Coffee Lake has a reason not to run on Z270

[Suika]

(click for source)

 

In a world ravaged by people who actually thought Intel would support Coffee Lake on Z270 (since when has Intel let a chipset support 3 generations in recent years LOL), Gamers Nexus has come along and pointed out Intel's reason for not doing so. Z270 is appears to be very capable of supporting Coffee Lake, and Intel has had early engineering boards capable of supporting both Coffee Lake and Kaby Lake, but were later disabled via firmware updates. So, what gives?

Quote

[...] the Z370 chipset is more of a stop-gap solution for immediate launch, and that Z370 contains trace optimizations for power delivery that are necessary to make Coffee Lake work well. This is the real reason that Coffee Lake isn’t meant to be compatible with Z200 chipsets. We learned through multiple sources that some earlier engineering boards supported intercompatibility between the products, but that this was eventually disabled in firmware to ensure Coffee Lake is used on platforms which have had their power delivery optimized properly.

Which answers an earlier question I was curious about, had Intel launched Coffee Lake with compatibility on older chipsets, how would lower end boards handle the extra strain that will come with a processor that has two more cores than the initially highest end processor the board supported? Intel has made that answer simple; there's no guarantee it would have worked, so it's easier for manufacturers and consumers if they completely blocked off support.

 

Despite Intel's patterns with processor support between generations of CPUs and chipsets (Z270 had "dead by next gen" written all over it imo), some users will still feel betrayed that the Z270 chipset doesn't support Coffee Lake. A bit understandable, given high end boards could still likely support Coffee Lake at full performance, but I can't say this was ever unexpected.

 

Apologies if this was posted or discussed before. I couldn't find any earlier discussion of this in Tech News.

[GDRRiley]

it is sort of a re-post but with new sources. I am more mad that z170 doesn't work as it means if I want a 6 core got to change cpu+mobo (while mine is not great it handles my OCed i5 fine.)

[BachChain]

So rather than leaving it up to the motherboard manufacturer, Intel just assumes that all of them only meet the bare minimum spec

[RadiatingLight]

dumb as fuck.

maybe they could have said that H110 boards can't but H270, B150, B250, etc. and higher-end locked chipsets could support it.

 

and bullshit on the Z170 and Z270 not supporting it.

if they can support overclocking, then can definitely support a CPU with a TDP that's 4 watts higher.

[Suika]

6 minutes ago, huilun02 said:

As expected of Intel. AM4 will last longer. AMD isn't nearly as big of a shill. 

This is a continued pattern of Intel, and it's a legitimate concern on their part and doesn't blatantly appear to be a money grab. By the sounds of it, they considered intercompatibility, but lower end boards are probably a concern. 

 

I mean, If Intel is legitimately forcing 7700k owners into buying new boards, what's to stop them from considering AMD? Intel has to have thought this through.

4 minutes ago, RadiatingLight said:

if they can support overclocking, then can definitely support a CPU with a TDP that's 4 watts higher.

TDP isn't everything, and isn't exactly representative of how much power the CPU actually uses. I wouldn't be shocked to see actual power usage closer to Intel's HEDT chips. Also, the question isn't only in high end boards, it's also in lower end boards. I've worked with a couple boards that I wouldn't trust to run Coffee Lake at stock settings.

[RadiatingLight]

1 minute ago, Suika said:

TDP isn't everything, and isn't exactly representative of how much power the CPU actually uses. I wouldn't be shocked to see actual power usage closer to Intel's HEDT chips. Also, the question isn't only in high end boards, it's also in lower end boards. I've worked with a couple boards that I wouldn't trust to run Coffee Lake at stock settings.

TDP relates to average power consumption, and even if the consumption was 30W higher, then it still would be fine on literally every Z-series board.

if anything, then at least intel should've let Z-boards support CL

[WereCat]

Good guy Intel doesn't want people to accidentally destroy the VRM on their overclocking Z170/270 motherboards and causing unexpected expenses so they created expected expenses by forcing people to use new Z370 which will literally be like Z270 but with newer BIOS and higher number in its name.

[Suika]

1 minute ago, RadiatingLight said:

TDP relates to average power consumption

"TDP is not the maximum power that the processor can dissipate. TDP is measured at maximum TCASE.1." - Intel

"Tcase: Temperature measurement using a thermocouple embedded in the center of the heat spreader." - Also Intel

4 minutes ago, RadiatingLight said:

if anything, then at least intel should've let Z-boards support CL

The sketchy boards I worked with are Z series boards. Intel still would have taken a risk allowing processors to run on motherboards that initially only supported at max a CPU with 2 less cores than Coffee Lake is offering.

[mr moose]

6 minutes ago, RadiatingLight said:

TDP relates to average power consumption, and even if the consumption was 30W higher, then it still would be fine on literally every Z-series board.

if anything, then at least intel should've let Z-boards support CL

TDP relates to power consumption in the same way radiator size in a car relates to fuel economy.

 

It doesn't.  It doesn't even indicate how much power the CPU will draw at stock.  In fact the only thing TDPis good for is designing a cooler for a specific CPU usage. 

 

 

[Master Disaster]

Hmmmmm, so that one guy who swore blind Intel had to change the socket layout to accommodate CLs extra cores and updated memory support was talking out of his ass.

 

I knew this was the case the second they announced LGA1151, it's like I said in the other thread, Intel WOULD have called it LGA1151-V2 (or something similar) had it been a different and electrically incompatible socket but it was nothing but Intel being Intel as usual.

 

So we now know the CL could have worked with Z170/270 and that SL/KL could have worked with Z370. Given its rumoured Ice Lake will be a new socket entirely anyone who buys into CL needs their head examining (imo).

[Dabombinable]

I wonder if custom firmware would enable compatibility? Its not like all Z270 boards have power delivery that couldn't handle Coffee Lake.

5 minutes ago, Master Disaster said:

Hmmmmm, so that one guy who swore blind Intel had to change the socket layout to accommodate CLs extra cores and updated memory support was talking out of his ass.

 

I knew this was the case the second the announced LGA1151, it's like I said in the other thread, Intel WOULD have called it LGA1151-V2 (or something similar) had it been a different and electrically incompatible socket but it was nothing but Intel being Intel as usual.

 

So we now know the CL could have worked with Z170/270 and that SL/KL could have worked with Z370. Given its rumoured Ice Lake will be a new socket entirely anyone who buys into CL needs their head examining (imo).

Especially considering that the current i7 7700K is still the best CPU for gaming, and that those who need more than 8 threads can get a far better deal with Ryzen (god AMD is really putting the screws on Intel)

[Suika]

5 minutes ago, Master Disaster said:

So we now know the CL could have worked with Z170/270 and that SL/KL could have worked with Z370.

Maybe, but there's no guarantee that every Z170/270 board would have worked with Coffee Lake, and it'd add a whole mess of confusion on the consumer side. There are some really shitty, lowend Z170/270 boards out there that probably don't even account for overclocking, let alone throwing in two more cores.

[RadiatingLight]

11 minutes ago, Suika said:

"TDP is not the maximum power that the processor can dissipate. TDP is measured at maximum TCASE.1." - Intel

"Tcase: Temperature measurement using a thermocouple embedded in the center of the heat spreader." - Also Intel

The sketchy boards I worked with are Z series boards. Intel still would have taken a risk allowing processors to run on motherboards that initially only supported at max a CPU with 2 less cores than Coffee Lake is offering.

 

9 minutes ago, mr moose said:

TDP relates to power consumption in the same way radiator size in a car relates to fuel economy.

 

It doesn't.  It doesn't even indicate how much power the CPU will draw at stock.  In fact the only thing TDPis good for is designing a cooler for a specific CPU usage. 

 

 

 

 

but where does the heat come from?

electricity!

 

all electricity that goes into a CPU will turn into energy, and no more.

 

if a CPU outputs 500W of heat, then it has consumed exactly 500W of electricity to do so. no more, no less. energy is not (and cannot) be used up in the process of calculating, it simply transitions to heat.

 

[Prox1ma]

4 minutes ago, RadiatingLight said:

if a CPU outputs 500W of heat, then it has consumed exactly 500W of electricity to do so. no more, no less.

 

Wouldn't that mean its an ideal system? Is everything turned to heat? I didn't think that such a system was possible....I thought to generate 500W of heat, it must have needed more than that to generate it in the first place. How much more, that I do not know.

[thorhammerz]

13 minutes ago, Ezilkannan said:

Wouldn't that mean its an ideal system? Is everything turned to heat? I didn't think that such a system was possible....I thought to generate 500W of heat, it must have needed more than that to generate it in the first place. How much more, that I do not know.

Actually, it'd be more like a "perfectly useless CPU" .

 

Heat is generated by a CPU when there is electrical resistance occurring at the transistors (in between the ON and OFF state a transistor acts as a resistor, preventing current from flowing, thus dissipating the energy as heat).

 

If a CPU converted all its incoming power into heat (as opposed to work, which is getting the transistor to switch from ON to OFF, or vice-versa), that would mean nothing actually gets done (and thus it would be fit for only making toast) .

 

There's a good reason why Intel spends so much R&D money on making more "power efficient" CPUs (especially for mobile): getting the same amount of work done for less power, which is achieved through the reduction of power loss via thermal radiation.

[Suika]

13 minutes ago, RadiatingLight said:

but where does the heat come from?

electricity!

 

all electricity that goes into a CPU will turn into energy, and no more.

 

if a CPU outputs 500W of heat, then it has consumed exactly 500W of electricity to do so. no more, no less. energy is not (and cannot) be used up in the process of calculating, it simply transitions to heat.

Here you are, appearing to have no faith in Intel's word on Coffee Lake's compatibility in Z270, putting a lot of faith in Intel's TDP measuring process, assuming they don't just run processors at base clock (which the 8700k has a lower base clock).

 

I also highly doubt that Intel can, in a single generation, take a 6 core, 12 thread chip, raise base and boost clocks significantly, and shave TDP from 140W to 95W. Something in me just doesn't trust that, and I'm not sure why nobody else is seeing that.

[Prox1ma]

4 minutes ago, thorhammerz said:

Actually, it'd be more like a "perfectly useless CPU" .

 

Heat is generated by a CPU when there is electrical resistance occurring at the transistors (in between the ON and OFF state a transistor acts as a resistor, preventing current from flowing, thus dissipating the energy as heat).

 

If a CPU converted all its incoming power into heat (as opposed to work, which is getting the transistor to switch from ON to OFF, or vice-versa), that would mean nothing actually gets done (and thus it would be fit for only making toast) .

That was what I was I thinking

[thorhammerz]

12 minutes ago, Suika said:

I also highly doubt that Intel can, in a single generation, take a 6 core, 12 thread chip, raise base and boost clocks significantly, and shave TDP from 140W to 95W. Something in me just doesn't trust that, and I'm not sure why nobody else is seeing that.

Lowering the base clock to 3.7GHz does count as "saving power" .

 

Rather, pity the poor S.O.B. who tries to "sync all core clocks" to the Turbo Boost rating of 4.7 GHz without a delid job.

[RadiatingLight]

4 minutes ago, thorhammerz said:

Actually, it'd be more like a "perfectly useless CPU" .

 

Heat is generated by a CPU when there is electrical resistance occurring at the transistors (in between the ON and OFF state a transistor acts as a resistor, preventing current from flowing, thus dissipating the energy as heat).

 

If a CPU converted all its incoming power into heat (as opposed to work, which is getting the transistor to switch from ON to OFF, or vice-versa), that would mean nothing actually gets done (and thus it would be fit for only making toast) .

 

14 minutes ago, Ezilkannan said:

Wouldn't that mean its an ideal system? Is everything turned to heat? I didn't think that such a system was possible....I thought to generate 500W of heat, it must have needed more than that to generate it in the first place. How much more, that I do not know.

I hate this misconception

 

work is not energy. energy gets converted directly to heat, and the by-product of that conversion is heat.

thermodymanics says that energy cannot be created nor destroyed, and work is not energy, so the energy must either go to heat, or to light, or to sound/movement. CPUs don't move or emit light, so...

[mr moose]

20 minutes ago, RadiatingLight said:

 

 

 

but where does the heat come from?

electricity!

 

all electricity that goes into a CPU will turn into energy, and no more.

 

if a CPU outputs 500W of heat, then it has consumed exactly 500W of electricity to do so. no more, no less. energy is not (and cannot) be used up in the process of calculating, it simply transitions to heat.

 

 

Electrical resistance, and resistance is only part of the equation. The CPU also drives all it's outputs (data lines for PCIe, Ram, etc), so a percentage of the power consumed by the CPU is passed out through the data lines, some is converted to heat and of that some is retained in the CPU while the rest needs to be dissipated. This is why TDP is not a good figure when discussing power draw because TDP only relates to a requirement to maintain a specific Tj. 

 

This is like the third thread in the last 24 hours where TDP has been brought up and used incorrectly.  Seriously @LinusTech. A techquickie please on TDP.

[thorhammerz]

5 minutes ago, RadiatingLight said:

 

I hate this misconception

 

work is not energy. energy gets converted directly to heat, and the by-product of that conversion is heat.

thermodymanics says that energy cannot be created nor destroyed, and work is not energy, so the energy must either go to heat, or to light, or to sound/movement. CPUs don't move or emit light, so...

Last I checked, converting a resistor from an ON state to an OFF state (or vice versa) technically counts as movement (even if it's on a nano scale of just a bunch of atoms/electrons... multiplied by however many billion are on a single die ).

[Prox1ma]

2 minutes ago, RadiatingLight said:

 

I hate this misconception

 

work is not energy. energy gets converted directly to heat, and the by-product of that conversion is heat.

thermodymanics says that energy cannot be created nor destroyed, and work is not energy, so the energy must either go to heat, or to light, or to sound/movement. CPUs don't move or emit light, so...

Then again, you are implying that 100% of the energy that goes into the system is converted as heat. That is a very ideal heat conversion system there. I did not know that microprocessors are ideal heat conversion systems.

Quote from wikipedia:
 

Quote

(CPUs) consume electrical energy, and dissipate this energy both by the action of the switching devices contained in the CPU (such as transistors or vacuum tubes) and by the energy lost in the form of heat due to the impedance of the electronic circuits.

 

[Suika]

6 minutes ago, VegetableStu said:

Although to be fair I think it's new information enough to make this thread

Yea, it's mostly just a clarification as to why Coffee Lake won't be compatible on Z270, but it seems a lot of users still don't trust Intel's judgement on this decision because we're all engineers.

[juretrn]

If Intel was fair, they'd certify "good" Z270 boards for CL compatibility.

1 hour ago, thorhammerz said:

Last I checked, converting a resistor from an ON state to an OFF state (or vice versa) technically counts as movement (even if it's on a nano scale of just a bunch of atoms/electrons... multiplied by however many billion are on a single die ).

*transistor

But that's mostly wrong. Electrons moving around isn't much of a "movement" - and it's never atoms in semiconductors that move. The currents involved in these processes are negligible, at east for the switching operation itself.

[Master Disaster]

2 hours ago, Suika said:

Maybe, but there's no guarantee that every Z170/270 board would have worked with Coffee Lake, and it'd add a whole mess of confusion on the consumer side. There are some really shitty, lowend Z170/270 boards out there that probably don't even account for overclocking, let alone throwing in two more cores.

So then allow manufacturers to add support for CL but mandate that OCing support be disabled. That way users on older platforms can still upgrade but they miss out on a feature unless they buy a 370 board.

 

Looking at the whole Skylake X & Kabylake X situation I'm inclined to believe that Intel doesn't care about consumer confusion at all.