Finer Details of Intel Core i7-9700K and Core i9-9900K Emerge

[btarunr]

Taiwanese tech site BenchLife.info scored finer details of Intel's upcoming premium LGA1151 processors through screenshots of leaked documents; revealing more about the Core i7-9700K 8-core/8-thread processor, and the top-dog 8-core/16-thread Core i9-9900K. The i7-9700K has the QDF number QQPK, and the i9-9900K "QQPP." The tables below also reveal their extended product code, CPUID, and iGPU device ID. There's also a confirmation that the TDP of both parts is rated at just 95 W. The next table provides a great insight to the clock speeds of the two chips.

Both chips idle at 800 MHz, and have an identical nominal clock speed of 3.60 GHz. The two differ with their Turbo Boost states. The i7-9700K has a maximum Turbo Boost state of 4.90 GHz, which it awards to 1-core. As a reminder, this chip is the first Core i7 SKU ever to lack HyperThreading support. 2-core boost frequency for this chip is 4.80 GHz. 4-core boost is up to 4.70 GHz. 4.60 GHz is the all-core boost (cores 5 thru 8). The i9-9900K gives both 1-core and 2-core the highest boost frequency of 5.00 GHz (that's up to 4 threads). The 4-core boost state is 4.80 GHz, and all-core (cores 5 thru 8) get 4.70 GHz. Intel is keeping its boost states rather high for this round of processors, as it tries to compete with the Ryzen 7 "Pinnacle Ridge" series.



View at TechPowerUp Main Site

 

[Caring1]

The big question is, will it compete as far as pricing!

 

[HimymCZe]

The big question is, will it compete as far as pricing!

it IS expected that Intel will charge 50% premium as a tax for AMD combo-breaking their 11 years streak of reselling same Quad-cores...

 

[rtwjunkie]

Running all 8 cores and thus 16 threads at 4.7Ghz with no effort is actually very good. There isn’t much the normal consumer runs that will even tax all of that.

 

[Assimilator]

They should've given the 9900K the QQMOAR moniker.

 

[Prima.Vera]

I wonder how much is that Wattage at full 5Ghz on both cores...

 

[rtwjunkie]

I wonder how much is that Wattage at full 5Ghz on both cores...

Yeah, because we know they don’t test 95w on an 8 core 4.7Ghz run. I bet with 5 you are looking at close to 200w TDP. That’s still within manageable by my air cooler though.

 

[windwhirl]

Finally, Intel is starting to move.

On another note, couldn't BenchLife use a slightly less glaring "watermark" (more like a stamp) than neon green? Honestly, it's somewhat annoying...

 

[wurschti]

Looks great and everything, but all these security vulnerabilities makes it hard to consider, not only these CPUs, but Intel in general. They have to make a solid and secure CPU, then can they start building fast ones imo.

 

[phill]

Interesting times for CPU's at the moment Be interesting to see how these compare

 

[bug]

I might have been tempted to try one of these, but Intel made sure I won't when they made them incompatible with my Z170 board.

 

[Valantar]

Yeah, because we know they don’t test 95w on an 8 core 4.7Ghz run. I bet with 5 you are looking at close to 200w TDP. That’s still within manageable by my air cooler though.

Intel rates TDP for base clocks only, so 95W is for 3.6GHz at all cores (although if history is anything to go by, actual power draw at those clocks might be slightly lower than 95W for most chips, as even the worst-performing chips need to meet TDP at base clocks). Still, a 1.1GHz bump over base clocks on 8 cores should indeed mean that all-core turbo draws a heck of a lot of power. I bet you'd need a particularly heavy-duty cooler to actually maintain these clocks in the first place. Considering ~5GHz CFL 6-cores often pull 200W or more, I'd be surprised if they managed to add two cores and stay beneath that number just by downclocking a few hundred MHz. Likely 4.7GHz is as far as they can push all-core turbo and get by passably with a bog-standard 240mm AIO.

The real surprise here will be whether these are soldered or not. Considering der8auer's very strong hints a while ago, I'm leaning towards thinking Intel aren't quite dumb enough to smear these with their bird-poop TIM.

 

[Berfs1]

I wonder how much is that Wattage at full 5Ghz on both cores...

I’d estimate about 40-60W, since it’s just two cores (assuming you enable only two cores).

Intel rates TDP for base clocks only, so 95W is for 3.6GHz at all cores (although if history is anything to go by, actual power draw at those clocks might be slightly lower than 95W for most chips, as even the worst-performing chips need to meet TDP at base clocks). Still, a 1.1GHz bump over base clocks on 8 cores should indeed mean that all-core turbo draws a heck of a lot of power. I bet you'd need a particularly heavy-duty cooler to actually maintain these clocks in the first place. Considering ~5GHz CFL 6-cores often pull 200W or more, I'd be surprised if they managed to add two cores and stay beneath that number just by downclocking a few hundred MHz. Likely 4.7GHz is as far as they can push all-core turbo and get by passably with a bog-standard 240mm AIO.

The real surprise here will be whether these are soldered or not. Considering der8auer's very strong hints a while ago, I'm leaning towards thinking Intel aren't quite dumb enough to smear these with their bird-poop TIM.

Or maybe a lot of overclocks were not performed by seasoned overclockers? Putting down 1.4V Vcore and slapping 5 GHz isn’t that impressive. The reason they pull more than 200W is because a lot of people have the time to put 1.4V but not the time to test whether or not that was way too high (1.35V is expected for 5GHz, with a .05V variance)

 

[PanicLake]

So the i9 9900k are just binned i7 7900k...

 

[dj-electric]

it IS expected that Intel will charge 50% premium as a tax for AMD combo-breaking their 11 years streak of reselling same Quad-cores...

Implying Q6600 and 7700K are the same quad-cores. Cars have 4 wheels. Time for the car industry to advance as well.

I might have been tempted to try one of these, but Intel made sure I won't when they made them incompatible with my Z170 board.

"Hack" your BIOS, just don't expect a great result running X2 the power consumption if you want to OC.

So the i9 9900k are just binned i7 7900k...

Aren't all downgraded chips of any kind the lesser binned ones essentially?

 

[mouacyk]

So the i9 9900k are just binned i7 7900k...

Not if the communications bus has anything to say about that.

 

[bug]

"Hack" your BIOS, just don't expect a great result running X2 the power consumption if you want to OC.

Neah, I'll just spend my money elsewhere. I don't actually need that many cores, it would have just been tempting.

 

[HimymCZe]

Implying Q6600 and 7700K are the same quad-cores. Cars have 4 wheels. Time for the car industry to advance as well.

..., but they did? Or you're saying the best engine in 2018 car is 177 C.I.D. (2.9 L) 20 hp I4 (Ford Model T engine fyi.)?
because, that's EXACTLY how it feels to buy Intel nowadays. You are STUCK with 100 years old, minorly polished old tech, and NO ONE is progressing us into V6, V8, higher horse power or better fuel efficiency and that is all right ???

 

[Valantar]

I’d estimate about 40-60W, since it’s just two cores (assuming you enable only two cores).

The ring bus still needs to be active, and given that it'll be bigger than it is for CFL, power draw will be higher for the uncore regardless of disabled cores. Also, power-binned KBL-R and CFL U-series chips pull more than 50w when turboing to 4.2-ish on two cores (with the other two active, of course). Achieving no increase in power draw on a bigger chip with a bigger uncore at 0.8GHz up from that, even with disabled cores, seems wildly optimistic. Even if they manage to eke out a 10% power improvement from their next tweaked node (which is by itself extremely unlikely), that sounds far, far too low.

Or maybe a lot of overclocks were not performed by seasoned overclockers? Putting down 1.4V Vcore and slapping 5 GHz isn’t that impressive. The reason they pull more than 200W is because a lot of people have the time to put 1.4V but not the time to test whether or not that was way too high (1.35V is expected for 5GHz, with a .05V variance)

Sure, most consumer OCs could probably be tuned quite a bit more. But getting even a majority of CFL chips to 5GHz all-core below 1.4V is really unlikely. Golden samples run 5.4GHz at 24/7 voltages, while even most 8086Ks (all of which are top-binned silicon, limited to 50 000 units) top out between 5.2-5.3. A minority of these seem to go above 5.1 below 1.4V. And you expect a majority non-top-bin silicon to reach 5 at 1.35? Sure, it can probably be done by, as you say, seasoned OCers who know exactly how to balance on that knife edge. That is a pretty low bar, though, and definitely not indicative of possible binning for mass-produced chips that aren't hand tuned,even if the 4th-gen 14nm process is significantly improved.

So the i9 9900k are just binned i7 7900k...

Only if you disregard... well, the entire design of the chip. Some hints: ring bus vs mesh interconnect, cache size and layout, binned) LCC die.

 

[R0H1T]

Neah, I'll just spend my money elsewhere. I don't actually need that many cores, it would have just been tempting.

The force welcomes you. You don't want that many either?

Only if you disregard... well, the entire design of the chip. Some hints: ring bus vs mesh interconnect, cache size and layout, binned) LCC die.

Except there is no i7 7900K, possibly a typo & that he meant i7 9700k (with HT)

 

[bug]

The force welcomes you.
You don't want that many either?

Nope. I rarely see my current four being 100% in use.

 

[Valantar]

Except there is no i7 7900K, possibly a typo & that he meant i7 9700k (with HT)

Typo? Yep. It's far more likely that what was intended was the i9 7900X (which I (with an uncanny ability to remember model numbers and names) couldn't remember the correct name for either) though, as your alternative is so obvious as to be utter nonsense. What would it otherwise be, a bespoke piece of silicon for a single SKU? Yeah, that doesn't happen outside of smartphones and AMD's semi-custom business.

..., but they did? Or you're saying the best engine in 2018 car is 177 C.I.D. (2.9 L) 20 hp I4 (Ford Model T engine fyi.)?
because, that's EXACTLY how it feels to buy Intel nowadays. You are STUCK with 100 years old, minorly polished old tech, and NO ONE is progressing us into V6, V8, higher horse power or better fuel efficiency and that is all right ???

Well, considering the 7700K was more than 3x faster than the fastest Core2Quad (I should know, I ran an OC'd C2Q 9450 @3.52GHz until last year), that statement is pretty far off. Intel can absolutely be blamed for intentionally reserving increased core counts to the HEDT market, but claims like that just undermine the entire debate. Did Intel screw over users? Sure. Did they not improve at all? Of course they did. Given that CFL must have been in the works at least a year before the launch of Ryzen (likely more), it's more or less natural to view it as a natural extention of the power and efficiency improvements over the last few generations. Should they have brought >4t CPUs into the mainstream a few years ago? Absolutely. But that doesn't mean they've been peddling 2008 tech for a decade.

 

[Hood]

i7 7900K

He's referring to the skylake-X, i9-7900X

 

[R0H1T]

He's referring to the skylake-X, i9-7900X

Possibly, can't be 100% certain unless the poster clarifies it himself.

 

[bug]

Meh, who cares about these at the same time the Threadripper refresh is upon us?