Thursday, January 28th 2021

At Stock All-Core Boost, i9-11900KF "Rocket Lake" Hits 98°C with 360mm AIO CLC Under Stress

Intel's 11th Gen Core "Rocket Lake-S" processor is shaping up to be one hot chip. Enthusiasts on Chinese tech forum ChipHell with access to a Core i9-11900KF sample report, that when stressed at [apparently] stock settings, in which the processor hits its all-core boost frequency of 4.80 GHz, even a 360 mm all-in-one liquid CPU cooler struggles to keep the chip from hitting 100 °C, with core temperatures reaching 98 °C. The i9-11900KF comes with a disabled iGPU, which means all of its heat is generated by the core and uncore components of the GPU. AIDA64 was used, to apply multi-threaded stress on the processor (the burn-in test), while CPU-Z reports a core voltage of 1.401 V, it's not known whether this is a manual setting, the chip's VID, or whether the motherboard is trying to stabilize the 4.80 GHz clocks.
Sources: 9550pro (Twitter), ChipHell Forums, VideoCardz
Add your own comment

91 Comments on At Stock All-Core Boost, i9-11900KF "Rocket Lake" Hits 98°C with 360mm AIO CLC Under Stress

#51
ThrashZone
qubitThat's horrendous.
I wonder if it runs anywhere near that hot while doing regular gaming. And it's not even overclocked.

I want something that runs much like my current 2700K, reasonable heat output with air cooling. Sounds like it's worth waiting for Alder Lake in September, which will be the real upgrade.
Hi,
Yes the news report is just that horrendously inaccurate I'd add and shouldn't of been given light of day on tpu.
Posted on Reply
#52
tabascosauz
10900K leaked: 93C under 240mm AIO!1!1!1
10900K review: 54C, 85C no power limit under 120mm air

11900K leaked: 98C under 360mm AIO!1!1!1!
11900K review:

Starting to see a pattern here...

Also the title just completely taking the leaker's word on being "stock" is just irresponsible. Anyone with 2 brain cells who knows even a shred of how Skylake/Coffee/Comet Lake worked can tell that 1.4V isn't "stock all-core", at best terrible pre-release board firmware, at worst an intentional attempt at FUD.
Posted on Reply
#53
Phabets70
ThrashZoneHi,
AVX testing is a good thing to test
But just at 5.0 vcore is only 1.163v on auto under stress not 1.4v and I'm hitting on cpu package 71c
Max vcore is only 1.261v atm
I'll edit and add a screen shot in a minute of my 10900k readings same test.

I agree with you 100% that a stresstest should be performed with AVX / AVX-512 enabled but to get a realistic temperature reading comparable with 100% load you should disable it. In my case I even get thermal throttling with AVX enabled (with a crappy aircooler - Mugen 5) but I would never reach those same temps under normal circumstances. I would like to see temperatures of a Ryzen 5xxxx with AVX enabled for comparison. I hear an 5800 (with 1 CCX) can get quiete hot.
Posted on Reply
#54
bonehead123
B-Real98C? I bring the eggs.
And I will bring the frying pan, hahahahaha :roll:
Posted on Reply
#55
Cobain
Phabets70I agree with you 100% that a stresstest should be performed with AVX / AVX-512 enabled but to get a realistic temperature reading comparable with 100% load you should disable it. In my case I even get thermal throttling with AVX enabled (with a crappy aircooler - Mugen 5) but I would never reach those same temps under normal circumstances. I would like to see temperatures of a Ryzen 5xxxx with AVX enabled for comparison. I hear an 5800 (with 1 CCX) can get quiete hot.
5800x is VERY HOT and in that kind of stress tests the all Core clock goes down a lot.
Posted on Reply
#56
Mistral
Not called a "Rocket" for nothing, I guess...
Posted on Reply
#57
qubit
Overclocked quantum bit
ThrashZoneHi,
Yes the news report is just that horrendously inaccurate I'd add and shouldn't of been given light of day on tpu.
I hope you're right. We'll know for sure in a few months when the official reviews come out.
Posted on Reply
#59
Frick
Fishfaced Nincompoop
XL-R8RSensationalised news post that intentionally leaves out any mention of the use of ES/QS sample chips running way above the intended speed (these are not retail CPU's with finalised design specs) with way over spec voltage of 1.4v....




Not much to see here.
To be fair the concept of news is not what it used to be.
Posted on Reply
#60
Tech00
FP on AIDA64 on a AVX-512 capable CPU becomes almost pure AVX 512 test (at least that will drive the temps)! Loading all 10 cores' AVX-512 units 100% simultaneously for 10's of minutes and with low or now negative offsets (4.8Ghz is shown) is incredibly taxing and heat generating. This is why (for X chips 10890XE etc) there is a dedicated negative offset setting for AVX512 workloads (in BIOS ). for 10980XE etc the negative offset would be 8-10 or so but in this image for Rocket lake it shows the core running at 4.8Ghz = AVX-512 unit is running at 4.8GHz @98C. This is actually pretty good and very over clocked. Setting the AVX512 negative offset to 2 or 4 would drop the temps significantly (10 degrees or more).
Posted on Reply
#61
Caring1
PapaTaipeiOf true, that is unacceptable.
If true it's still within Intel specifications, just like most past generations.

Edit: Of course this could explain why the Integrated Graphics is being moved away from the Processor, to spread the heat distribution around.
Posted on Reply
#62
Fluffmeister
tabascosauz10900K leaked: 93C under 240mm AIO!1!1!1
10900K review: 54C, 85C no power limit under 120mm air

11900K leaked: 98C under 360mm AIO!1!1!1!
11900K review:

Starting to see a pattern here...

Also the title just completely taking the leaker's word on being "stock" is just irresponsible. Anyone with 2 brain cells who knows even a shred of how Skylake/Coffee/Comet Lake worked can tell that 1.4V isn't "stock all-core", at best terrible pre-release board firmware, at worse an intentional attempt at FUD.
Yeah come on now, this is getting petty. This isn't the EU Commission.
Posted on Reply
#63
Aquinus
Resident Wat-man
Sounds like Intel is tuning the boost algorithm to bounce against the CPU's upper thermal limit, a lot like mobile chips. The only difference is that it's a desktop chip and it eats power like a fat kid eats cake.
Posted on Reply
#64
Minus Infinity
Yep this is what happens when you are solely focused on winning a stupid benchmark. All pretense at power efficiency out the window to try and beat AMD in single threaded performance for 1080p lamers.

Backporting Ice lake cores to 14nm is a joke and it's not even total backport.

Rocket Lake is a joke and if you really want Intel wait for Alder Lake, that will be a much much better and more efficient product
Posted on Reply
#66
InVasMani
Certainly 1.4v isn't the default VID short of a ES and even then seems doubtful unless Intel intentionally did so to prevent leaks from divulging too much relevant information of what to expect. This seems to be a absolute worst case scenario across the spectrum. No rational thinker is pushing all at 8 cores all at once for a extended period at 4.8GHz and that amount of voltage. It probably runs noticeably better with 6 or 7 cores with that same VID at that same frequency though at the same time. This is just poor testing and a disinformation lower the core count one by one and show us the data. There isn't any practical conclusions from this other than 14nm at 1.4v across 8c is going to run hot which is at all something that would surprise anybody.

This isn't even about single thread performance for people remarking about that this is all core performance there is a difference you'd disable the extra cores and push single core frequency on a single core if that was your objective. Also you could technically test that on each core and determine which one has the best frequency and stability and assign the .exe to that specific core permanently with ImageCFG for the record though I haven't heard of anyone going to that extent and same is true of AMD chips.
Posted on Reply
#67
lesovers
ThrashZoneHi,
AVX testing is a good thing to test
But just at 5.0 vcore is only 1.163v on auto under stress not 1.4v and I'm hitting on cpu package 71c
Max vcore is only 1.261v atm
I'll edit and add a screen shot in a minute of my 10900k readings same test.

Everyone is forgetting about ambient temperature which should be subtracted to get the real performance and comparison result. The best indication of ambient air temperature in the case about the CPU is the motherboard temperature or other powered down device temperatures.

Therefore i9-11900KF results:

Td = 98C – 37C = 61C

P = 250W

Your i9-10900K results

Td = 71C – 26C = 45C

P = 209W

In the end we are just comparing the efficiency of the 360 RAD coolers including how flat and good the interface is for the CPU hotspots. The other factor is just the CPU wattage which is just heat generation (watts are just watts). This whole article is basically a non-story!
Posted on Reply
#68
watzupken
I was expecting the Rocket Lake to be hot, just like the name. While its lost 2 cores as compared to its predecessor, the new chip is actually being pushed very hard. If I look at a comparable example, i.e. Ice Lake, they run at a very low clockspeed. I presume its mostly because of the less mature 10nm, but I feel the architecture may not be meant for very high clockspeed in the first place.

It won't be long before we know how hot this chip runs. I am not optimistic because 250W PL2 is not something easy to cool. I assume Intel would have soldered this instead of using lousy thermal compound again.
Cobain5800x is VERY HOT and in that kind of stress tests the all Core clock goes down a lot.
It is true that the 5800X runs very hot. But it is possible to undervolt it quite a fair bit and also lower the power fed to it, and it still retains the same performance, and in some cases that I heard, improve performance. The chip is being fed too much power in the first place.

On the other hand, Intel really pushed the 11900K to the max, just like they did with 10900K. So its expected to run much hotter. The fortunate thing is that the die itself is quite big, so contact with the IHS and heatsink is better as compared to the chiplet design used in the Ryzen 3000 and 5000 series. But hopefully Intel will not scrimp on contact with the IHS.
Posted on Reply
#69
lexluthermiester
btarunr
This picture is not trustworthy. It looks deliberately grainy and seems to have been altered. Very suspect... Not call YOU out, @btarunr, just where-ever this pic is from.
Posted on Reply
#70
Midland Dog
you know its bullsch1t because AIDA is the most pathetic load ever
Posted on Reply
#71
InVasMani
These testing methodologies are terrible in terms of demonstrating anything credible that holds up to real scrutiny. The lack of easily individually inspected and validated information is the best form of disinformation. Results credibility and validity unknown.
Posted on Reply
#72
Makaveli
Cobain5800x is VERY HOT and in that kind of stress tests the all Core clock goes down a lot.
I just upgraded from a 3800X to a 5800X which test are these I would like to try them.
Posted on Reply
#73
R0H1T
qubitAlder Lake in September, which will be the real upgrade.
Real upgrade for who, Apple's M1 users?
Posted on Reply
#74
laszlo
Solid State Soul ( SSS )"Rocket lake" Because your fans will spin so fast, they'll launch your PC to the moon
Lava lake...
Posted on Reply
#75
qubit
Overclocked quantum bit
R0H1TReal upgrade for who, Apple's M1 users?
What? You know it's a big upgrade over Rocket Lake, don't you?
Posted on Reply
Add your own comment
Dec 23rd, 2024 02:26 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts