Ryzen 7000 Temperature Scaling - Overclocking the 7950X to over 6GHz with Dry Ice

[@TL....]

i just gone done watching a patio powerwashing video and now i'm here and i have no idea whats going on but i like it

[@donaldexley1852]

the amount of effort you put into these videos is amazing, whenever a new video comes up its a must watch!

[@shotybumbati]

It's insane to think that the normal boost clocks of modern CPUs are what we used to get only under extreme LN2 cooling- amazing times we're in

[@goldnoob6191]

Euh, 1Ghz was already a serious limit for pentiums. 🧐

[@anurag7058]

As impressive as it looks but i try to look at it differently. these cpu makers have hit the physical limit of adding more transistors hence they are only left with the option of increasing clocks to gain more performance. they cant do this forever. This indicates we need a breakthrough in cpu making technology maybe go quantum or something.

[@simonhazel1636]

@@anurag7058 well that's untrue, we've still got several lithography nodes to go before we start hitting a hard limit with silicon and even then smarter design, chiplets, 3d stacking will carry us a ways further.

[@wazaagbreak-head6039]

@@anurag7058 no just no we have an estimated 15 -20 years before we hit caps

[@hawkshot1025]

@@anurag7058 unless i'm wrong i'm pretty sure quantum computing as it is now isn't really viable for mainstream pc usage.

[@bobbymoss6160]

2 videos in 1 day! What a mad man!

[@Kvantum]

Remember, he's usually doing two versions, English und Deutsch, for each video, so that's basically 4 videos for him on a day like this.

[@bobbymoss6160]

@@Kvantum Roman is not human!

[@SixDasher]

Apparantly AM5 boards run through memory timings when you change the memory, which takes a long time and several automatic reboots. It looks for the best timings on it's own.

[@petert3355]

Yeah, I've heard of up to 10 boot loops to make sure the memory timings are as tight as possible.

[@Lishtenbird]

Say what you want about ASRock, but they bothered to put a simple "hey, this will take really long" sticker to inform people of this behavior.

[@Caboose2563]

It seems like Z690 boards have a similar thing too. My MSI one cycles through the same bunch of startup codes for 2-3 minutes before booting to windows (or telling me that the OC failed).

[@iPhixef]

My AM4 board powers up and down a few times before the screen actually displays anything whenever there's a power outage, is this related to what you're saying?

[@CSJSyr]

@@iPhixef Yep, but AM4 and ?Z690?(not sure) just do memory training for auto settings. But very loose timings.

[@ryzenforce]

That C5 code I think is typical of these new boards with RAM and CPU as it tries the best timing possible. It is not a bug and you have to let it do it thing as it will reboot a couple of times. It can take up to 20 min I've seen.

[@Hostilenemy]

According to AMD, the internal temperature sensors can read up to (or down to) -50C,

[@psychotictactoe]

Roman on fire today, unlike the R9! 😁

[@17themaster17]

Love the fact that, as a true man of science, you're using Kelvin degrees for scale purposes. Keep up the good work

[@seansean8008]

Thank you for your video. I have the 7950X, with PBO set to negative 30, and I have been getting all core boost between 5.3-5.4 GHz, 175-ish watts used and temps at 95c on a very robust custom loop while using Handbrake for rendering video. I was afraid that I was doing something wrong, but watching your video here made me realize that the system is running 'normal.' Thank you again.

[@halrichard1969]

I am really excited to see what you can do with a V-cache Model. Also of course testing that will reflect gaming performance.

[@nykraftlemagnifique]

Hail from France ! Love your T-Shirt with the inscription in french : "sans pareil" ! ;-)

[@kaloyantsonev130]

Поздрави от България 👍

[@cosminpopica]

I feel like you are onto a well paved way into becoming a great youtuber sir :D

[@lugga413]

He already has a big German Channel

[@LerRhann]

As always a great video from der8auer. Good stuff man.

[@Fosanrulz]

Your content is literally GOLD

[@JumpingJoseph]

Super informative, love your work, thanks as usual for the xtra effort for the English ❤

[@lordseph]

Makes you wonder how far he can push the delided chip (7900x) with the dry ice!~

[@mylittlepwny3447]

You can't use liquid metal with dry ice. It cracks at around -10C. And mounting an LN2 pot to the physical die itself without the IHS is dangerous and really hard to maintain contact due to thermal paste cracking. LN2 overclockers use CPUs with the IHS still installed for this reason

[@Scouse29K]

@@mylittlepwny3447 Cracking is mostly an issue at full pot, probably won't get full pot most of the time on these chips. Most of us don't direct die on ln2 due to it being easy to crush/crack the die if you even get it slightly wrong. Roman would use his Kryo Extreme paste on the dies if he tried dry ice cooling it, not LM

[@GAnimeRO]

@@mylittlepwny3447 Maybe with a ground down IHS it would be slightly better.

[@gg-gn3re]

@@mylittlepwny3447 crack kills

[@tarfeef_4268]

Probably not worth the chance of condensation on the chip, you'd want the "free seal" of the IHS and glue as a barrier to airflow to prevent moisture buildup. I'd use a new cpu.

[@theluggage7131]

Wish you would test normal boost behavior vs thermals while you did these cooling experiments and not just manual OC. Since that will be questioned all the time after release.

[@xeridea]

Probably similar. You can observe a slow frequency drop running an all core workload as temperature rises. AMD has likely done internal testing and has CPUs auto scale based on temperature, though at a much more granular level.

[@Edward135i]

Just want to let you know that the You Tube Algorithm worked in your favor, its been years since I've watched one of your videos back when you didn't have a English channel, so glad to rediscover your content. Also I've got to give you probs for shooting two videos in two different languages for every video that you do, that is a insane amount of work.

[@th3master]

would be interesting to try dry ice with pbo to see how much it does in auto oc mode.

[@MrStoffzor]

I bet GN will do this soon

[@darthkarl99]

Amen.

[@shahrukhwolfmann6824]

4:39 This channel is just EPYC!

[@swizzler8053]

Just finished your first video😁

[@miladesn]

Interested to see the degradation test on zen4 running at 95c after a year.

[@4gbmeans4gb61]

It hit 108c on an 360 aio doing simple cinebench, not good.

[@BBWahoo]

@@4gbmeans4gb61 Hotter means better :)

[@4gbmeans4gb61]

@@BBWahoo according to amd fanboys, intel cpus running that hot are pure garbage lolz

[@lasthopelost9090]

@@4gbmeans4gb61 to me that’s just Intel pushing it’s cpu design to the limit just to keep up where as the AMD cpu is just boosting its frequency till it hits the thermal limit so if you limit the frequency it shouldn’t get so hot

[@4gbmeans4gb61]

@@lasthopelost9090 lol? Have you seen gaming benchmark? New amd cpu gets beat by intel and last gen ryzen. All while being at 95-108c on watercooling. This seems like epic fail.

[@Дмитрий-с3п4ы]

Great analysis, thanks!

[@mylittlepwny3447]

This is good Content!

[@hugsun5918]

I would love to see the followup mentioned at 14:30 ❤

[@wolf1438]

Few hours ago JayzTwoCents explained AM5 AMD CPU has really loooooooong boot time memory training with 10+ reboots during process and it caught him off-guard. Almost 1 minute waiting for POST? Motherboard manufacturers should make some disclaimer about it or else there will be a lot of "faulty" systems reports on the line.

[@Maekiii]

Thats why they have done exactly that already.

[@stoneymahoney9106]

I saw an image of a board with a huge label stuck over the CPU socket/pcie slots (can't remember which) with exactly such a warning on it.

[@Jsteeezz]

@@stoneymahoney9106 that was an asrock board. It’s a smart idea so consumers dont freak out.

[@impuls60]

That will probably get fixed with bios updates. Ram stability got way better on Alder lake after a few bios revisions.

[@toast1797]

shows how much power can be shoved into these monsters, personally I would favour eco mode but it's nice to have the option to go full bore if required

[@RGBeanie]

Hopefully getting a 7950x was a good choice. At least future proofing my stuff for a bit with gaming/streaming/editing etc

[@stevesloan6775]

Radical content!!! Id love to own one of these amazing CPUs... It has all the hallmarks of a wild ZBrush processor. ZB loves ram so DDR 5 should be a noticeable speed up on big complex files. : )

[@DJSammy69.]

Ryzen 7000 most definitely need that de-lidding kit. Quite much potential performance hiding!

[@RyanF470]

Sounds like delidding or lapping the IHS you should be able to push further.

[@anthonyc417]

You can put the acetone in a squeeze bottle makes putting it in the pot so much easier IDK how safe it is though so @ me lol

[@DoObs]

Roman just dominating the Guru status of OC information.

[@Eveningbreeze78]

@der8auer Code 15 should be memory training.. it will cycle timings until expo finds the best stable timings

[@billgaudette5524]

I can take multiple minutes and reboots along the way, as well. Just let it go for 5 minutes at least and see what happens.

[@Eveningbreeze78]

@@billgaudette5524 just saw review from Jay2cents .. he had 10 boot cycles but after that mem was fine-tuned. Wendel at l1techs was also impressed with the auto-timings

[@steelfalconx2000]

I like that the new platform can flash bios without a CPU. It'll avoid many of the bios issues people had in Am4 where they couldn't update the bios because the CPU they got was the wrong generation.

[@michalp1602]

its a feature available on gigabyte motherboards, also on am4. idk about intel

[@Lucasbrlvk]

Good job🎉

[@bjn714]

That chipset temp at 10:12! 152.6C on the chipset! That has to be erroneous, as there is no way AMD raised the chipset thermal shutdown temp by over 50C from the previous 100C on x570. I wonder what the actual chipset thermal shutdown temp is...

[@___DRIP___]

I missed that! You’re right though, there’s no way that’s accurate lol. I wonder how it could get it so wrong though? Maybe it does a calibration on board before it does a reading, and it’s throwing the baseline off.

[@XeeDooX]

Is there a place where I would be able to buy 7950x already delidded ? Thanks

[@BeezyKing99]

if there is, be prepared to get scalped for one.... people reselling delidded CPU's are insane!.... tired of seeing a R5 2600 going for $800CAD when it's base price isn't anywhere near $800... it's $280CAD with everything in the CPU box as one would expect... otherwise buy one and get your hands on the delidder der8auer EN has and do the same thing you see him do during the delidding process.

[@XeeDooX]

@@BeezyKing99 Shortly said ... thank you, but no thank you. No way I am paying more than 100-300$ more for delidding service.

[@xThetomhawk]

@@XeeDooX 100$ is probably fair, but more not really.

[@XeeDooX]

@@xThetomhawk You are probably right. I am just adding some more extra for people who do not want to take the risk of braking the CPU when delidding. Also for top of the line products.. for 5800x for example 100$ is fine, although the risk is the same.

[@BeezyKing99]

@@XeeDooX I do agree with your point, I'll just never understand why people think it's okay to markup at least 50% on top of retail price just for delidding service.

[@haylspa]

90-95c standard temp can have limit set to 115c before throttling back, they have already got 6450mhz out of this cpu/chip soooo not sure whare your having an issue other then assuming temp ceiling

[@BBWahoo]

I like your hair, nice and poofy

[@NVMDSTEvil]

he wasnt pushing for max clocks, just looking at scaling. He stayed at 1.224v.

[@jlelelr]

My man is always one step ahead of other channels

[@4gbmeans4gb61]

97c with custom cooling ? WTF?

[@Diesr]

dry ice sounds quite nice

[@puciohenzap891]

Roll the dice!

[@huhneat8908]

doesnt taste great

[@BaH4o3eH]

don't you 4gate 2 ventil8

[@Wasmachineman]

@@BaH4o3eH Dice sublimates into CO², so yes, definitely ventilate properly.

[@Apollo-Computers]

I'm excited to do some xoc with these! Are you selling those gaskets?

[@m4dizzle]

Every Gigabyte product I've ever bought has been unreliable junk, glad to see you had better results -- although I suspect you're not using these long term

[@Gastell0]

JayzTwoCents has noted on the boot times, first time boot on AM5 will stay on code 15 for a long time and will restart multiple times - it is constantly adjusting timings on memory to get best results (instead of doing that manually in bios with xmp profiles or other settings)

[@weirdo911aw]

imagine still watching jaytwoshills 😭😭😭

[@MotoCat91]

@@weirdo911aw Imagine believing he shills for anyone other than ifixit.. and boats... and his own branded waterblock Same shit with Hardware Unboxed - every video showcasing AMD they're called AMD shills, every Nvidia video they're Nvidia shills but god forbid they say something bad about AMD CPUs one time and BAM they're Intel shills now I guess

[@roberthulmejr9520]

I would like to see if you can do better with direct die cooling with lm2 or dry ice

[@chloedevereaux1801]

lm2 ????? you mean LN2.

[@roberthulmejr9520]

@@chloedevereaux1801 yes ln2

[@fjnagle2nd]

The best thing I've found for removing your old thermal paste is hand sanitizer. Brand doesn't matter. Paper towel and q-tip to get into the hard to reach areas, this gel dissolves the old hardened thermal paste. Hope this helps. god bless and stay well.

[@user-bf5sc8pn8x]

Might as well use (99%) isopropyl alcohol (the main ingredient of hand sanitizer), although if you don't have any around you can use hand sanitizer

[@Calango741]

@@user-bf5sc8pn8x I always use a dry cloth first before using IPA because it it much cleaner and faster to not have all of the thermal paste liquified by the IPA.

[@HamBown]

99% IPA is a much better option. It is more effective and does not leave a sticky residue behind when it evaporates.

[@fjnagle2nd]

@@HamBown Thanks for the response. Didn't have any IPA available at the time. The thing about the HS is it doesn't evaporate as quickly, so you use it to soften up the thermal paste before you wipe it off. Also, I haven't noticed any sticky residue when using the HS.

[@fjnagle2nd]

@@user-bf5sc8pn8x Exactly what happened, no IPA around when I needed to clean off the thermal paste.

[@physxme]

6:56 the Chinese version of Ryzen Master also has broken translation. It's ridiculous for such a big company like AMD to ship software like this.

[@-opus]

9:33 dry ice dancing on the ram

[@khayman9574]

You speak English at native level and you complain about Ryzen Master's horrible German translation. You made my day there, bruv... 😂😂

[@SullySadface]

I wonder what this thing could hit on just one core

[@LayerZeroDesign]

I just rebuilt 3 days ago and I'm on the asus x670e crosshair gene with a 7950x and I have been having abnormally long POST times also. A few times, it actually would never boot into windows and I had to hard reboot. These were all after the PC had been sitting off during the day. Once I get it up and running and go through reboots it works ok, other than the long POST times. It seems like I get random debug codes/lights each time. I haven't had any real instability or other issues, but it has been a little concerning. I just updated to the latest bios version and when it was doing it's flash procedure it was being really weird. Eventually, I got it back up on the newest version and everything seems ok, but I seriously have no clue what is going on. Another matter entirely is with the newest bios version it seems Asus has activated their AI overclocker which has overriden the stock CPU behavior and even my own manual PBO settings. Out of the box my 7950x was boosting to 5.7ghz on 7 or 8 cores and like 4.8 on the rest iirc. (Not under heavy multithread obviously). After bios update it seems to be running it at 5.5ghz all core. This is happening on expo 1 and 2. I loaded defaults and the only thing I did was enable expo, and went into PBO in both areas of the bios and set the curve optimizer negative 16 and set some custom power limits. Not quite sure what I'm doing wrong. I'm just running a 280mm AIO for the moment but I have some EK hardware on the way, I'm excited to see how big of a clock speed difference I get off the custom loop. Also I ordered your Ryzen 7000 gaskets but I could not get it to line up properly with the notches on the CPU and eventually gave up in frustration lol. I still have 2 left, maybe I'll give it another shot when I install the new loop now that I saw your method. Thanks for all the trail blazing you do for all of us as always.

[@michaelthompson9798]

QQ: do you think AM5 socket closing mechanism design will affect even cooling like Alder Lake requiring a modified closing mechanism to reduce temps (well boost cpu frequency considering AMD7000 is designed to run @ 95c)?

[@Lishtenbird]

Yeah - after all the talk about Intel's "mistake" and how "AMD already knew how to do it right", I was pretty surprised to see a single-lever design here. I wonder if they managed to "fix" it with just minor pressure point tweaks, and without going for a multi-lever design.

[@zqzj]

Epic!!!

[@earthtaurus5515]

That gasket will come in handy, especially many folks will be looking to reign in thermals abit. Just seen GN's review of the 7950X and apparently AMD have designed Zen 4 to hit the thermal wall before a power wall thus undervolting may not be as useful. I'm curious, would undervolting help with thermals at all?

[@earthtaurus5515]

@Jeroen I'm aware thanks. I've been running my 2700X undervolted and oced to 4075 - the same with my ram - 4x8GB@2800Mhz too (from testing I found increasing the voltage on the RAM also increases the voltage to the memory controller albeit marginally but it does affect stability). According to Steve at GamersNexus apparently AMD said Zen 4 is supposed to run at TJ Max consistently. Therefore heating a thermal wall first before a Power wall. I've now just seen Gordon's (from PCworld) power comparison with Eco mode and he has shown there is 10% hit at 105 watt and 25% at 65 watt modes. Both of which still beat the 12900K in R23.

[@lasthopelost9090]

@Jeroen undervolting might make you lose frequency since this cpu will just keep boosting till it hits 95c

[@lasthopelost9090]

@Jeroen it very will could but how does reducing or limiting the voltage effect the boosting on these CPU’s? Is the question does it need thermal head room and voltage or will it still boost with more thermal headroom and less voltage to work with

[@kharak6204]

@Jeroen undervolting reduces temperature and POWER usage. zen4 increases POWER till it hits that thermal limit... and just like the turbo boost and all those things it is achieved mainly by higher voltage, so you're not really doing anything there on zen4, other than some minor efficiency adjustments since the cpus and mobos come shipped with certain always safe and stable settings, but what you get out of that can vary from cpu to cpu i really don't think there will be room for any real undervolting anymore, which is why the ECO mode exists to begin with

[@CL-rm6sb]

This has always been the case and hence why using curve optimizer in addition to PBO was always beneficial.

[@mikebruzzone9570]

There needs to be a tutorial on why extreme cooling to discover the processor frequency limits associated lithography shrink because on the passion of the pursuit for the engineering reason for has been lost. mb

[@mikebruzzone9570]

The engineering reason for extreme cooling has been lost explain this please or I am mistaken. mb

[@Andy-Andeee]

Why not use comformal coating on the exposed metal parts on the cpu to prevent electrical conductivity risk

[@yunodiewtf]

Hey man, have you ever tried actually indium-soldering a waterblock to a CPU? I understand that would be the most impractical and absurdly dangerous operation but I can't just get the idea out of my head but I'm kinda short on whole factorys worth industrial equipment and it'd be very cool if you did it instead with some old power hog CPU, especially comparing best liquid metal direct die vs indium-soldered wb.

[@talibong9518]

I don't think it's as difficult or crazy as it sounds. Remove the retention bracket, pop in the CPU and apply precise amounts of indium solder to each die and surround each die with silicone, carefully place a disassembled water block with just the mounting clips left attached on top of the CPU, heat up the water block until the solder melts. Now reassemble the water block and it can be dropped into the socket and bolted to the motherboard. I'm gonna have to try this on an old Q9400 I have as an experiment.

[@GTFour]

@@talibong9518 amazing, let us know how that goes! Video or write up with photos would be v cool

[@lord_khufu]

man these new hardware looks like they're from a different planet, hard to imagine stuff like this are gonna be outdated in the next decade or so

[@cokeforever]

more like two years or so

[@Christopher_S]

I'm watching this with my 2 year old daughter on my lap, and at 8:52 she starts shouting "Witch, witch. A scary witch is coming!" Haha.

[@bakedbeings]

Sorry if this has been answered many times: could your replace the original lid with a copper lid over liquid metal paste? Could manufacturers offer a soldered copper lid and would it improve cooling?

[@der8auer-en]

Stock IHS is made of copper

[@BBWahoo]

@@der8auer-en I wonder if shortening it would help any

[@bakedbeings]

@@der8auer-en Haha thanks 🙏 I guessed from the colour they were using a steel or aluminium alloy - next time I'll Google, I promise.

[@creaturedanaaaaa]

@@bakedbeings It's usually nickle plated copper.

[@lasthopelost9090]

My guess it maybe board vendors just wasn’t liking the AM4 socket and now there LGA sockets they feel more comfortable pushing it finally

[@tarekmahmudbhuiyan7368]

do a deep digging in 7000 series please.

[@Speak_Out_and_Remove_All_Doubt]

Does no one ever use those phase change freezer coolers anymore? They used to be all the rage in the early 2000's but super loud. I think LD Cooling still makes them but I have no idea what they are like with 300w CPUs.

[@Speak_Out_and_Remove_All_Doubt]

@Jeroen I really fancy making a submerged system but not sure how quickly those can dissipate heat from the dies, should be really good though I would have thought.

[@denvera1g1]

2:10 Yes, AMD socket was neglected, good luck getting 5000MT/s RAM to run on anything other than an ITX board

[@ThunderingRoar]

could be also related to amd locking down the agesa stuff whereas on intel boards manufacturers have more freedom

[@denvera1g1]

@@ThunderingRoar Could be, but then again, it works fine on ITX boards and specialty overclocking boards with only 2 physical slots. My 5700G runs great in 1:1 with 4933C17-17-17-28, actually faster than What wendle showed with the 7950x DDR5 6000 C30 with custom timings, he got something like 72GB/s 63ns and i get 74GB/s 49.1ns Oh and this is with a kit that costed me $90, IIRC that DDR5 kit was $340

[@ThunderingRoar]

@@denvera1g1 well you re comparing a monolothic cpu to chiplet one, on 7950x the 2 seperate core chiplets have to communicate with eachother so infinity fabric will often be the bottleneck

[@denvera1g1]

@@ThunderingRoar 3 chiplets

[@denvera1g1]

@@ThunderingRoar i'm not comparing the 5700G to the 5950x, i'm comparing the 5700G on half a dozzen ITX board to the same 5700G on really high end ATX boards All of the ITX boards will hit at least 4800, the ATX Boards struggle to hit 4200 unless they are 2 slot only boards specifically for RAM overclocking

[@Crushonius]

we need a lapping video please help us out brother

[@vibonacci]

Dry ice? Why not liquid helium?

[@d3n1ss_ok]

the chipset heats up like an infernal oven😀😀

[@paulchilders9969]

Still cooler than Intel.

[@klounader]

Freeze delidded processor!

[@zorrozalai]

It would be interesting to see an underclocking & undervolting championship, where the lowest stable voltage wins. 1st place would be something like 400 MHz 550 mV.

[@Dom_Mason]

Can't wait until the 13900K comes out. So we can do a true comparison between the 7950x and 13900K. I am hearing 13900k will have a 5.8Ghz turbo and should hit 6.0Ghz overclock just fine.

[@QRWASUPER]

Have you tried to undervolt ?

[@mmedo155]

I'm waiting for your PUBG benchmark with 7600x

[@Alex96194]

Me: watches this Also me: looks at my pc My pc: look we can talk about this

[@Slimmeyy]

What if you sanded down the IHS thickness to a more reasonable level? I would imagine it'd start at quite a bit higher than the Intel parts

[@mattyc7003]

Damn

[@a3j4s7]

Why not use Cinebench R23?

[@moasto02]

The 6GHz Project!

[@ErnestJay88]

Remember back when Intel Core 2 Duo mostly 1,6 - 1,8 Ghz, nowadays most modern CPU both AMD and Intel easily run at 5+ ghz on STOCK.

[@superpwnageable]

Where can we buy that CPU guard from ?

[@1BigBen]

best bet at caseking in due time

[@superpwnageable]

@@1BigBen thought so just wanted to make sure

[@tyronhamamoto5044]

I was wondering how exactly to deal with liquid metal on these CPU. i figure you needed to protect the capacitors around the IHS but not sure how much else if anything needs to be covered.

[@NinjaForHire]

I'm wondering if just hot glue would be a good option for sealing off the die test traces.

[@cleengreeny]

Just imagine what you can do with dry ice or LN2 with a de-lidded 7950x.

[@SuperLC1010]

Do you think we will have 20ghz by 2062?

[@hzpcs1319]

Gotta get me one of those gaskets.

[@crissnickers_frog6689]

a good scaling,.........thx

[@jezaustevas7170]

Can you test it dellided now? The thick IHS might be responsible for the curve

[@georgeyu798]

I am an ASIC design guy. We do active voltage scaling in our projects by monitoring transistor speed. What has always puzzled me is that, the transistors are faster when it is HOT (e.g. from -5C to 105C) so that we are able to lower power supply level when the die/junction gets hotter/closer to 100C. However in the CPU overclocking world, it seems that the processor can run faster when the junction temperature is colder? We are using the same TS*C transistors...any insights?

[@Spirit532]

Probably to do with electron mobility being higher and quantum tunneling being statistically less likely at elevated temps or something like that in your ASICs, but in CPUs you're just limited by extreme density and power output. The chip is able to clock this fast without cooling, but it'll just immediately overheat if not cooled, hence why this is perceived as colder=faster.

[@georgeyu798]

@@Spirit532 yea I was suspecting that, although the "case temp" is very low, the "junction temperature" is already at ~100C then that would make sense. However the screen shot from the video seems to suggest that the "junction temp" is very low at -30C while running. I am assuming AMD's temp sensor is the junction sensor, not the case sensor.

[@jestificated]

This guy hit 6ghz on water. Not stable but enough for benchmark. He did some unique settings to get there. "6 GHz Water Cooled Ryzen 9 7900X From BIOS to Benchmark"

[@attilavs2]

It's crazy how it's got 50% more. Cinebench performance than my 5600X

[@dominic.h.3363]

I don't know what boards you were looking at in the past, but whenever I wanted to buy something and started filtering for features, I always ended up with Intel boards filtered out almost immediately. AMD was always way better equipped. If anything, Intel started catching up with the 6xx chipsets TO AMD.

[@stoneymahoney9106]

Understand that when he's talking about "features", he's probably referring to things like VRM configuration and build quality, memory topology, PCIe bifurcation, stuff most people don't need to even understand, and wouldn't be included in motherboard retail listing filters.

[@6QG2P9X77W]

also depends the area of the world u are located in so can be different for everyone fr

[@dominic.h.3363]

@@stoneymahoney9106 I look at half of those as a regular ass user.

[@dominic.h.3363]

@@6QG2P9X77W I live two countries over, and therefore have the same suppliers.

[@stoneymahoney9106]

@@dominic.h.3363 That means you're not a regular-ass user ;)

[@pavanbiliyar]

I wonder if one-liter of Novec 7000 could get 6GHz at room temperature...

[@abrfilho]

What's the difference between Cinebench R20 and R23?

[@stoneymahoney9106]

R23 is exactly the same test, but runs loops for 10mins+ making it more suitable as a thermal testing and stability testing tool.

[@hadikhan63]

Please delid this cpu and use a watercooler . I am curious on how much it can be pushed.

[@HowDidThisHappenNow]

Kelvin and not Celsius ?

[@fnnhh]

A 10 kelvin change is the same as a 10 Celsius change

[@fluffyct100]

Roman couldnt help himself .. had to do 2 videos today lol

[@unijabnx2000]

Did you OC the delidded one with dry ice yet????

[@81formann]

40-50w _less_ power consumption with 4-600Mhz higher clocks... are those numbers correct? 5200Mhz/5400Mhz on water was 230w/238w 5800Mhz on dice was 190w with about 0c core temp. 5400Mhz with 0c would be... 165-175w ? With a potmeter to control your fans, you can adjust CPU speed and room temperature at the same time. Placed next to the volume knob for the 2.1 speakers for your convenience.