First Comprehensive Review of the Core i7-11700K (ES) Surfaces

[btarunr]

Lab501 posted the first comprehensive review of an Intel Core i7-11700K "Rocket Lake-S" engineering sample. The ES has clock speeds matching the rumored clock speeds of the retail version, and should give you a fair idea of how the finished product should perform. The i7-11700K, which is an 8-core/16-thread chip, was tested to be being consistently behind the AMD Ryzen 7 5800X in synthetics such as WPrime, rendering tests such as Blender, video-encoding tests such as Handbrake, and was negligibly trading blows with the 5800X at gaming ±1%. The chip does post leads over the previous-gen i7-10700K in all these areas, though.

Performance aside, the Core i7-11700K is shown to have significantly higher power draw, with the whole-system power draw being 27% higher than a 5800X-based whole-system, when measured using Prime95 (which only adds a CPU load). In a real-world scenario such as gaming, where GPU power draw is added, this whole-system power draw percentage difference should come down. Interestingly, the i7-11700K isn't a "hot" processor, running up to 18°C cooler than a 5800X under Prime95 load. Check out this, and other invaluable early insights into "Rocket Lake" by hitting the source link below.



View at TechPowerUp Main Site

 

[xenocide]

Something doesn't add up here. How does it use way more power, run way cooler, and perform just about exactly the same as the previous gen i7?

 

[tabascosauz]

Something doesn't add up here. How does it use way more power, run way cooler, and perform just about exactly the same as the previous gen i7?

Indicated CPU temperature =! heat output =! power draw

It's a little late to be beating the dead horse that is "but it's Intel, how can it run cooler??". Comet Lake released months ago with die thinning and a new heatspreader, none of this is news. Intel has the monstrous power draw without the thermal density, AMD has the efficiency without the die area to properly dissipate it.

One thing is for sure, Superfin and Golden Cove needs to come to desktop ASAP, because the power draw is going to keep driving VRM throughput and already insane board prices up until Intel can get a handle on its PL2 numbers.

 

[xenocide]

Indicated CPU temperature =! heat output =! power draw

It's a little late to be beating the dead horse that is "but it's Intel, how can it run cooler??". Comet Lake released months ago with die thinning and a new heatspreader, none of this is news. Intel has the monstrous power draw without the thermal density, AMD has the efficiency without the die area to properly dissipate it.

I misread. I thought the comparison was to the 10700k, not the 5800X, hence my confusion. If Intel managed to sidegrade the 10700k to the 11700k by having it use way more power and run way cooler I would be very confused.

 

[Searing]

Indicated CPU temperature =! heat output =! power draw

It's a little late to be beating the dead horse that is "but it's Intel, how can it run cooler??". Comet Lake released months ago with die thinning and a new heatspreader, none of this is news. Intel has the monstrous power draw without the thermal density, AMD has the efficiency without the die area to properly dissipate it.

One thing is for sure, Superfin and Golden Cove needs to come to desktop ASAP, because the power draw is going to keep driving VRM throughput and already insane board prices up until Intel can get a handle on its PL2 numbers.

I keep on seeing people post things like this, yet I have a 10600k and 5600x and the 5600x consumes less power and is way cooler. I don't find AMD's temps very high. I find this "18 degree difference" totally suspect. Not the same cooler and fan speed imo.

 

[kapone32]

I keep on seeing people post things like this, yet I have a 10600k and 5600x and the 5600x consumes less power and is way cooler. I don't find AMD's temps very high. I find this "18 degree difference" totally suspect. Not the same cooler and fan speed imo.

Probably a Wraith Prism or some such lol.

 

[tabascosauz]

I keep on seeing people post things like this, yet I have a 10600k and 5600x and the 5600x consumes less power and is way cooler. I don't find AMD's temps very high. I find this "18 degree difference" totally suspect. Not the same cooler and fan speed imo.

You don't have a 5800X. Stock 5800X says hi. 5800X is the comparison here. Lol

5600X is the one chip in the stack that runs uncharacteristically cool for various reasons, one of which is that it's actually closer to its TDP than PPT.

Although, in a sense, it's a bit of a strange comparison and unfair in a way since the 5800X is the uncharacteristically hot one in the stack.

 

[Cobain]

Seems very good. 20fps more than 10900k in some games, and that's an ES without any overclock.

Can only imagine its performance with 4133mhz c16 ram, it Will beat zen 3 In games for sure.

If 11600k comes at 250€-280€, and with stock, we have a winner for gaming.

11400f for 160€-180€ Will also be an interesting budget option paired with a B560 motherboard.

 

[Logoffon]

There's a comma at the end of the title. Was there supposed to be anything after that?

 

[1d10t]

Finally, Intel show what everything should be, 6 core musn't faster than 8 core.

 

[watzupken]

In my opinion, the difference in temps should not come as a surprise when comparing Comet Lake, Rocket Lake against the Ryzen 7 5800X. The Intel chips have monolithic die, which means there is considerably more surface area to dissipate heat to the IHS and to the heatsink. The Ryzen chips are utilizing chiplets which is likely to be more than 2x smaller than the Intel die. Having said that, it is possible to tweak the chip using curve optimizer that will reduce heat and potentially maintain or inprove performance.

As these chip companies start moving into chiplet design, I feel heat issues will become more common.

 

[yeeeeman]

Something is not right in MT...

 

[Cobain]

In my opinion, the difference in temps should not come as a surprise when comparing Comet Lake, Rocket Lake against the Ryzen 7 5800X. The Intel chips have monolithic die, which means there is considerably more surface area to dissipate heat to the IHS and to the heatsink. The Ryzen chips are utilizing chiplets which is likely to be more than 2x smaller than the Intel die. Having said that, it is possible to tweak the chip using curve optimizer that will reduce heat and potentially maintain or inprove performance.

As these chip companies start moving into chiplet design, I feel heat issues will become more common.

Specially 5800x, because it is a really hot CPU.

 

[InVasMani]

The lower CPU temps should mean pushing the IMC harder might be very feasible by extension potentially it's hard to say because the tcase value isn't particularly high at 72° and from what I can tell tcase is like a stability crossover threshold point for a chip itself. I always have instability soon as my CPU exceeds tcase it's not always instant though I guess that's due to current leakage of the silicone lottery grade perhaps.

More on that subject as you push the IMC harder at a given frequency and reducing CL by 1 or conversely at a given CL speed you using a higher memory frequency strap it raises CPU temps in turn quite progressively. The end result of course is a lower tcase chip will hit a IMC system memory bandwidth/latency limitation more rapidly far as I can tell. Which considering how involved cooling the FSB was when the IMC was on the chipset and not integrated on the CPU itself that certainly makes logical sense. I actually think Intel should remove the IMC on it's big.LITTLE approach for the Core chip die (RIP) and retain it on the Atom chip that they place side by side on substrate in the same socket. The Atom chip die should run cooler so the IMC could be pushed more aggressively without causing as much CPU temperature rise in the process when you use a higher memory frequency strap on the IMC or if you reduce the CL latency at the same frequency strap.

I think there are pro's and con's to that approach, but if it means a higher memory frequency strap you end up with higher bandwidth which gives you more true latency upside than reducing CL latency so it balances out a bit. The Atom die would be closer to the Core die in the first place as well on the same substrate on the same socket much more so than in the old days where the memory controller resided on the NB MILES AWAY literally not figuratively from the CPU socket. The latency penalty wouldn't be anywhere close to as pronounced as it was with the IMC on the NB in reality. The other aspect is any tasks run by the Atom chip itself would suffer any latency drawback in the first place. The big thing is that the Atom chip die should run cooler which should make moving the IMC to that chip die beneficial in terms of dealing with cooling.

The other option is Intel could put 1 channel of bandwidth on each chip die to spread the heat output from the IMC between them in some sort of variable rate memory channel configuration between both dies where one channel runs quicker than the other channel, but can still access the other. You might be able to push each IMC on each die chip higher though if their each running a single channel however much like on a motherboard where if you populate more DIMM slot's pushing peak frequency scaling drops a bit for additional DIMM's.

In my opinion, the difference in temps should not come as a surprise when comparing Comet Lake, Rocket Lake against the Ryzen 7 5800X. The Intel chips have monolithic die, which means there is considerably more surface area to dissipate heat to the IHS and to the heatsink. The Ryzen chips are utilizing chiplets which is likely to be more than 2x smaller than the Intel die. Having said that, it is possible to tweak the chip using curve optimizer that will reduce heat and potentially maintain or inprove performance.

As these chip companies start moving into chiplet design, I feel heat issues will become more common.

I have the solution to that issue mix and match die sizes and scale the heat and performance of them around the die size, but arrange them on a substrate a bit like mipmaps for texturing.

 

[Vego]

who with this amount of money cares about performance in 1080p
give me CPU that wil boost performance in 4k on 144hz panel and ill consider changing my 5950x thats boosts to 4,85-5,09 with top temps of 53'C on my watercooling

 

[Dredi]

Something doesn't add up here. How does it use way more power, run way cooler, and perform just about exactly the same as the previous gen i7?

The temperature reporting between manufacturers or even between product generations are not comparable. The probe locations and calibration curves can be modified to fit any need and there is no easy way for a thrid party to test the accuracy of the measurement.

 

[ZoneDymo]

who with this amount of money cares about performance in 1080p
give me CPU that wil boost performance in 4k on 144hz panel and ill consider changing my 5950x thats boosts to 4,85-5,09 with top temps of 53'C on my watercooling

I mean...at that res you are gpu limited/bound so until we have much faster versions of those you wont see much influence from the cpu.

On the actual topic:

yikez....I was hoping for a bit more there Intel...makes it even more logical to just wait for Alderlake, course in the end, it all comes down to price

 

[InVasMani]

who with this amount of money cares about performance in 1080p
give me CPU that wil boost performance in 4k on 144hz panel and ill consider changing my 5950x thats boosts to 4,85-5,09 with top temps of 53'C on my watercooling

GPU bottleneck CPU bottleneck learn the difference.

 

[Caring1]

So the 5800X wins all around and uses much less power, good job Intel, that showed them.

 

[Bruno Vieira]

Something doesn't add up here. How does it use way more power, run way cooler, and perform just about exactly the same as the previous gen i7?

7nm way denser, so, way harder to cool, even conuming that much less power. You can OC an FX-8350 to the moon conuming 300W and yet is very easy to cool. But a r5 3600 at 4.4ghz consuming 80w is VERY hard to cool, you need to lap the ihs and put very good thermal paste / lm to keep it cool.

 

[AnarchoPrimitiv]

who with this amount of money cares about performance in 1080p
give me CPU that wil boost performance in 4k on 144hz panel and ill consider changing my 5950x thats boosts to 4,85-5,09 with top temps of 53'C on my watercooling

It's the "Walmart Effect"...basically, Wal-Mart sells a handful of items at the lowest prices (mostly cheap items that no one REALLY wants), but it has the psychological effect of making people think "if this is cheapest price on this item, then every item in the store must be the cheapest price too"...when in reality, those other items are more expensive than with competitors, but it's already too late, and the consumers have convinced themselves otherwise....the same thing for CPUs...all intel has to do is convince some fools that being fastest by 5% at 1080p with a $1500 videocard (honestly, I'd bet the percentage of actual gamers out there with a 3080/3090 who game at 1080p with an ultra high refresh monitor is probably less than 1% of the demographic and that a 5% difference, e.g. 132fps vs 138fps, is completely indistinguishable by basically every human being living, but I digress) means that they're the fastest at everything else...even if the consumer doesn't have that $1500 videocard, or an ultra high refresh monitor, or games at 1080p...this is marketing 101.

 

[Aretak]

Another Massive Disappointment from Intel it seems. Guess it's time for the fanboys to start shilling Alder Lake as the second coming.

 

[THU31]

Something doesn't add up here. How does it use way more power, run way cooler, and perform just about exactly the same as the previous gen i7?

I am sure they put liquid metal under that heat spreader, while having the thinnest die possible.

 

[DeathtoGnomes]

negligibly trading blows with the 5800X at gaming ±1%

Somewhere here, in reference to an old Intel PR thread, is an "I told you so" coming because it boasted double digit gains over AMD..

 

[RedelZaVedno]

R20 single core score of 608 is impressive (vs 514 of 10700K), on pair with 5800X, but it will beat it in gaming because most PC games are still better optimized for Intel CPUs. It makes sense since Intel still owns 82% of laptop/desktop market share vs AMDs 12%. Don't get me wrong, Zen3 is awesome arch+node and will only get better, but software optimization is still it's Achilles heel (though less and less). AMD needs to get at least third of laptop market share to get the same level software optimization support as Intel does. It's just the way it is atm.

There is no hardware reason for AMD CPUs to be so much slower in sims like MS Flight Simulator 2020, DCS World, IL2 for example, in fact they should be faster based on benchmarks than intel counterparts, but they're not. It all comes down to game code optimization in the end. No matter how good hardware potential is, it still comes down to software utilization of it. Same goes with AMD's RDNA/RDNA2 and some productivity apps support like Adobe Creative Suite unfortunately. 94% of Adobe users use Nvidia so they don't bother with AMD GPUs optimization.

Btw, I'd opt for 11700F instead of 11700K if I were not waiting for Alder Lake / Zen4 to upgrade. It will have much better power consumption profile and still be within 5% 11700K/5800X performance in low res gaming. Even 10700F is very tempting at current 256 bucks price tag versus hard to get 5800X with hard to swallow $448 price tag. Sure it's a bit slower, but looking at the price difference, it's a no brainer what to buy if you you're looking for optimal price/performance ratio imho.