ValantarWhile both of those are true, GloFo 14nm was not a particularly high-clocking node (being based on Samsung's mobile/low power focused 14nm), and Intel 14nm has seen a series of revisions with explicit design goals being increased clocks (mostly through optimizing design libraries for higher voltages, which in turn lowers efficiency). Intel 14nm didn't clock particularly high at first, and TSMC 7nm still doesn't. Also, we are fast approaching a point where production nodes are small enough that their ability to handle the voltages needed for high clocks drops off significantly. This is why AMD's current chips still struggle with higher all-core clocks than ~4.3GHz even on well-binned chips - they simply can't handle the voltage necessary. This won't be getting better with upcoming smaller nodes either. High-power 10-16nm-class nodes are likely to be the highest clocking nodes we'll ever see.

And I remember hearing how we would never see hard drives larger then 10GB, that it just wasnt possible to put more sectors closer together without overwriting them, and how going below 100nm was going to pose such significant struggles with quantum tunneling that we'd never see them in consumer products, and how pushing memory speeds above 800 mhz was going to produce so much latency they would be useless, then how DDR4 would be useless due to latency, and now the peanut gallery is chanting on how DDR5 will be high latency and not worth it, and how with 28nm it was getting so difficult to produce large dies we'd never see a GPU over 600mm2 ever again, and yada yada yada yada.

Hey, remember how sub 20nm nodes would never be able to handle voltages over 1.2V? Remember how Moore's law was dead, and we'd never see significant performance/watt improvements over ivy bridge quad cores? I remember hearing all that garbage on forums and in tech news.

Much like samsung and intel's 14nm, just because TSMC 7nm doesnt do it well NOW doesnt mean it never will. 7nm EUV is supposed to allow for 15-20% higher clocks or reduced power consumption, on a node that supposedly was impossible to make back in 2010, when sub 10nm was written off for being too close to electron size.

Notice a pattern here? Now the chant is "high clocks will never happen again below 14nm!!!!". I'm not going to make that basis on ONE GENERATION of CPUs from AMD that dont hit 5 ghz, from an arch that has NEVER clocked up well. That is how many tech "predictions" end up eating elephant sized crow.

watzupkenI am skeptical that it will even do 3.8 to 4Ghz on all 8 cores to be honest. From what I observed, most laptops improve cooling by slapping more heatpipes, criss-crossing it everywhere to 2 heatsinks. While heat is quickly moved to the heatsink because of the heatpipes, the bottleneck is always the physical size of the heatsink and the blower (more heatsink). I've used a few gaming laptops before and despite the elaborate cooling solution, generally the CPU will hit high 80s to 90s easily under load. When this happens, even with the fan ramped up to 90 or 100%, the CPU will always throttle to the base speed. My observations are based on 4c/8t 45W processors few years back. So with 2x the cores and higher clockspeed + power, keeping 8 cores cool and running substantially higher clockspeed is not possible on a laptop no matter how you cut it. Otherwise, there will not be some laptops that comes with watercooling, like the one from Asus.


Personally, laptops with top end desktop processors makes the least sense. I think you can read reviews of how they perform. Physically, laptops don't have the luxury of space for massive heatsink to cool the components. Laptops of this class generally comes with some high end graphics as well, which adds on to the cooling woes. Under load, the end result is that both the CPU and GPU will suffer due to extremely high temps and substantial throttling is to be expected.

One way to help increase core clock is by lowering the cache ratio. Not a lot of laptop folks know how much this can impact, but lowering from 42x to 32x, reduced my temps by around 10C. Not a big difference eh? But how did it affect my performance? Virtually zero impact, so there isn't a reason to run it at full speed unless you have awesome cooling. Reduce the cache ratio if throttling, that will help you more than reducing core ratio.

yotano211The cooling on current laptop is really bad and I kinda like it hot in my house. I prefer the heat. i7 8750h, the specs are over there
<<<

Haha just noticed, alrighty, you should be able to run 41/41/40/40/39/39 on your 9750H max. Those are the turbo ratios for 1c/2c/3c/4c/5c/6c respectively. In your scenario, you SHOULD run your computer hotter (since you like heat), and also, max out your clock speeds. If it is capping at 3.3 GHz, I think there is some manual limit in place that can be overridden; use ThrottleStop to increase your clocks. Also I recommend undervolting your processor but by all means go for as high as it can clock! I have a 9750H and I have an undervolt of around -150mV, all cores go at 4.0 GHz as it is a slightly higher clocked 9750H, but temps go to almost 100C when cache is 39x-42x, and ~90C when cache is 32-33x.

sergionographyMelting laptops will be the new trend. Or laptops with cup holders to keep your coffee hot

hatI've explained this before. The reason that we're "against boosting", in this scenario, is because it's simply not going to happen. Nobody cares if a laptop boosts to 5.3GHz for a fraction of a second, before power or thermal limits kick in, to load a word document or a web page 1/10 of a second faster than it would have if it just ran at base clock all the time. As such, the feature is useless... and who wants features that don't work? No, we don't get a working feature... instead we get "5.3GHz!!111!111" plastered everywhere, when in reality it can only reach that speed for a very short time, i.e. applications where 5.3GHz performance isn't necessary anyway.

Now, take that same chip and put it in a desktop where power and thermal limits aren't woefully constrained like they are in a laptop, and you get a different story. Nobody complains about the 9900k boost clock, because it can actually reach it. Nobody complains about the boost feature on Nvidia graphics cards, because they actually do it. People complain about the advertised boost clocks on these laptops, because they don't do it. That's the key difference.

Have you also explained that more clocks and smaller die sizes that "are plastered everhwere" do not increase gaming performance ? ... or in pretty much everything else 98+% of what people do on a daily basis ? Why are cores and die size relevant when TPU testing shows CPU overclocking is relevant. Sure, if you are a user trying to game on a 3.5 pound ultralight but again, on a properly designed and preferabley custom built lappie, no problem. There's a very active overclocking community on the Clevo website with lapped CPUs, delidding, cooler mods and high sustainable OCs from simple things to extreme. Before Alienware was bought by Dell, they were simply Clevo custom built laptops.

When you say "Nobody complains about the 9900k boost clock, because it can actually reach it.", you should be aware that the fact is custom built laptopss have the option of using desktop CPUs 9900k so nobody should be complaining by your reasoning. The cooling systems are designed to handle the load.

Clevo P775TM1-G 17" Laptop
9th Generation Intel® Core™ i9-9900K Processor (16M Cache, up to 5.00 GHz)
17.3" Full HD 144Hz Wide View Angle 72% NTSC Matte with G-SYNC Technology
NVIDIA® GeForce™ RTX™ 2080 GPU with 8GB GDDR6
32GB Dual Channel DDR4 3000MHz (PC4 24000) - 2 X 16GB
Windows® 10 Pro 64-Bit Edition Preinstalled, (with 64-Bit USB Recovery Media)
Thermal Grizzly Kryonaut Thermal Compound - CPU + GPU
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (OS DRIVE)
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (SLOT 2)
SAMSUNG® 860 PRO™ 4TB SATA III 3-D Vertical SSD
Intel® Wi-Fi 6 AX200 M.2 AX + Bluetooth® 5 Combo Card

Clevo PB71RF-G
9th Gen Intel® Core™ i7-9750H Processor (12M Cache, up to 4.50 GHz)
17.3" Full HD (1920 x 1080) 144Hz, Wide View Angle 72% NTSC Matte with G-SYNC Technology
NVIDIA® GeForce™ RTX™ 2070 with 8GB GDDR6 Video memory
32GB Dual Channel DDR4 3000MHz (PC4 24000) - 2 X 16GB
Thermal Grizzly Kryonaut Thermal Compound - CPU + GP1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (OS DRIVE)
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (SLOT 2)
SAMSUNG® 860 PRO™ 4TB SATA III 3-D Vertical SSD
Intel® Wi-Fi 6 AX200 M.2 AX + Bluetooth® 5 Combo Card

Should also be aware that the performance difference between mobile and desktop CPUs / GPUs has considerably narrowed. Let's look at some numbers:

Card: Ranking - 3D Mark Ice Storm / 3D Mark Cloud Gate / 3D Mark Firestrike

NVIDIA GeForce RTX 2070 (Desktop) : Ranked 20th - 424385 / 126874 / 23373
NVIDIA GeForce RTX 2070 (Mobile): Ranked 23rd - 425550 / 116232 / 20036.5

Gaming performance as tested, ranges from 119 desktop / 118 mobile (Escape from Tarkov) to 184 / 154 (Doom Eternal)

Imagine that ... only 2 GPUs in the world faster than the 2070 Mobile and slower than the 2070 desktop... GTX 1080 and RTX 2070 Super.

We have been buying Clevo laptops exclusively for, I'd guess, going on 20 years ...1) because nothing we've found performs better, 2) unequal freedom in component selection, 3) no artificial limitations on performance tweaking and 4) less expensive than gaming laptops from MSI. After installing OS / running RoG Real Bench and Furmark is the 1st thing done and when I put the effort in, have even managed to get a notable OC.

Berfs1One way to help increase core clock is by lowering the cache ratio. Not a lot of laptop folks know how much this can impact, but lowering from 42x to 32x, reduced my temps by around 10C. Not a big difference eh? But how did it affect my performance? Virtually zero impact, so there isn't a reason to run it at full speed unless you have awesome cooling. Reduce the cache ratio if throttling, that will help you more than reducing core ratio.


Haha just noticed, alrighty, you should be able to run 41/41/40/40/39/39 on your 9750H max. Those are the turbo ratios for 1c/2c/3c/4c/5c/6c respectively. In your scenario, you SHOULD run your computer hotter (since you like heat), and also, max out your clock speeds. If it is capping at 3.3 GHz, I think there is some manual limit in place that can be overridden; use ThrottleStop to increase your clocks. Also I recommend undervolting your processor but by all means go for as high as it can clock! I have a 9750H and I have an undervolt of around -150mV, all cores go at 4.0 GHz as it is a slightly higher clocked 9750H, but temps go to almost 100C when cache is 39x-42x, and ~90C when cache is 32-33x.

MSI usually throttle their gaming laptop starting at 94C. 3.3ghz is with undervolt at 100%.

John NaylorHave you also explained that more clocks and smaller die sizes that "are plastered everhwere" do not increase gaming performance ? ... or in pretty much everything else 98+% of what people do on a daily basis ? Why are cores and die size relevant when TPU testing shows CPU overclocking is relevant. Sure, if you are a user trying to game on a 3.5 pound ultralight but again, on a properly designed and preferabley custom built lappie, no problem. There's a very active overclocking community on the Clevo website with lapped CPUs, delidding, cooler mods and high sustainable OCs from simple things to extreme. Before Alienware was bought by Dell, they were simply Clevo custom built laptops.

When you say "Nobody complains about the 9900k boost clock, because it can actually reach it.", you should be aware that the fact is custom built laptopss have the option of using desktop CPUs 9900k so nobody should be complaining by your reasoning. The cooling systems are designed to handle the load.

Clevo P775TM1-G 17" Laptop
9th Generation Intel® Core™ i9-9900K Processor (16M Cache, up to 5.00 GHz)
17.3" Full HD 144Hz Wide View Angle 72% NTSC Matte with G-SYNC Technology
NVIDIA® GeForce™ RTX™ 2080 GPU with 8GB GDDR6
32GB Dual Channel DDR4 3000MHz (PC4 24000) - 2 X 16GB
Windows® 10 Pro 64-Bit Edition Preinstalled, (with 64-Bit USB Recovery Media)
Thermal Grizzly Kryonaut Thermal Compound - CPU + GPU
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (OS DRIVE)
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (SLOT 2)
SAMSUNG® 860 PRO™ 4TB SATA III 3-D Vertical SSD
Intel® Wi-Fi 6 AX200 M.2 AX + Bluetooth® 5 Combo Card

Clevo PB71RF-G
9th Gen Intel® Core™ i7-9750H Processor (12M Cache, up to 4.50 GHz)
17.3" Full HD (1920 x 1080) 144Hz, Wide View Angle 72% NTSC Matte with G-SYNC Technology
NVIDIA® GeForce™ RTX™ 2070 with 8GB GDDR6 Video memory
32GB Dual Channel DDR4 3000MHz (PC4 24000) - 2 X 16GB
Thermal Grizzly Kryonaut Thermal Compound - CPU + GP1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (OS DRIVE)
1TB SAMSUNG® 970 PRO™ M.2 PCIe NVMe V-NAND SSD (SLOT 2)
SAMSUNG® 860 PRO™ 4TB SATA III 3-D Vertical SSD
Intel® Wi-Fi 6 AX200 M.2 AX + Bluetooth® 5 Combo Card

Should also be aware that the performance difference between mobile and desktop CPUs / GPUs has considerably narrowed. Let's look at some numbers:

Card: Ranking - 3D Mark Ice Storm / 3D Mark Cloud Gate / 3D Mark Firestrike

NVIDIA GeForce RTX 2070 (Desktop) : Ranked 20th - 424385 / 126874 / 23373
NVIDIA GeForce RTX 2070 (Mobile): Ranked 23rd - 425550 / 116232 / 20036.5

Gaming performance as tested, ranges from 119 desktop / 118 mobile (Escape from Tarkov) to 184 / 154 (Doom Eternal)

Imagine that ... only 2 GPUs in the world faster than the 2070 Mobile and slower than the 2070 desktop... GTX 1080 and RTX 2070 Super.

We have been buying Clevo laptops exclusively for, I'd guess, going on 20 years ...1) because nothing we've found performs better, 2) unequal freedom in component selection, 3) no artificial limitations on performance tweaking and 4) less expensive than gaming laptops from MSI. After installing OS / running RoG Real Bench and Furmark is the 1st thing done and when I put the effort in, have even managed to get a notable OC.

I love Clevo laptops, my last Clevo laptop was a P370sm3. It had one of the 1st 120hz screens on any laptop. The Nvidia 970m sli was a nightmare to get it running correctly.

EarthDog5.3 ghz.. yikes!!!


They'll use sTIM I'd imagine.

Mobile chips are mostly bare dies. No IHS.

btarunrMobile chips are mostly bare dies. No IHS.

Oof, yep!

TheinsanegamerNAnd I remember hearing how we would never see hard drives larger then 10GB, that it just wasnt possible to put more sectors closer together without overwriting them, and how going below 100nm was going to pose such significant struggles with quantum tunneling that we'd never see them in consumer products, and how pushing memory speeds above 800 mhz was going to produce so much latency they would be useless, then how DDR4 would be useless due to latency, and now the peanut gallery is chanting on how DDR5 will be high latency and not worth it, and how with 28nm it was getting so difficult to produce large dies we'd never see a GPU over 600mm2 ever again, and yada yada yada yada.

Hey, remember how sub 20nm nodes would never be able to handle voltages over 1.2V? Remember how Moore's law was dead, and we'd never see significant performance/watt improvements over ivy bridge quad cores? I remember hearing all that garbage on forums and in tech news.

Much like samsung and intel's 14nm, just because TSMC 7nm doesnt do it well NOW doesnt mean it never will. 7nm EUV is supposed to allow for 15-20% higher clocks or reduced power consumption, on a node that supposedly was impossible to make back in 2010, when sub 10nm was written off for being too close to electron size.

Notice a pattern here? Now the chant is "high clocks will never happen again below 14nm!!!!". I'm not going to make that basis on ONE GENERATION of CPUs from AMD that dont hit 5 ghz, from an arch that has NEVER clocked up well. That is how many tech "predictions" end up eating elephant sized crow.

There are always new innovations moving things forwards, but the past decade has been one long story of major progress due in large part to better designs and more money for more ambitious designs (compared to the previous decades where a lot of development was slow simply because the companies involved couldn't afford more) on the one hand, with the behind-the-scenes stuff slowing down on the other. No matter if previous predictions have been wrong, there are serious innovations needed to make upcoming 5nm and 3nm nodes clock as high as current ones, let alone higher. I'm not saying it can't or won't happen, but it will be slow and expensive. We've been hearing of new materials revolutionizing IC production for two decades now, and none of them have really panned out yet. We're reaching the practical limits of some fundamental underpinnings of all silicon manufacturing - such as the use of copper for internal wiring - and have yet to really come to grips with replacing it, even if we theoretically know how this might be done. Another major factor is cost - HDDs are a good example of this. There has been massive progress in the HDD space in terms of capacity over the past decades, but prices per GB have stagnated, meaning the capacity increases don't actually benefit users. The biggest HDD I have is 6TB, simply because anything bigger gets stupidly expensive and price/GB actually rises. Similar developments will hit other sectors of the PC space too, with new innovations being so costly as to continuously drive prices upwards. I do think the HDD market is a worst case scenario (high base cost for materials and parts, other limitations shrinking the market, extreme commodification), and I don't think we'll ever see a situation where GPU or CPU price/perf stagnates for a decade, but as innovation slows prices will inevitably suffer. We're already at a point where a CPU can realistically perform adequately for mixed use (including gaming) for nearly a decade. GPUs make bigger strides, but that too will inevitably taper off as clocks and die sizes start hitting various limits. Then we'll have to look to exotic packaging for performance increases, which will boost prices, and is only a stopgap solution. There's always something that can be done, but with every refinement and development made, there's less left to do, so improvements become slower and more expensive.

Reviews are in for the 4900HS, these intel chips look silly now.

VaderReviews are in for the 4900HS, these intel chips look silly now.

I haven't seen any like-for-like comparisons for gaming (same chassis won't happen, but same GPU would be nice), but for anything CPU-bound they look amazing. Of course It will definitely be interesting to see what 10th gen H-series chips being to the table to compete with this. Nonetheless that Asus G14 is a mighty attractive package.

VaderReviews are in for the 4900HS, these intel chips look silly now.

Not to play devils advocate, but the i9-9880H is NOT intel's top chip, its the 9980HK. There is a rather large difference with the two, the 9980HK is binned AND unlocked. As far as I know, AMD's mobile chips are not unlocked.

Berfs1Not to play devils advocate, but the i9-9880H is NOT intel's top chip, its the 9980HK. There is a rather large difference with the two, the 9980HK is binned AND unlocked. As far as I know, AMD's mobile chips are not unlocked.

I think AMD doesn't have locked or unlocked CPUs like Intel does. They are simply CPUs. Desktop CPUs don't have that and I'm sure mobile as well. Besides, who would OC a mobile processor in a laptop with so many constraints? Power envelope is important.

ratirtI think AMD doesn't have locked or unlocked CPUs like Intel does. They are simply CPUs. Desktop CPUs don't have that and I'm sure mobile as well. Besides, who would OC a mobile processor in a laptop with so many constraints? Power envelope is important.

(for Ryzen) Desktop consumer chips are for the most part unlocked. Desktop prosumer chips (pro) are locked. Mobile chips I think are locked, not 100% on that but from my datasheet all of them are locked (at least for Ryzen era). The reason overclocking a mobile processor makes sense is it helps increase performance when plugged in. obviously when unplugged you generally want to improve efficiency, but when plugged in, that's when overclocking would make sense.

ratirtI think AMD doesn't have locked or unlocked CPUs like Intel does. They are simply CPUs. Desktop CPUs don't have that and I'm sure mobile as well. Besides, who would OC a mobile processor in a laptop with so many constraints? Power envelope is important.

Epyc and mobile cpus are locked.

Berfs1(for Ryzen) Desktop consumer chips are for the most part unlocked. Desktop prosumer chips (pro) are locked. Mobile chips I think are locked, not 100% on that but from my datasheet all of them are locked (at least for Ryzen era). The reason overclocking a mobile processor makes sense is it helps increase performance when plugged in. obviously when unplugged you generally want to improve efficiency, but when plugged in, that's when overclocking would make sense.

Notebooks have no BIOSes that can overclock.
Also, if you don't want to fry your CPU/APU, better don't try overclocking on a notebook.
I am not sure but their cooling solutions must be worse than the worst BOX cooler.

champsilvamobile cpus are locked.

Then what makes them cTDP and how is that altered?

Caring1Then what makes them cTDP and how is that altered?

I would guess by the specific OEM firmware?

ARFNotebooks have no BIOSes that can overclock.
Also, if you don't want to fry your CPU/APU, better don't try overclocking on a notebook.
I am not sure but their cooling solutions must be worse than the worst BOX cooler.

Not true, there are overclockable BIOSes that exist (though the BIOS itself may not have the features visible). For intel platforms, you can use ThrottleStop to overclock the processor, as well as Intel Extreme Tuning Utility (XTU), if your processor supports it. From my experience, the iGPU is overclockable only in XTU, but TS is better for everything else. For AMD systems however, Ryzen Master apparently doesn't work on mobile Ryzen chips (not 100% sure about that), so there isn't any way to modify it other than editing the power profile. If your manufacturer provides drivers, use those. Those can potentially allow tuning.