An Analysis of the Z390 Socket Proves the Extra Pins Aren't Necessary

Indeed. A friend and I were trying to figure out why... I think it was his 1090T, wouldn't work in his old AM2+ board. I told him that mine was reporting (via AIDA64) to indeed have the DDR2 and DDR3 controller, so technically it should have worked fine. I'm sure it was all down to a BIOS incompatibility for his. Even the Athlon II, which were still PhII based, had the DDR2 controller intact.

Here's another similar tidbit that I had just came across recently that wouldn't surprise me if no one knows...
The Intel DDR4 CPUs still support DDR3, up until the 8th gen at least.
Examples: 6700K, 7700K both list DDR3 for "Memory Types" :D
There were quite a few socket 1151 motherboards at release supporting DDR3L and plenty still on sale.
 
They literally reduced the amount of power pins and said intel is screwing the customer because it works without them. This is exactly the same as dropping cylinders. Just because it works doesn't mean it's right.

Maybe Intel is jacking us around and it's a blatent 'fuck you, because we can' or maybe there is some obscene edge case that caused the v1 socket to burst into flames.

I take it you don't know how engines work. The 6-core part is still able to run on all cores (or cylinders using your analogy). Taking off half your spark plugs would make your truck run like crap. The 8700k does not run on crap using a Z-170 if Intel didn't mess with it.
 
They literally reduced the amount of power pins and said intel is screwing the customer because it works without them. This is exactly the same as dropping cylinders. Just because it works doesn't mean it's right.

Your analogy would've worked better had you said you removed one spark plug from a V12, though. (I"m not saying you were wrong, although modern V8s that are capable of shutting off some cylinders on the fly weaken it.)
 
I can disconnect four of the eight sparkplugs on my v8 and still make it home, doesn't make the extra cylinders a cash grab.

I think you should consider replacing your 1/4 inch gas line with a 2 inch diameter gas line on your car... That will make it go really fast! After all that is essentially what intel is doing.

If you don't get the sarcasm then let me be blunt. You really don't know what you are talking about!
 
If it's that simple to test the uselessness of the added pins, shouldn't it just as simple to retromod a Firmware on a Z270 board and drop a 9900K in? Men did it with an ABit BP6 and P3 processors. Intel continually gets flack for their short lived 2 generation support for chipsets. I'm betting that AMD R9 flag ship will shit bricks if it works at all on a X370 chipset board.

For the people that think the car analogy was crap: You can take a Stock Evo9 and tune 30+ more HP out of it but your injector duty cycle will 99% and fuel pump will be maxed out. How long before they fail, engine runs lean and you shred the engine block?
 
I take it you don't know how engines work. The 6-core part is still able to run on all cores (or cylinders using your analogy). Taking off half your spark plugs would make your truck run like crap. The 8700k does not run on crap using a Z-170 if Intel didn't mess with it.

I think you should consider replacing your 1/4 inch gas line with a 2 inch diameter gas line on your car... That will make it go really fast! After all that is essentially what intel is doing.

If you don't get the sarcasm then let me be blunt. You really don't know what you are talking about!

Let me make it clear for everyone who lacks reading comprehension, just because you can take something away from a system and it still works doesn't mean it's right. If you take away spark, fuel, whatever from part a motor it will run like shit but still get you home. You can't disrupt a precisely engineered system and expect it to be right.


I'm inclined to believe a multi-billion engineering facility who invented the fucking product at question would have a bit more understanding of what is happening than some hack with a roll of tape, a couple spare hours with a video camera and a chip on his shoulder.
 
No one should take this amateur analysis seriously. Did he analyze the socket and CPU under elevated temperature conditions? Did he artificially age the socket and CPU to introduce oxide, sulfide, or nitride contaminates at the socket/CPU interface? Did he run the test for years? Did he factor in the potential impact of socket or CPU warp over time or from manufacturing variations, which might cause some pins to lose or have reduced contact? Did he look at how much extra power was wasted by the increases heat generated by reducing the number of pins? Did he consider the effect of accelerated reaction rates due to pin heating on long term reliability?

And let's crank the numbers: the revised LGA 1151 adds 18 more power pins (going from 128 to 146).
The video demonstrates that each bin has a little more than 0.045 ohms resistance (0.23V voltage drop at 5A).
That makes the aggregate power pin resistance for 1151V1 14% higher (0.35 milliohms) the 1151V2 (0.31 milliohms).
Doesn't sound like much, BUT those Coffee Lake processors are pulling 138A through those pins.
The extra power pins reduce the resulting voltage drop across the pins (not to mention the drop across the wiring connected to those pins) from 48mV to 42mV, with a corresponding decrease in power dissipated just by the pins of almost 1 Watt. And that's without overclocking. Sure, it's only 1W, but it's 1W at a place that's hard to get extra cooling to -- what the thermal conductance of the material they make sockets from, anyway?

So it's one thing to say "there's no need for these extra pins if you're just going to run your system for a week."
But it's an entire different matter when your customers expect the socket-CPU interface to not be an issue for three years.
Or when your customers expect to be able to feed 150A to an overclocked CPU without the socket melting.


He first showed what the overclocked current would be on a Z-170:
pin.PNG


He then ran 5 amps through a SINGLE PIN with no signs of damage.

power 2.PNG

Finally reduced the pins until he was down to 69 power pins were blocked and ran prime95 overnight while oveclocked. Going from Z-370 to Z-170 only reduces the the power pins by 18.
 
Wished he had covered the VRM phase balancing question with reduced pins. Did certain phases on the motherboard get hotter, while others were disconnected? In line with this and just speculating... may be Intel over-engineered the pins to support max 32-phases, even though no board wires in more than 16 so far?
 
Please do take a look at Apple's laptops, they've been using DDR3LP on the latest and greatest intel chips.

There were quite a few socket 1151 motherboards at release supporting DDR3L and plenty still on sale.
I stand corrected then! Clearly showing my Green-Camp colors here with how little I pay attention to Intel's stuff :X3::eggface: lol
 
Let me make it clear for everyone who lacks reading comprehension, just because you can take something away from a system and it still works doesn't mean it's right. If you take away spark, fuel, whatever from part a motor it will run like shit but still get you home. You can't disrupt a precisely engineered system and expect it to be right.


I'm inclined to believe a multi-billion engineering facility who invented the fucking product at question would have a bit more understanding of what is happening than some hack with a roll of tape, a couple spare hours with a video camera and a chip on his shoulder.

You act as if nothing in the world is over engineered. If 128 power pins is "precisely needed" for the 7700k and 146 pins are "precisely needed" for the 8700k, why aren't 164 pins needed for the 9900k?? A 6R80 transmission works for the Ford V-6. Guess what, a 6R80 works just fine for the V-8. Your unhooking spark plug analogy made no damn sense.
 
Not surprised in the least. Intel has been manufacturing obcelesence into their chipsets for years.

As someone else said, I’d have upgraded if I didn’t have to replace my motherboard. I’m sure they’ve done the numbers though.
 
He first showed what the overclocked current would be on a Z-170:
View attachment 132623

He then ran 5 amps through a SINGLE PIN with no signs of damage.

View attachment 132625
Finally reduced the pins until he was down to 69 power pins were blocked and ran prime95 overnight while oveclocked. Going from Z-370 to Z-170 only reduces the the power pins by 18.

Exactly. amazing what people can learn by actually watching the video rather than just blindly defending Intel.
 
Actually it is a bad analogy because in his testing there isn't performance loss. You have lost 50% of work in yours.

It might not be prefect, but the idea he was trying to get across stands.

It might have been better to use a fuel pump, as it is what supplies the fuel, just like these pins do to the chip. In most cases fuels pumps are close to double the flow needed, my autox car reached almost double OEM power with the stock pump, does that mean its not needed? Without a doubt no, as running a pump at half the flow potential means a higher duty cycle of the part, which can result in more problems even if it would still run without power loss, not even getting into degradation over time that the overhead helps mitigate. There are lots of factors as to why something is engineered a given way, from lots of testing and looking at worse case situations and designed life spans.

To find that it works without them under limited and short testing is fine, but to make a sweeping statement that they are not needed at all is a bit premature.
 
Let me make it clear for everyone who lacks reading comprehension, just because you can take something away from a system and it still works doesn't mean it's right. If you take away spark, fuel, whatever from part a motor it will run like shit but still get you home. You can't disrupt a precisely engineered system and expect it to be right.


I'm inclined to believe a multi-billion engineering facility who invented the fucking product at question would have a bit more understanding of what is happening than some hack with a roll of tape, a couple spare hours with a video camera and a chip on his shoulder.
If you believe what you wrote there you need some education or to just quit reading the [H]. As the video points out, overclocked CPUs regularly draw close to double the current of stock CPUs - for years at a time and without any melted sockets. Explain to me how 14% more current handling capability is absolutely required when experience has shown that the previous design was capable of handling double the rated current. If you do watch the video I'm sure you noticed the point where he tested an individual pin up past 5a (a 900% increase in current over stock if you're wondering) without melting the pin... Yet you'd rather trust the guy with a pretty obvious reason to sell more chipsets? Ok.

Also, your engine analogy shows a complete lack of comprehension - since the new pins are VCC and ground, the ONLY thing they can claim is that the new chips require more current. This isn't cylinders of an engine, this is the diameter of your air intake hose. A better analogy would be the wire gauge of speaker cables: Intel (aka Monster) tried to claim that you need their special "oxygen-free" 12awg speaker cables for your new speakers to work. Why? Because current and stuff. Of course these speakers are 8ohm jobs, and the cable length is only 2.5' so realistically you really only need 16awg, but the billionaire told you you need 12awg cables so... Just like the Z390 it's 100% bullshit.
 
"The conclusion is that the LGA-1151v2 is absolutely unnecessary."

Anyone shocked by this?

Not in the least.

This is ancient news. AMD got big as a second source licensed x86 clone for IBM. Once AMD started selling enough 486 CPUs (often dropped into Intel chipset boards in IBM clones) that it was taking a noticeable bite out of their pocket books, Intel decided to become a complete chipset platform company. It has been making it difficult to not have to change to a matching chipset with new processor releases ever since.

All true.

This is why Intel is in decline.

Hardly. It isn't even remotely that simple.

I can disconnect four of the eight sparkplugs on my v8 and still make it home, doesn't make the extra cylinders a cash grab.

No it doesn't, but that doesn't mean that the extra pins are actually needed either.

Indeed. A friend and I were trying to figure out why... I think it was his 1090T, wouldn't work in his old AM2+ board. I told him that mine was reporting (via AIDA64) to indeed have the DDR2 and DDR3 controller, so technically it should have worked fine. I'm sure it was all down to a BIOS incompatibility for his. Even the Athlon II, which were still PhII based, had the DDR2 controller intact.

Here's another similar tidbit that I had just came across recently that wouldn't surprise me if no one knows...
The Intel DDR4 CPUs still support DDR3, up until the 8th gen at least.
Examples: 6700K, 7700K both list DDR3 for "Memory Types" :D

Those CPU's might have had the DDR2 controller intact, but that doesn't mean that it was actually connected to anything. It could have been disabled or AMD chose not to use it for a number of reasons. Intel and AMD CPU's often contain hardware that's ignored or disabled to modify the processor for certain markets. Disabled cache, disabled cores, removed iGPU, etc. have all been done before. Even if AIDA64 was right, it doesn't mean that a theoretical AM2+ board could use a newer AMD CPU that was designed for AM3. VRM specifications, microcode and other variables come into play.

As for Intel's 7700K supporting DDR3, it only supports low voltage DDR3 typically found in mobile applications. It was an option for the desktop market, but one that wasn't utilized a whole lot. It certainly wouldn't have been worth doing on enthusiast systems which used higher clocked DDR3 modules which took far more voltage than low-voltage DDR3 modules do.
 
I've seen many a scorched board... maybe this needs to be run full tilt for over a year or so before drawing any conclusions?

What would be the point of that? Intel will have a new socket by then.
 
How blatant this is they don't get a pass on this one. Forget the 9900k, 8700k, or even 7700k.
If it's that simple to test the uselessness of the added pins, shouldn't it just as simple to retromod a Firmware on a Z270 board and drop a 9900K in? Men did it with an ABit BP6 and P3 processors. Intel continually gets flack for their short lived 2 generation support for chipsets. I'm betting that AMD R9 flag ship will shit bricks if it works at all on a X370 chipset board.

For the people that think the car analogy was crap: You can take a Stock Evo9 and tune 30+ more HP out of it but your injector duty cycle will 99% and fuel pump will be maxed out. How long before they fail, engine runs lean and you shred the engine block?
It's been done, Intel just wanted to sell more chipsets and or motherboards with the aibs.
 
Also, your engine analogy shows a complete lack of comprehension - since the new pins are VCC and ground, the ONLY thing they can claim is that the new chips require more current. This isn't cylinders of an engine, this is the diameter of your air intake hose. A better analogy would be the wire gauge of speaker cables: Intel (aka Monster) tried to claim that you need their special "oxygen-free" 12awg speaker cables for your new speakers to work. Why? Because current and stuff. Of course these speakers are 8ohm jobs, and the cable length is only 2.5' so realistically you really only need 16awg, but the billionaire told you you need 12awg cables so... Just like the Z390 it's 100% bullshit.

"But, but if you use that 16 Ga non oxygen free cable, it might degrade over time!" I love how these guys keep on pushing these crazy car analogies. Now they are comparing AFTERMARKET tuning on a car and how long the fuel pump lasts or something.
 
If it's that simple to test the uselessness of the added pins, shouldn't it just as simple to retromod a Firmware on a Z270 board and drop a 9900K in? Men did it with an ABit BP6 and P3 processors. Intel continually gets flack for their short lived 2 generation support for chipsets. I'm betting that AMD R9 flag ship will shit bricks if it works at all on a X370 chipset board.

Maybe it will, but guess what...AMD is ALLOWING you to use it (although you will not be able to overclock as well as an X570 or whatever). Also, AMD is not stopping people from upgrading to a 3700x using their very capable X370/B350. That is the difference.

There is absolutely nothing wrong with Intel releasing a Z-370 and Z-390 with high capabilities (or at least percieved ones). But stopping people from upgrading their CPUs without buying a new $150 motherboard for no reason is a dick move.
 
I always want the features from a new motherboard everytime I get a new CPU anyways, this isn't an important revelation to me. When intel was talking about going BGA I was fine with that too...

Maybe I don't buy often enough to see this as a problem.
 
I still want an official statement from Intel.

I also want to see somebody run their 9900k with all those taped pins for a few years and then report back. My guess is that either the CPU will die or the motherboard will die... but I could be wrong.
 
As for Intel's 7700K supporting DDR3, it only supports low voltage DDR3 typically found in mobile applications. It was an option for the desktop market, but one that wasn't utilized a whole lot. It certainly wouldn't have been worth doing on enthusiast systems which used higher clocked DDR3 modules which took far more voltage than low-voltage DDR3 modules do.
In complete fairness, I have a 2x2GB kit of KingMax Nano Gaming DDR3-2200 10-10-10 1.5V (did 2400 at that voltage) which was definitely low voltage compared to my 2x1GB OCZ Reapers DDR3-1600 7-7-7 kit that was 1.9V (with those Micron ICs.
The KingMax were released in 2008ish and so their speeds were respectable for the time, at least IMO They're not as good as my 2x4GB Hyper-X DDR3-2666 1.65V released not long after, but those also used big heatsinks whereas the KingMax did not, so there's that. Granted, the Hyper-X didn't run what I'd say "hot", not like those OCZ did, but they weren't cool to the touch like the sink-less KingMax either.

I know that the Intel IMC was far more finicky to RAM voltage compared to AMD, and so comparatively 1.5V is still quite a lot higher than the 1.35V in mobile.
 
<cough>...... Crickets? No official Intel response? That's what I thought ;)

Actually, I think if you look at the feature list, you'll find it quite extensive. In fact the column for Spectre version numbers had to be made larger, that's how many new features have been added!!
 
I always want the features from a new motherboard everytime I get a new CPU anyways, this isn't an important revelation to me. When intel was talking about going BGA I was fine with that too...

Maybe I don't buy often enough to see this as a problem.

While I am inclined to agree with this sentiment, you get almost nothing going from Z270 to Z370 and to Z390. The changes are minor throughout. The only real difference between Z390 and Z370 is the integrated USB 3.1 Gen 2 controller, which isn't any better than Alpine Ridge or Titan Ridge. It certainly costs less, but it also lacks Thunderbolt 3 support. Not that this matters to most of us. This chipset benefits Intel and really no one else. It was probably billed as a cost savings to the motherboard manufacturers as if they were getting a deal or had any choice as to whether or not to produce a motherboard or series based on the new chipset.
 
Indeed. A friend and I were trying to figure out why... I think it was his 1090T, wouldn't work in his old AM2+ board. I told him that mine was reporting (via AIDA64) to indeed have the DDR2 and DDR3 controller, so technically it should have worked fine. I'm sure it was all down to a BIOS incompatibility for his. Even the Athlon II, which were still PhII based, had the DDR2 controller intact.

Either his board didn't support 125W CPU's or he never bothered to do a bios update, assuming it was a decent brand that still released updates
 
I've been in the electronics and IT biz for a very long time. There is truth to both sides of this argument.

Is Intel being a bit greedy by forcing the new chipsets and mobos so often? Yes. (duh)

Is there a real reason why power/ground/signal pins need to be added or moved sometimes? Yes.
It's a bit more complex than some hacks on a website saying, "Look, it's not burning.".

Intel does want to sell chipsets and help their partners sell mobos with every new CPU family release.
There's no mystery on that. And yes, many times the incompatibilities are artificial.

But to start taping off lots of power & ground pins and claim they are pointless is silly. This "testing" by
disconnecting pins and saying "Look the rest of them aren't burning." is not how electronic engineering is
undertaken. That's not how it works, there's a bit more to it. lol

Reminds me of a customer many years ago who was upset at the cost of a replacement server mobo.
He said, "Those metal legs on the chips are just jumpers that go through the plastic, why are they so
damn expensive?". He had no concept of the micro-circuitry inside the chips. His eyes glazed over as
I started to explain what was inside the chips, so I just ended with, "It's a bit more complicated than that.".


And I have to say that beating people up over their analogies is asinine.

.
 
Christ, zealots on both sides are annoying. I bet an Intel engineer that actually worked on this would laugh at the notion that it was purely an arbitrary construct to milk people (not that Intel hasn't done it before). I'd lay money it has to do with bolstering power delivery and raising ceiling on spec - they *are* energizing 2 additional physical cores at each step from 7700k -> 8700k -> 9900k, and on unlocked CPUs have to factor the worst case overclocking scenarios stretching the limits.

My next build will be a Ryzen because AMD will sell me 8 or 16 cores for cheaper than Intel is willing to, but the idea that Intel did this purely to compel hardware upgrades with no engineering rationale is hard to believe.

In fairness, der8auer is the real deal, he's not any old YouTube hack.
 
Last edited:
There were quite a few socket 1151 motherboards at release supporting DDR3L and plenty still on sale.

Yeah I have 2-3 year old Dell convertible type with a skylake i3-6100u running DDR3L 1600 ram.
 
Didn't watch the video.. but would taping off said pins effectively reduce the number of power supply phases that normally feed power to the CPU?

Power supply phases feed into a common power bus. Multi-phase power supplies are about delivering a cleaner power supply signal and aren't fed separately into the CPU. This is why you can have great variance in the number of power supply phases across motherboards. Basically each phase creates a high frequency sine wave and they overlap to simulate a continuous supply.

Even though those "extra" pins are not necessary for the CPU to function, I kinda think that they are there for a purpose.

Socket pins specs are specified over a long term life at a wide range of environmental conditions with significant margin. Individual pins can run well out of spec and function with a limited life time and limited environmental conditions with reduced margin but then so can most things. Its like saying that a car company is short changing you because a tuner can get an extra 200 HP out of an engine by removing all margins and only needing it to last 5k miles.
 
I've been in the electronics and IT biz for a very long time. ... "It's a bit more complicated than that.".


And I have to say that beating people up over their analogies is asinine.

.

Bragging about being an IT expert and bashing a video and expert overclocker while saying nothing more than "it's more complicated than that" is like being a mechanic defending a truck company for not allowing you to put the new tires on an older vehicle without giving any technical reasons.

How do you like that analogy?
 
Let me make it clear for everyone who lacks reading comprehension, just because you can take something away from a system and it still works doesn't mean it's right. If you take away spark, fuel, whatever from part a motor it will run like shit but still get you home. You can't disrupt a precisely engineered system and expect it to be right.


I'm inclined to believe a multi-billion engineering facility who invented the fucking product at question would have a bit more understanding of what is happening than some hack with a roll of tape, a couple spare hours with a video camera and a chip on his shoulder.


They're not saying you're wrong, per-se... Just that your analogy is absolutely terrible, lol.

Intel specs for the lowest common denominator, not enthusiast boards.

You can literally push 250w of 12v through a SINGLE gpu connector strand... Modern high end GPU's use 6 conductor pairs. For a reason. I'm sure everyone has run them off molex>6 pin adaptors, hell, even some of us have run them off Sata>6 pin adapters. Ignoring slot delivery, that means cards power supplies are rated for 1500w... Yet pull <300w during normal use.

Doesn't make it right, but the spec damn well sure needs to be rated for it. To an order of magnitute, conservatively...
 
They did load testing to prove the pins could handle the load without the additional ones. This makes it seem like Intel wired in a thermostat for a furnace with jumper cables.

Edit: nevermind, I'm way off base. Seems like an increase of 14% of pins, which doesn't sound unreasonable for doubling core count over Skylake. Unless Skylake was already way overkill for power delivery.

Everything is designed for way overkill power delivery because it is a critical link that forward flows into all the margins everywhere in the chip design. It affects di/dt curves, etc. All that leads to margin assumptions in power management, decap, transistor performance ratios, etc. All of which leads into FIT rates for the product. And those FIT rates leads into overall reliability for the system. Sure, you probably can run the pins at 5A, but then your pin FIT rates jump through the roof, not to mention all your di/dt margining goes into the toilet also increasing FIT.

And all this is based off of probabilities. The probability of a pad failing to contact in the socket, the probability of an arc, metal fatigue, etc.

Best way to think about it is, you probably can get away with replacing all your 15/20 amp breakers with 30/40 amp breakers in your house without going from 14/12 gauge wire to 10/8 gauge wire, but I doubt you'll be able to get your home owners insurance to cover you with it.
 
He first showed what the overclocked current would be on a Z-170:
View attachment 132623

He then ran 5 amps through a SINGLE PIN with no signs of damage.

View attachment 132625
Finally reduced the pins until he was down to 69 power pins were blocked and ran prime95 overnight while oveclocked. Going from Z-370 to Z-170 only reduces the the power pins by 18.

Except that doesn't actually prove anything. When he steps up to cover peoples warranties when they do it, maybe he'll have an argument.
 
All these car analogies are pretty awful, the fuel pump one almost made sense if you ignore that unlike a socket it has moving parts that wear and more importantly that he found the pins have a ton of overhead. I am curious how many of the people that seem to think he doesn't know what he's talking about have been consulted by MB manufacturers/engineers on designs like he has.


That being said this really isn't news as much as it's further confirmation of what we already knew. I don't mind upgrading MBs if there's a legit reason or for extra features and most of my MBs have never seen a second CPU but the ability to upgrade to future CPUs was a selling point for me on AM4, I wasn't sure if I would upgrade but if the 3000 series lives up to the rumors there's a high probability I will and it's nice having the option.
 
....Its like saying that a car company is short changing you because a tuner can get an extra 200 HP out of an engine by removing all margins and only needing it to last 5k miles.

Brilliant. Another spot on analogy by the IDF. I suppose increasing the current per pin by around 10% even though each pin can handle like 500% would have the same long term effect as running an extra 200 hp in some aftermarket tuned engine.
 
Bragging about being an IT expert and bashing a video and expert overclocker while saying nothing more than "it's more complicated than that" is like being a mechanic defending a truck company for not allowing you to put the new tires on an older vehicle without giving any technical reasons.

How do you like that analogy?

You do realize that being an overclocking "expert" really doesn't mean all that much, right? Its not like one needs an EE background to overclock. Overclocking is primarily about cash at the competitive levels. Christ, there a multitude of reasons that people don't put 5A though an LGA pad on any design in existence as spec. Like, you can put 40A though 18 gauge if you really want to, but I'd never recommend that anyone actually do it.
 
Back
Top