is not much there. I can see the end of 3 phase boards, but that's about it, some of them are actually worse. For example, the ASRock Taichi, even the Ultimate on Z390 is still worse in the VRM department than the Z170 OC Formula ATX which has a doubled 6 phase (totaling 12). I will reference your video in my community, it's really good, thanks again for bringing knowledge to us! Cheers from Brazil
The link Jing Kang posted is a worse tool. CoffeeTime is a better tool and does not fail. You can see walkthrough and all links updated in my channel. Link to video = th-cam.com/video/kK503X1Wh1U/w-d-xo.html
The Asus TUF series likely has better VRMs right? I'm curious because I want to upgrade my 6700k but dropping such a nice mainboard seems insane to me just to get to the 8700k or 9900k... at that point, I may as well go AMD.
Well there may be some specifications regarding the pin involved but the effect of 30% higher Amps/pin seems to pale into insignificance when nearly 100% more A/pin causes no effect on stress test stability. (I assume if Roman had to increase voltage to remain Prime95 stable he'd have mentioned that in the video.)
i highly doubt the main motivation is thermal. They probably wanted more pins to reduce overall impedance (each pin is an inductor and a resistor, and adding more in parallel reduces the overall impedance between the motherboard's decoupling caps and the massively unstable load that is the CPU). Remember, intel has to care about things like EMI that are totally off your radar as an extreme overclocker.
This was the exact same thought I had. Number of pins probably related to impedence and frequency response of the PDN. Also a cheaper motherboard might not perform as well as the one Der8auer used. //Embedded electronics engineering student
Find the PIII discord group - there's a lot of mobo bios patches already done. I run 9900K on a Z170 Clevo laptop, completely stable and safe up to thermal limit of 160W so any decent Z170/Z270 desktop board would be fine.
This was interesting. One thing you haven''t really taken into account in the first test was the wires heatsinking the pin. It's possible the thick wires, and the fact it was soldered would help wick away heat from it. Soldering also creates better contact. I wonder if the results would have varied a fair bit if only the board side was soldered, and the pin side was touching a piece of perforated stripboard, to simulate a cpu substrate. I bet you wouldn't be able to fire 5A through it like that the same. probably good for 2A, and I bet the design safe amps would be 1A for each pin. It'll also be for redundancy, the number of pins, so that a few bad contacts won't matter so much.
der8auer...for me the volume was a little quiet and I had to turn you up and then music on youtube would blast into my ears after this video XD. Just letting you know so you can turn it up for next time. Happy New Year!
It's important to remind that soldering the PIN you reduced the resistance since the PIN is connected to the CPU by Surface contact. But the rest of the test is pretty conclusive.
Thank you. Excellent video. It would be nice and interesting to hear from a CPU engineer to explain why such care is being taken by intel to reduce per-pin current. Maybe the contact point to the CPU itself is so small that it can mean a big difference to the result.
Pins can handle current. Good to know. But more pins means less resistance, therefore less V droop. I you have the pin still soldered to the cables. You can solder another two wires to the ends of pin to measure V droop on the pin. Should be interestiong to see. I know the power supply shows 0,2V but that is not accurate and it measures total V droop acros all the wires.
I'd guess the additional pins would be primarily to reduce impedance and resulting EMR, etc. Could also be futureproofing the platform for future chips that require more power.
The little round circles imprinted in the plastic of the socket are marks from holes in the injection mold that this part is made in. After the plastic is injected in the mold and has cooled sufficiently enough to hold shape, metal rods poke through holes in the mold that push against the part and pops them out of the mold so that the next part can be made. It's the holes in the mold that exist for the push pins(aforementioned metal rods) that actually leaves the circular imprint as the injected plastic expands into the edges of them(as it takes every single feature of the mold of course). If you pay attention, you'll notice those little circles on E-V-E-R-Y-T-H-I-N-G, as almost everything in our lives contains at least some part or component that has been mass produced via plastic injection molding
@@Razzaq585 Yup, I've had my fair share of it. Although most of my experience comes from operating twin screw extruders, I've had a little time around some other equipment too.
That Chinese and Japanese on your shirt is super strange. The Chinese reads "eye, entertainment festival, today". The Japanese reads "Please do extreme drying". Is this some meme? And how did you get the pin layout diagram? Is it public available from Intel?
SuperDry is all about weird phrase in different languages. The Chinese character can be rearranged to “今日娛樂節目” (in Traditional Chinese) which translates into “Today’s Entertainment Show”
You can't because the voltage regulator is inside the cpu on that socketso that can't work.. but intel could totally make a LGA1155 Z77 DDR3 version of the 9900K if they wanted to, skylake memory controller can do ddr3.
The comments below miss how much more is involved here. So if the pin gets up to 60C people say, oh No Big Deal, its a scam. But those pins are then transferred to much smaller wires to the circuit, and then orders of magnitude smaller as it gets to the IC. 1A is actually an incredible amount for those wires. Plus you have the heat from the IC itself- Der8auer should know this well. You dont want additional heat just from getting the power to the chip adding to the heat of the chip itself. It takes away headroom. Its really the transfer from the package to the IC that I would be worried about. The size of those wires is so tiny. Take a look at whats under those if youve never seen it. Normally you dont because of the flip chip method and the epoxy that covers it.
Yes I was thinking about the same thing. More pins is not for the current is fo the resistance. Resistance means voltage drop. Traces inside pcbs are relatively smaller than the pins. And inside the die of the CPU this means even more difficulties. Having more power pin also mean have more power distribution on the die. Pins are also used to remove hear from CPUs. I don't think that Intel made this change only for those reasons. If they really wanted to screw things up they simply could swap two data pin like they did on 771 and 775.
TuttoCrafting he der8auer is a smart dude but he’s testing totally the wrong thing here. I’m sure spambot is totally correct. I’m sure my info is a decade out of date. But regardless that pin is 100x larger than the trace it connects to. Which means 100x the resistance per unit of measurement. It’s the cross sectional area that matters. It’s somewhat irrelevant how much power a pin can take. It’s not the bottle neck. I guess you’d have to crack open a chip and see how big those traces are and see then. I’m sure there’s probably a ground plane and a Vcc plane to try and minimize droop and drift. But not everywhere. To me it looks like the beefed up to Vcc and Vdd to accommodate more power and generate less (useless) heat.
Hear what you're saying- but remember putting a coffee lake CPU into an older socket would still only be about 15% extra current draw per pin. Do you really think Intel is already running so close to the such a red line? Especially considering the LN2 overclockers are pulling 300% extra current sometimes and no traces melt. The CPU/mobo either dies from too high a voltage or condensation/water causing a short.
Come one, Roman! The video is awesome and I thank you for it, but the question "will it melt" has a simple answer: Has anyone melted the socket of a Z270 board while overclocking an i7 7700K to over 200W? Well, no!
Interesting video, glad you made it! Happy 2019 to you, hope to see a lot more of your stuff in the coming year. =) Enjoy Vegas, and make sure to give Steve a bro-hug on camera. He needs one.
is there perhaps an importance for extreme uses? does the extra surface area for power from these extra Pins make some difference for Chiller or LN2 uses? they surely didn't build the Socket this way for extreme Overclockers, but even so.
sometimes a ground pin is only connected externally and not to any of the other ground pins of the device - so missing a power or ground connection can cause device functions to fail depending on the topography of the device
Awesome video man. Very interesting, I wonder what Intel's response to this would be. BTW, love the music but damn it's such and intense beat for the video. Seems more suited to a wing-suit video, or someone going to space lol. Need more content like this!
Could be a problem in the maximum amount of current can be transfered to the die, I don't know if still use thin gold wires or tracks but may be there are too thin.
When you taped the pins, did you check if it didn't get punctured after testing? Images shoes that they do quite a bit of pressure when the socket locked down. I see some small white point.
Considering LGA775 & 771 in the past and how they actually are cross compatible, this doesn't surprise me. Even less so since a group in China has also managed to make BGA CPU socket mountable (only functional with certain boards & BIOS however). I've had an LGA775 Xeon rig running for several years now, with any issues being caused by a bad PSU (replacing it fixed the rigs instability).
This is not the same thing. This is power delivery. Back then they just flipped 90deg the CPU layout to make Xeon CPUs not working on standard mb. If they wanted to do the same thy could simply have changed a couple of pins and not a lot. Changing pins is not cheap on CPU design. (That why back in time they only swapped two) I think that in this case they have seen some problems with power distribution. This is a single test on a single cpu. This does not address all the cases. This may be a lucky CPU.
You have to take the fact that your thick copper wires are very good thermal conductors into consideration. Also the resistance between the pin and the pad will certainly be different than your soldered connection. Please make a new test setup where you use a dead cpu and a cut in half socket on a pcb to properly test the thermal scenario. I would also like to see a proper resistance measurement of one of those pins using a Kelvin-clamp resistance meter.
he could run the current through 2 ground pins, so that they'll be connected by the interposer's ground plane. if he wants to measure voltage drop he could have separate probes on each of the lands and subtract the internal voltage drop from the total voltage across the pins themselves. would get a bit cluttered with 6 wires total but would be a proper test
I think a pin would heat up more if embedded in the plastic, rather than open to the air. And you say the metal can handle high temperature, but what about the plastic around it?
Please correct me, i believe the CPU socket you refer to as "plastic" which i doubt is actually made from some formulation of fiberglass & epoxy thus would make it much more resistant to high temp than ordinary plastic. Observe that as you cut with the Dremel it cuts like PC board material.
Perhaps only one or two phases of the vcore vrm were still connected to the cpu? It'd be interesting to see the temps of individual vrm phases as you tape off pins, one would expect some phases would no longer be connected to the cpu and thus the other phases pick up the slack and get hotter.
So all phases are simultaneously connected to all the relevant pins in the socket? There's maybe 8 vcore phases but 18 or more pins transferring that power to the cpu. So there's dozens of traces from each phase of the vrm to each of the cpu pins?
@@ericwright8592 there's not "dozens of traces"; there's a large copper rectangle underneath the output of the vcore, like a PCB version of a bus bar. it's then usually stitched with lots of vias (copper plated holes to interconnect layers) to another large copper rectangle in one of the internal layers (PCBs tend to have 8+ different copper layers, each with their own design). Then each vcore pin on the cpu is connected to this internal plane by one or more vias each. so you can think of it as a big sheet of copper that all the vcore phases connect to and all the pins connect to.
@@spambot7110 very interesting, I had no idea. Kind of want to tear apart an old motherboard now just to see. Or perhaps someone like Der8auer could do it for us?
@@ericwright8592 it would be pretty tricky to get those layers apart; they're very thin and laminated together with very strong glues. it would be very cool though!
Awesome! Thanks for yet another interesting innovative experiment! If the 9900K in my Z170 Clevo laptop is reduced to smoking rubble I can rest assured it wasn't because of the missing vcore pins :)
as in F.lux? i hate losing Color Accuracy, personally. instead, turn the Brightness down from eye searing to a more normal level, don't cram your face into the Display (but sit a foot or two back - or ballpark 0.6m if you prefer), always have ample ambient lighting in the room, Et Cetera.
I would have thought it would be to do with stability for certain parts of the core/cores. Due to the volt drop under higher currents. More pins = less volt drop = hight stability, especially if your going to do LN2
Moin 8auer, ich habe eine eventuell ungewöhnliche Frage - ich habe im Moment ein Asus Z170i (Mini ITX) Mainboard. Ich habe mir allerdings letztens ein neues Case gekauft (NZXT Source 340 Elite) weil ich ein Case wollte welches eine Glasseite hat damit ich anfangen kann mit modden. Allerdings wusste ich zu dem Zeitpunkt noch nicht das ich ein Mini ITX board besitze (ich hab mir meine PC's immer zusammenstellen lassen, ich habe sie nie selber gemacht). Nun brauche ich also ein neues Mainboard mit einem z170 Chipsatz und ich sehe in so vielen Videos bei dir den Karton des Maximus Vlll Hero - das einzige Board, bei dem all meine jetztigen Komponenten passen würden - wäre es irgendwie möglich das ich dir dieses Board abkaufen kann? :P Liebe Grüsse Sharleg
One difference between his test and a motherboard with an actual CPU is that the pin isn't going to be soldered onto the CPU. The non-soldered contact between the pin and pad is going to have higher resistance than soldering it to a wire, which should cause it to heat up more.
Remember the soldering part was only 1 of 2 experiments performed in the video -- he covered 68 pins with elec tape and ran his CPU at a much higher wattage than a 7700k or 6700k would be capable of pulling under full load. After 24hrs there was still no degradation under the microscope (even in his extreme example).
Maybe GPU display pins and/or motherboard manufacturer specified trade secerts for what those pins are for but most likely SATA controller maxed out, USB 3.0 controller connector, DMI for full speed, direct connect of M.2, and maybe a few other things that you cannot find by just covering pins on the CPU without other stuff also plugged in.
uh, they're vcore and ground pins. they're documented in the datasheet, and if he wanted to verify he could probe them with a multimeter in continuity mode; he'll get a little beep for all of them because they're all internally connected together.
Wow thank you for the info.. I just have to ask was this fun for you? Because I don't mind watching but to run these test would be way too much for me. Thank You again for your time and effort
Unless i wasn't paying attention : 1.35v x 1.16a = 1.5525w per pin oced your bench psu test 0.23v x 5.098a = 1.17254w so you missed the current draw mark by 25% but more than likely that pin could go to 5w fairly easy.
And in a surprise to no one, Intel "excuse" for Z370 was actually nothing. Still, more than 60 pins taped, damn that is overengineered AF. Also, thanks for the English content. =)
AM4 does fine with Ryzen vs. Ryzen 2. We will see how it does with Ryzen 3 and if any problems arise with power delivery to the chip (I don't think they will). They said it was for the next 10 years so Intel could definitely do the same. Pins have never been a valid excuse. Happy New Year!
I'm thinking it should be fine. AM3+ was fine on 500+w from FX series, so I doubt it will be a big deal (I can't even hit that on my TR 1950x, but I haven't used LN2 on it yet). AM3+ was "plus" sized over AM3 with fat pins for more current, and now AM4 looks like spider webs by comparison (AM3+ was pretty safe to just set the CPU on its pins, even small drops wouldn't bend it), however, I think AM4 will be just about as capable since it has so many more pins. Thinner, but many more. Also the power draw is not through the roof like FX, so that's the biggest thing.
How about the voltage on each pin? The heat from wattage is the real measure for this test. Unless you KNOW there is only 0.23v like you tested through each pin.
I would be much more likely to buy a new CPU if it "just worked" on my Z170 board. Buying both board + cpu again so soon after upgrading is just not a good return on investment.
Hello, I have to somewhat disagree that z270 and z170 could be usable for 9900K, even z370 is NOT. I own a Gigabyte Z370 AORUS Gaming 7, after updating it to latest BIOS I got 9900K, I did BIOS reset and LOAD BIOS Defaults. The Board is on all default settings. with my 9900K it ALWAYS boots with 1.34-1.36 vcore and under load in windows cross 1.4 I assumed that its fine for this CPU but then googled Apparently on z390 boards at default setting the 9900K boots at 1.2v I googled some more and found otehr people having same issues as I do [BTW seating Manually vcore to 1.2v wont boot at all] So I got Z390 AORUS Master P.S. NONE OF OTHER below 9900K CPUs have such issue on z370,the i7 and below all work FINE, only i9 have this over voltage
I have an i7 6700K that I purchased 2 years ago, thinking I would build a computer, but I never did. Should I build a new computer with it now ? Or should I just accept the $ 300 loss, and start over with a new CPU on a Z390 ?
As people had expected right at Z370 launch. Intel purposely made "8th/9th Gen" CPUs incompatible, not because there was an electrical need for it, but because it both makes more money to sell people another Motherboard, and saves money in support not having to deal with people buying boards without appropriate BIOS Versions for newer CPUs, as AMD had to deal with with Zen+ CPUs on 300 Series AM4 Boards.
11:00 I see a flaw in your test method. You are using a CC/CV power supply and so at 5 amps your power supply is delivering only .23 volts and that is only 1.15 watts which is not the correct voltage or wattage that would be flowing through the pin. Even at the lower amperage setting your power supply is limiting the voltage to get a specific current. This is not how the CPU and VRM will be working and so this would not be a valid test of the pin. However, with all that information we can calculate the resistance of the pin which is between 0.04222 and 0.04512 ohms at the temperatures measured but that does NOT include the contact resistance between the pin and the pad on the CPU interposer which will significantly contribute to its power handling capabilities. And the contact resistances will not always be constant, they can change with vibrations. Also, the contact resistance will change OVER TIME especially where there is heat cycling and this is the same no matter what type of mechanical contacts are used. The pins will begin to deform over time because heat cycling has an annealing effect on the metal pins. So the contact pressure will lower over its lifetime causing more contact resistance to build over time. Meaning more pins are necessary for the power delivery to be stable over its total predicted lifetime and expected operating environment. That doesn't mean that Intel was right in changing the socket pinout to add more power pins and I'm not defending it, it just means your initial test method was flawed. This annealing effect on contacts is the main reason why taking a computer apart and putting it back together can bring an unstable computer back to life. It doesn't reset the pins back to their original glory but it does re-seat the contacts allowing it an extended life. Bending the contacts up slightly also helps, and the same goes for PSU connectors and connectors between the boards on hard drives and the motors. Over time they will begin to settle into position and more resistance builds between the contacts. It isn't always the circuitry in the PSU that makes it unstable, the female Molex pins on PSU connectors can become loose over time and cause stability problems especially if the PSU uses one of the many voltage lines on the mother board connector for feedback. So if your computer shuts off unexpectedly and you suspect power supply, try getting those female Molex pins in the connectors to hold on tighter. The pin connectors on hard drive boards can also become loose over time and cause the motor controller to error out which means the hard drive continuously tries to spin up but can't. And again, removing the board and bending the contacts so they make more contact pressure can bring the hard drive back to life but eliminating the connector altogether by hard wiring the motor to the board can make it last until the heads decide to crash. Even on modern PCs and laptops connectors are inevitable but they are also inevitably a problem given enough time. If you could eliminate mechanical connections, exposed metal traces/contacts and electrolytic capacitor aging then you can theoretically create a system that could last a lifetime without physical maintenance other than dusting...
It's more to give headroom for reliability issues, I'm guessing. Not for physical reliability wrt mounting and dismounting, but from actual long term wear and tear from the electrical current flowing through either the pins, solder balls between the package and die, or some part of the interconnect structure. The extra pins can also be an effort to improve the stability of the power grid on the chip. Not everything is a ploy to extract more money out of customers. Chip designers usually know better than overclockers when it comes to what the design actually needs to function like it's promised for a long as it's warrantied
if you test latest ryzen pin from old fashion outdated tech off the cpu from substrate ...i bet it couldn't do 5 amps at all...it goes burn't and overheated on amd pins...so 3 amp is max limit for amd ryzen per pin while intel can push it further in temps and amps per lga ball pin ...to 20 amp max
hey guys, i wanna buy a ryzen 2600 and 2x8 gb ddr4 ram. can you guys recommend a cheap mainboard? dont need to overclock, also i dont want to need to update the bios. :)
I appreciate that you did in English also!
TheRealSamPreece videos with marry are in de only, scientific ones in english too
The sound volume is kinda low.
I second that as well. appreciate you doing it in English.
yeah and amd are debating whether to have the 3000 series work on current boards. why? because after you put 2 generations of cpus worth of bios info on a bios chip theres not much room left. so then you have to either have mulitple bioses per board or you have to just break it off after 2 gen and start new. which is what intel does. ive heard the mobo makers are actually getting a bit pissy about it and might put their foot down. intel gives you 2 generations interchanging. amd has given 2 so far and maybe 3 but maybe not. there wont 4 for sure. wow big difference. its almost like intel had good reasons for some of the things it does. wow what a concept
@@bobhumplick4213 Not that big of an issue, if you view the entire microcode for CPUs that Intel have released, it all fits into a package of 1.3Mb of size.
This goes from old socket 775 and up until todays CPUs, and it is easy cramming 1.3Mb onto a BIOS flash chip if you want to.
shame intel, shame
@Max Pain who ain't greedy? How much money do you leave on the table at the end of the day? Money that you know you could have walked away with scot-free.
Didn't think I'd see a Derrin Hinch slogan here ,but there you go..
@Max Pain Gamer's chipsets, they don't make big profits on them, needing to pay the R&D people too!
shame on nerdy noobs, not understanding , less and less need PC's now, the smaller the market, the more money they need for to produce them! Need Pro parts, for gamer's only.... The cloud is NOT needing this, that's were they sell!
@Knut Gerber really you're better off getting a new motherboard anyways. While the performance improvements of a CPU are important additional features usually don't hurt either.
Shame Intel, or shame Intel and motherboard manufacturers. Is this tacit collusion between motherboard manufacturers and Intel?
Thank you for bringing some science into this! I've been using a Biosmodded z170 for 8400 and now 8700k (OC'd to 5GHz) since March and July 2018 with no Issues whatsoever. I've done a few biosmodding videos, now there is actually an automated tool for it, which does SKU Swap, VBios, GOP and Microcode all from a menu. As for the VRM upgrades from Z170->Z370, there is not much there. I can see the end of 3 phase boards, but that's about it, some of them are actually worse. For example, the ASRock Taichi, even the Ultimate on Z390 is still worse in the VRM department than the Z170 OC Formula ATX which has a doubled 6 phase (totaling 12). I will reference your video in my community, it's really good, thanks again for bringing knowledge to us! Cheers from Brazil
Where can you find this?
www.win-raid.com/t3987f16-TOOL-Easy-automated-Mod-tool-for-Coffee-Lake-bios.html
The link Jing Kang posted is a worse tool. CoffeeTime is a better tool and does not fail. You can see walkthrough and all links updated in my channel. Link to video = th-cam.com/video/kK503X1Wh1U/w-d-xo.html
Casual? Yeah right.
The Asus TUF series likely has better VRMs right? I'm curious because I want to upgrade my 6700k but dropping such a nice mainboard seems insane to me just to get to the 8700k or 9900k... at that point, I may as well go AMD.
extra pins totally debunked
Well there may be some specifications regarding the pin involved but the effect of 30% higher Amps/pin seems to pale into insignificance when nearly 100% more A/pin causes no effect on stress test stability. (I assume if Roman had to increase voltage to remain Prime95 stable he'd have mentioned that in the video.)
Intel being Intel.
i highly doubt the main motivation is thermal.
They probably wanted more pins to reduce overall impedance (each pin is an inductor and a resistor, and adding more in parallel reduces the overall impedance between the motherboard's decoupling caps and the massively unstable load that is the CPU). Remember, intel has to care about things like EMI that are totally off your radar as an extreme overclocker.
Yes, this needs more likes
@@Malus1531 i guess i can add a like if u insist
This was the exact same thought I had. Number of pins probably related to impedence and frequency response of the PDN. Also a cheaper motherboard might not perform as well as the one Der8auer used. //Embedded electronics engineering student
Like!
Not to mention the transients, ripples. They work on mV precision
that's true but Intel could have done a better job from the beginning on the specs, like amd did with AM4, TR2 and SR2 platforms.
here is the reason: $$$$
$$$$ makes the world go round.
Next up, guide video how to make 9900k work on z170 motherboard. :)
Oh the good old Luumi? Need to check that one out asap!
please dont try
Don't tell me how to live my life! You are not my real mom!
Find the PIII discord group - there's a lot of mobo bios patches already done. I run 9900K on a Z170 Clevo laptop, completely stable and safe up to thermal limit of 160W so any decent Z170/Z270 desktop board would be fine.
Oh that's pretty cool! Cheers.
After watching so many of Roman's videos I'm slowly learning more German. Keep up the great content
This was interesting. One thing you haven''t really taken into account in the first test was the wires heatsinking the pin. It's possible the thick wires, and the fact it was soldered would help wick away heat from it. Soldering also creates better contact. I wonder if the results would have varied a fair bit if only the board side was soldered, and the pin side was touching a piece of perforated stripboard, to simulate a cpu substrate. I bet you wouldn't be able to fire 5A through it like that the same. probably good for 2A, and I bet the design safe amps would be 1A for each pin. It'll also be for redundancy, the number of pins, so that a few bad contacts won't matter so much.
der8auer...for me the volume was a little quiet and I had to turn you up and then music on youtube would blast into my ears after this video XD. Just letting you know so you can turn it up for next time. Happy New Year!
It's important to remind that soldering the PIN you reduced the resistance since the PIN is connected to the CPU by Surface contact. But the rest of the test is pretty conclusive.
Next project: soldering the CPU directly to the mobo!
"how to delid your socket."
There actually was some sockets that would melt small beads on the cpu so it gets stuck on the motherboard
That’s called BGA, it’s how most GPUs and mobile chips are mounted.
And the RAM and the storage! Wait.. am I Apple?
@Max Pain pretty sure i have an asus laptop with a socket inside and not soldered cpu (from like 2010 or something, gtx600 era as)
Thank you. Excellent video. It would be nice and interesting to hear from a CPU engineer to explain why such care is being taken by intel to reduce per-pin current. Maybe the contact point to the CPU itself is so small that it can mean a big difference to the result.
Pins can handle current. Good to know. But more pins means less resistance, therefore less V droop. I you have the pin still soldered to the cables. You can solder another two wires to the ends of pin to measure V droop on the pin. Should be interestiong to see. I know the power supply shows 0,2V but that is not accurate and it measures total V droop acros all the wires.
I'd guess the additional pins would be primarily to reduce impedance and resulting EMR, etc. Could also be futureproofing the platform for future chips that require more power.
yea bud, dream, future chips will get a new chipset,gotta ripoff customers some more
"future proofing" next release is gonna be a new socket lmao
The little round circles imprinted in the plastic of the socket are marks from holes in the injection mold that this part is made in. After the plastic is injected in the mold and has cooled sufficiently enough to hold shape, metal rods poke through holes in the mold that push against the part and pops them out of the mold so that the next part can be made. It's the holes in the mold that exist for the push pins(aforementioned metal rods) that actually leaves the circular imprint as the injected plastic expands into the edges of them(as it takes every single feature of the mold of course).
If you pay attention, you'll notice those little circles on E-V-E-R-Y-T-H-I-N-G, as almost everything in our lives contains at least some part or component that has been mass produced via plastic injection molding
This was hard to follow
RyTrapp0 I worked in plastic injection molding and that's right.
@@Razzaq585 Yup, I've had my fair share of it. Although most of my experience comes from operating twin screw extruders, I've had a little time around some other equipment too.
I love the way you communicate, and you did a very logical and well illustrated test here. I love your videos.
That Chinese and Japanese on your shirt is super strange. The Chinese reads "eye, entertainment festival, today". The Japanese reads "Please do extreme drying".
Is this some meme?
And how did you get the pin layout diagram? Is it public available from Intel?
I think the pinout is available on Intel, I've found the LGA1155 pinout years ago when I got bored and did some heavy searching
SuperDry is all about weird phrase in different languages. The Chinese character can be rearranged to “今日娛樂節目” (in Traditional Chinese) which translates into “Today’s Entertainment Show”
Actually all these phrases have no sense and they don’t follow any grammatical rules of that language.
reminds me of the porn actress riley reid, who tried to have "when life gives you lemons make lemonade" tattooed in chinese on her back. @@ChiHinNg
Intel gets slapped right into face.
Yes very cool, but can I fit a 9900K in my Z97
You can't because the voltage regulator is inside the cpu on that socketso that can't work.. but intel could totally make a LGA1155 Z77 DDR3 version of the 9900K if they wanted to, skylake memory controller can do ddr3.
Thanks for english version! RESPECT! +1
The comments below miss how much more is involved here. So if the pin gets up to 60C people say, oh No Big Deal, its a scam. But those pins are then transferred to much smaller wires to the circuit, and then orders of magnitude smaller as it gets to the IC. 1A is actually an incredible amount for those wires. Plus you have the heat from the IC itself- Der8auer should know this well. You dont want additional heat just from getting the power to the chip adding to the heat of the chip itself. It takes away headroom. Its really the transfer from the package to the IC that I would be worried about. The size of those wires is so tiny. Take a look at whats under those if youve never seen it. Normally you dont because of the flip chip method and the epoxy that covers it.
it's a flip chip, there haven't been bond wires in big x86 CPUs in probably over a decade
Yes I was thinking about the same thing. More pins is not for the current is fo the resistance. Resistance means voltage drop. Traces inside pcbs are relatively smaller than the pins. And inside the die of the CPU this means even more difficulties.
Having more power pin also mean have more power distribution on the die. Pins are also used to remove hear from CPUs.
I don't think that Intel made this change only for those reasons. If they really wanted to screw things up they simply could swap two data pin like they did on 771 and 775.
TuttoCrafting he der8auer is a smart dude but he’s testing totally the wrong thing here. I’m sure spambot is totally correct. I’m sure my info is a decade out of date. But regardless that pin is 100x larger than the trace it connects to. Which means 100x the resistance per unit of measurement. It’s the cross sectional area that matters. It’s somewhat irrelevant how much power a pin can take. It’s not the bottle neck. I guess you’d have to crack open a chip and see how big those traces are and see then. I’m sure there’s probably a ground plane and a Vcc plane to try and minimize droop and drift. But not everywhere. To me it looks like the beefed up to Vcc and Vdd to accommodate more power and generate less (useless) heat.
spambot71 you’re right. I knew that, I just wasn’t thinking. But yea this just furthers my point. Because what you described is even smaller.
Hear what you're saying- but remember putting a coffee lake CPU into an older socket would still only be about 15% extra current draw per pin. Do you really think Intel is already running so close to the such a red line? Especially considering the LN2 overclockers are pulling 300% extra current sometimes and no traces melt. The CPU/mobo either dies from too high a voltage or condensation/water causing a short.
Come one, Roman! The video is awesome and I thank you for it, but the question "will it melt" has a simple answer:
Has anyone melted the socket of a Z270 board while overclocking an i7 7700K to over 200W? Well, no!
Please keep this up. Intel needs to pay attention. Thank you for all you do der8auer
Interesting video, glad you made it!
Happy 2019 to you, hope to see a lot more of your stuff in the coming year. =)
Enjoy Vegas, and make sure to give Steve a bro-hug on camera. He needs one.
What a number of pins taped xD
Enjoy the night and happy 2019!
Btw catch up with Steve from GN to give him a german class lol
Thank you for the analysis. Happy new year der8auer.
Hey Roman, does the pin reduction have any effect on the motherboard's VRM phases? How many remain used and temperature changes?
Thank you for the work and for the English translation!
Very! interesting! Happy New Year to you!
is there perhaps an importance for extreme uses? does the extra surface area for power from these extra Pins make some difference for Chiller or LN2 uses?
they surely didn't build the Socket this way for extreme Overclockers, but even so.
When can US residents order the oc frame for the 9900k? Thanks for the videos btw, and have a Happy New Years!
sometimes a ground pin is only connected externally and not to any of the other ground pins of the device - so missing a power or ground connection can cause device functions to fail depending on the topography of the device
Enjoy CES, and thanks for the great content in 2018. Happy New year!
Awesome video man. Very interesting, I wonder what Intel's response to this would be. BTW, love the music but damn it's such and intense beat for the video. Seems more suited to a wing-suit video, or someone going to space lol. Need more content like this!
Could be a problem in the maximum amount of current can be transfered to the die, I don't know if still use thin gold wires or tracks but may be there are too thin.
When you taped the pins, did you check if it didn't get punctured after testing? Images shoes that they do quite a bit of pressure when the socket locked down. I see some small white point.
Yep checked that
@@der8auer So why there are soo many pins if they aren't really needed, save proof or ? Happy new year btw :)
Best wishes and a happy new year 2019, Roman :)
thank you very much for your videos and work
I just got an add that was a full length linus tech tips video. It was his video about the shadow game streaming device. Wow.
Roman. Can you try solder the cpu to the board without the socket for maximum delid?
So it still works in the end, what a drag Intel, as you have always been.
I know, right. So much!
Considering LGA775 & 771 in the past and how they actually are cross compatible, this doesn't surprise me. Even less so since a group in China has also managed to make BGA CPU socket mountable (only functional with certain boards & BIOS however).
I've had an LGA775 Xeon rig running for several years now, with any issues being caused by a bad PSU (replacing it fixed the rigs instability).
Running an SFF 771 Xeon-modded Lenovo Thinkcentre 24/7 as my primary internet gateway for the entire last year only rebooting for kernel updates :)
This is not the same thing. This is power delivery.
Back then they just flipped 90deg the CPU layout to make Xeon CPUs not working on standard mb.
If they wanted to do the same thy could simply have changed a couple of pins and not a lot.
Changing pins is not cheap on CPU design. (That why back in time they only swapped two)
I think that in this case they have seen some problems with power distribution. This is a single test on a single cpu. This does not address all the cases. This may be a lucky CPU.
You have to take the fact that your thick copper wires are very good thermal conductors into consideration. Also the resistance between the pin and the pad will certainly be different than your soldered connection.
Please make a new test setup where you use a dead cpu and a cut in half socket on a pcb to properly test the thermal scenario.
I would also like to see a proper resistance measurement of one of those pins using a Kelvin-clamp resistance meter.
he could run the current through 2 ground pins, so that they'll be connected by the interposer's ground plane. if he wants to measure voltage drop he could have separate probes on each of the lands and subtract the internal voltage drop from the total voltage across the pins themselves. would get a bit cluttered with 6 wires total but would be a proper test
What happened to your phase change AIO that was going to be release end of 2018?
Another great wideo from a top overclocker.
I think a pin would heat up more if embedded in the plastic, rather than open to the air. And you say the metal can handle high temperature, but what about the plastic around it?
any idea what will be after LGA 1151 v2 ? v3 or something different ?
GW, great year and all the best for 2019 tschuss
Happy New Year der8auer ;)
Please correct me, i believe the CPU socket you refer to as "plastic" which i doubt is actually made from some formulation of fiberglass & epoxy thus would make it much more resistant to high temp than ordinary plastic. Observe that as you cut with the Dremel it cuts like PC board material.
This is a worst case fault scenario too. Wouldn’t having a resistive load on one side of the pin more accurate?
Happy New year, Great Vid.. I will be trying to run a I7 8700K, on a Maximus VIII ITX board, if I can find a moded BIos for the board.
very informative content dude! keep it up!
Respect to your test methods and video topics
Very interesting video Roman. Thanks for sharing!
Can you do a video bout that microscope setup? I'm looking to upgrade.
Perhaps only one or two phases of the vcore vrm were still connected to the cpu? It'd be interesting to see the temps of individual vrm phases as you tape off pins, one would expect some phases would no longer be connected to the cpu and thus the other phases pick up the slack and get hotter.
no one designs vrms like that. all the phases are connected together with a big beefy copper plane
So all phases are simultaneously connected to all the relevant pins in the socket? There's maybe 8 vcore phases but 18 or more pins transferring that power to the cpu. So there's dozens of traces from each phase of the vrm to each of the cpu pins?
@@ericwright8592 there's not "dozens of traces"; there's a large copper rectangle underneath the output of the vcore, like a PCB version of a bus bar. it's then usually stitched with lots of vias (copper plated holes to interconnect layers) to another large copper rectangle in one of the internal layers (PCBs tend to have 8+ different copper layers, each with their own design). Then each vcore pin on the cpu is connected to this internal plane by one or more vias each. so you can think of it as a big sheet of copper that all the vcore phases connect to and all the pins connect to.
@@spambot7110 very interesting, I had no idea. Kind of want to tear apart an old motherboard now just to see. Or perhaps someone like Der8auer could do it for us?
@@ericwright8592 it would be pretty tricky to get those layers apart; they're very thin and laminated together with very strong glues. it would be very cool though!
My mobo i still use is next to der8auer's left shoulder (right side of video) with mildly OCed i7 6700k to 4.5ghz no delid
That is amazing. So changing the number of pins is an excuse to make you buy a new motherboard. How many data pins are there on a typical CPU?
Those are really good scissors! What is the manufacturer so I can buy them too.
So is it possible for me to get a z370 or z390 board and use my 7700k at some point?
guys, have you instruction for modified MB Asus Maximus Hero IX (z270) for Coffe Lake ??? If you have, pls help.
Awesome! Thanks for yet another interesting innovative experiment! If the 9900K in my Z170 Clevo laptop is reduced to smoking rubble I can rest assured it wasn't because of the missing vcore pins :)
Stock 95w ? Yeh ok intel , whatever you say.
Also , more flux..... or as in .... use flux.
as in F.lux?
i hate losing Color Accuracy, personally.
instead, turn the Brightness down from eye searing to a more normal level, don't cram your face into the Display (but sit a foot or two back - or ballpark 0.6m if you prefer), always have ample ambient lighting in the room, Et Cetera.
taiiat he's talking about solder flux.
probably also power distribution to get it as even throughout the socket/die as possible
I would have thought it would be to do with stability for certain parts of the core/cores.
Due to the volt drop under higher currents.
More pins = less volt drop = hight stability, especially if your going to do LN2
So i CAN upgrade my g4560 to 8th or 9th gen cpu?
Moin 8auer, ich habe eine eventuell ungewöhnliche Frage - ich habe im Moment ein Asus Z170i (Mini ITX) Mainboard. Ich habe mir allerdings letztens ein neues Case gekauft (NZXT Source 340 Elite) weil ich ein Case wollte welches eine Glasseite hat damit ich anfangen kann mit modden. Allerdings wusste ich zu dem Zeitpunkt noch nicht das ich ein Mini ITX board besitze (ich hab mir meine PC's immer zusammenstellen lassen, ich habe sie nie selber gemacht). Nun brauche ich also ein neues Mainboard mit einem z170 Chipsatz und ich sehe in so vielen Videos bei dir den Karton des Maximus Vlll Hero - das einzige Board, bei dem all meine jetztigen Komponenten passen würden - wäre es irgendwie möglich das ich dir dieses Board abkaufen kann? :P
Liebe Grüsse Sharleg
Great job, Roman!
One difference between his test and a motherboard with an actual CPU is that the pin isn't going to be soldered onto the CPU.
The non-soldered contact between the pin and pad is going to have higher resistance than soldering it to a wire, which should cause it to heat up more.
Remember the soldering part was only 1 of 2 experiments performed in the video -- he covered 68 pins with elec tape and ran his CPU at a much higher wattage than a 7700k or 6700k would be capable of pulling under full load. After 24hrs there was still no degradation under the microscope (even in his extreme example).
can you make a tutorial on how to mod z170 and z270 bios to take cofee lake cpus :D
So Jimmy Neutron is doing computers nowadays! Good video, I love it.
Maybe GPU display pins and/or motherboard manufacturer specified trade secerts for what those pins are for but most likely SATA controller maxed out, USB 3.0 controller connector, DMI for full speed, direct connect of M.2, and maybe a few other things that you cannot find by just covering pins on the CPU without other stuff also plugged in.
uh, they're vcore and ground pins. they're documented in the datasheet, and if he wanted to verify he could probe them with a multimeter in continuity mode; he'll get a little beep for all of them because they're all internally connected together.
Z390 is more about peripherals (much more USB3.1 and USB3.0 ports than the Z370) than the Vcc pin Intel plans to use. Good video by the way ;)
Wow thank you for the info.. I just have to ask was this fun for you? Because I don't mind watching but to run these test would be way too much for me. Thank You again for your time and effort
Yea I really enjoy doing this kind of thing :D Exploring and experimenting
Unless i wasn't paying attention : 1.35v x 1.16a = 1.5525w per pin oced your bench psu test 0.23v x 5.098a = 1.17254w so you missed the current draw mark by 25% but more than likely that pin could go to 5w fairly easy.
Very informative video as usual, thanks for sharing.
Very skillfully demonstrated
the background wall has only artefacts :O
i have every GPU from there but my GOD those motherboards
And in a surprise to no one, Intel "excuse" for Z370 was actually nothing. Still, more than 60 pins taped, damn that is overengineered AF. Also, thanks for the English content. =)
AM4 does fine with Ryzen vs. Ryzen 2. We will see how it does with Ryzen 3 and if any problems arise with power delivery to the chip (I don't think they will). They said it was for the next 10 years so Intel could definitely do the same. Pins have never been a valid excuse.
Happy New Year!
I'm thinking it should be fine. AM3+ was fine on 500+w from FX series, so I doubt it will be a big deal (I can't even hit that on my TR 1950x, but I haven't used LN2 on it yet). AM3+ was "plus" sized over AM3 with fat pins for more current, and now AM4 looks like spider webs by comparison (AM3+ was pretty safe to just set the CPU on its pins, even small drops wouldn't bend it), however, I think AM4 will be just about as capable since it has so many more pins. Thinner, but many more. Also the power draw is not through the roof like FX, so that's the biggest thing.
Huh pin looks alot like a part from a disposable lighter
Could You show how to run 8th gen on z170?
youve got 2 soldered contacts though. if you just had it touching a pad resistance would have been more
How about the voltage on each pin? The heat from wattage is the real measure for this test. Unless you KNOW there is only 0.23v like you tested through each pin.
Was tun wenn ich die Abstandhalter und die schraube für eine M. 2 ssd verloren habe.
My 9900k is in an EVGA Z370 Classified K. Can't seem to get it past 5.1GHz (all cores) though :)
And that is pretty freaking awesome if you can do that stable. Are you disappointed that you can only do 5.1Ghz all core?
@@KungPowEnterFist - nah but I would like slightly lower temps (already on a full custom w/c loop)
I would be much more likely to buy a new CPU if it "just worked" on my Z170 board. Buying both board + cpu again so soon after upgrading is just not a good return on investment.
Could be possible for your motherboard. Custom BIOS flash and pin tape mod (like what roman did here but only a few pins). And better cooling ;p
Hello, I have to somewhat disagree that z270 and z170 could be usable for 9900K, even z370 is NOT.
I own a Gigabyte Z370 AORUS Gaming 7, after updating it to latest BIOS I got 9900K, I did BIOS reset and LOAD BIOS Defaults.
The Board is on all default settings.
with my 9900K it ALWAYS boots with 1.34-1.36 vcore and under load in windows cross 1.4
I assumed that its fine for this CPU but then googled
Apparently on z390 boards at default setting the 9900K boots at 1.2v
I googled some more and found otehr people having same issues as I do [BTW seating Manually vcore to 1.2v wont boot at all]
So I got Z390 AORUS Master
P.S. NONE OF OTHER below 9900K CPUs have such issue on z370,the i7 and below all work FINE, only i9 have this over voltage
wouldnt lga sockets suffer from oxidization/corrosion?
pga never ever had those issues.......?
4:39 ....... 😭
Is you power supply only putting out .23v at 5 amps? Why aren't you running 1.35v at 1.15+ amps?
I have an i7 6700K that I purchased 2 years ago, thinking I would build a computer, but I never did. Should I build a new computer with it now ? Or should I just accept the $ 300 loss, and start over with a new CPU on a Z390 ?
wtf?
now this is quality from intel boards and package itself
As people had expected right at Z370 launch. Intel purposely made "8th/9th Gen" CPUs incompatible, not because there was an electrical need for it, but because it both makes more money to sell people another Motherboard, and saves money in support not having to deal with people buying boards without appropriate BIOS Versions for newer CPUs, as AMD had to deal with with Zen+ CPUs on 300 Series AM4 Boards.
11:00 I see a flaw in your test method. You are using a CC/CV power supply and so at 5 amps your power supply is delivering only .23 volts and that is only 1.15 watts which is not the correct voltage or wattage that would be flowing through the pin. Even at the lower amperage setting your power supply is limiting the voltage to get a specific current. This is not how the CPU and VRM will be working and so this would not be a valid test of the pin.
However, with all that information we can calculate the resistance of the pin which is between 0.04222 and 0.04512 ohms at the temperatures measured but that does NOT include the contact resistance between the pin and the pad on the CPU interposer which will significantly contribute to its power handling capabilities. And the contact resistances will not always be constant, they can change with vibrations. Also, the contact resistance will change OVER TIME especially where there is heat cycling and this is the same no matter what type of mechanical contacts are used. The pins will begin to deform over time because heat cycling has an annealing effect on the metal pins. So the contact pressure will lower over its lifetime causing more contact resistance to build over time. Meaning more pins are necessary for the power delivery to be stable over its total predicted lifetime and expected operating environment. That doesn't mean that Intel was right in changing the socket pinout to add more power pins and I'm not defending it, it just means your initial test method was flawed.
This annealing effect on contacts is the main reason why taking a computer apart and putting it back together can bring an unstable computer back to life. It doesn't reset the pins back to their original glory but it does re-seat the contacts allowing it an extended life. Bending the contacts up slightly also helps, and the same goes for PSU connectors and connectors between the boards on hard drives and the motors. Over time they will begin to settle into position and more resistance builds between the contacts. It isn't always the circuitry in the PSU that makes it unstable, the female Molex pins on PSU connectors can become loose over time and cause stability problems especially if the PSU uses one of the many voltage lines on the mother board connector for feedback. So if your computer shuts off unexpectedly and you suspect power supply, try getting those female Molex pins in the connectors to hold on tighter.
The pin connectors on hard drive boards can also become loose over time and cause the motor controller to error out which means the hard drive continuously tries to spin up but can't. And again, removing the board and bending the contacts so they make more contact pressure can bring the hard drive back to life but eliminating the connector altogether by hard wiring the motor to the board can make it last until the heads decide to crash.
Even on modern PCs and laptops connectors are inevitable but they are also inevitably a problem given enough time. If you could eliminate mechanical connections, exposed metal traces/contacts and electrolytic capacitor aging then you can theoretically create a system that could last a lifetime without physical maintenance other than dusting...
It's more to give headroom for reliability issues, I'm guessing. Not for physical reliability wrt mounting and dismounting, but from actual long term wear and tear from the electrical current flowing through either the pins, solder balls between the package and die, or some part of the interconnect structure. The extra pins can also be an effort to improve the stability of the power grid on the chip.
Not everything is a ploy to extract more money out of customers. Chip designers usually know better than overclockers when it comes to what the design actually needs to function like it's promised for a long as it's warrantied
haha riiight.
if you test latest ryzen pin from old fashion outdated tech off the cpu from substrate ...i bet it couldn't do 5 amps at all...it goes burn't and overheated on amd pins...so 3 amp is max limit for amd ryzen per pin while intel can push it further in temps and amps per lga ball pin ...to 20 amp max
8100 works fine in 2xx, biosmod. This pins is for a more processors in die.
hey guys, i wanna buy a ryzen 2600 and 2x8 gb ddr4 ram. can you guys recommend a cheap mainboard? dont need to overclock, also i dont want to need to update the bios. :)
80's sounding music to go with the shirt