Intel "Tiger Lake-U" Processor with Relatively High Clock Speed Spotted

[btarunr]

An unnamed Intel "Tiger Lake-U" quad-core processor was spotted on Futuremark database by _rogame, featuring 2.80 GHz nominal clock-speeds. Barring the 28 W i7-1068NG7 and i5-1038NG7, which are exclusive for MacBooks and aren't considered U-segment, all current-gen "Ice Lake" client chips have their nominal clock speeds ranging between 1.00 to 1.30 GHz. Given this, 2.80 GHz would qualify as a big jump for a U-segment "Tiger Lake" chip. We know from a separate report that "Tiger Lake" could also offer Turbo Boost frequencies as high as 4.70 GHz for the top Core i7-1185G7 part, a similar jump from the 3.90 GHz max boost of the current-gen i7-1065G7, all while retaining a 15 W nameplate TDP.

The Futuremark database listing only mentions nominal clock of 2.80 GHz, and the CPU core configuration of 4-core/8-thread. The hardcoded CPU name string of this prototype specifies "Tiger Lake U," confirming this is a 15 W part, and not a 28 W part that will be gobbled down by Apple. Intel's newfound clock-speed headroom could be attributed to the company's refined 10 nm+ silicon fabrication node. "Tiger Lake" combines "Willow Cove" CPU cores with an iGPU based on the company's ambitious new Xe graphics architecture, marking its commercial debut. "Tiger Lake" is expected to launch around September-October, 2020.



View at TechPowerUp Main Site

 

[dicktracy]

Apple always eat first. Golden Cove should be exciting next year.

 

[FreedomEclipse]

"as high as..." is a bit of a meme for Apple at this stage as their recent macbooks like to do a lot of overheating and throttling of the CPUs. Its just how inadequate the cooling system is in their macbooks

 

[TheinsanegamerN]

Interesting, Curious to see if these speeds could indicate that Intel has partially fixed the 10nm scaling problem.

 

[RandallFlagg]

If they can get these out by Sept/Oct in shipping products, it will do a lot for Intel's laptop space.

This is what 10nm Ice Lake (much lower clocks than Tiger Lake) did for the MacBook Air vs the previous 14nm parts in Geekbench (this is single thread, multi thread is much faster due to more cores) :

 

[watzupken]

"as high as..." is a bit of a meme for Apple at this stage as their recent macbooks like to do a lot of overheating and throttling of the CPUs. Its just how inadequate the cooling system is in their macbooks

Looking forward to see if Intel can pull this off. I agree that based on recent products from Intel, they tend to boast very high clockspeed, but in real life, that claim falls flat due to thermal constraints. While I know that this is going to be a 10nm chip instead of the usual 14nm, we need to consider that the underlying CPU and GPU architecture is expected to improve a lot. The GPU in particular is expected to take up more die space as compared to the current Ice Lake UHD graphics.

At least for the Ice Lake U i5 processor I am using now, I am happy to report that it can hit the boost clockspeed of 3.6Ghz for awhile, and able to somewhat maintain around 3.2Ghz with a moderate load. In games, this may go lower to 1.4Ghz. Which is why I am somewhat skeptical about Intel's high clockspeed claim.

 

[efikkan]

Interesting, Curious to see if these speeds could indicate that Intel has partially fixed the 10nm scaling problem.

Intel claims the yields are better than expected, so they have fixed the yields problem.
But the other problem (which most ignore) is probably not "fixable"; production capacity. Intel simply didn't account for the increased demand in server and laptop segments, so even with the best yields, 10nm can't cover their entire lineup. I just wish they could spare some of it for a few (enthusiast) desktop SKUs, that would have come a long way.

 

[Vayra86]

The point of 4.7 Ghz turbo with a TDP ceiling of 15W completely eludes me. So they can do burst loads... great, what's new? They're still stuck with terrible performance after that. Sub or near 1 Ghz is just shit no matter how you slice this. Those are frequencies we used to idle at... and IPC has NOT improved substantially over the past five years.

This also does not confirm Intel 'fixed' their high performance on 10nm. We know efficiency is already out the window on 14nm+++... Intel doesn't give a shit if your battery level drops by a few % every time Tiger Lake bursts to 4.7... This number is highly suspicious wrt power draw and given its timing. They're also still just duals and quads

I think going forward I will be purchasing my CPUs based on the base clock spec and not turbo at all. I've just finished troubleshooting a Macbook Pro with a hexacore Intel in it... apparently 'normal use' is now having a laptop you can't place on your lap without burning a hole in your pants (and not just from the 3K price tag!). 'In spec'....

That together with the recent Throttlestop issues / undervolt lockdown on recent CPUs does not bode well.

Intel claims the yields are better than expected

The irony could not be greater. And you believe this? Expectations must have been pretty damn low then.

 

[tabascosauz]

The point of 4.7 Ghz turbo with a TDP ceiling of 15W completely eludes me. So they can do burst loads... great, what's new? They're still stuck with terrible performance after that. Sub or near 1 Ghz is just shit no matter how you slice this. Those are frequencies we used to idle at... and IPC has NOT improved substantially over the past five years.

This also does not confirm Intel 'fixed' their high performance on 10nm. We know efficiency is already out the window on 14nm+++... Intel doesn't give a shit if your battery level drops by a few % every time Tiger Lake bursts to 4.7... This number is highly suspicious wrt power draw and given its timing. They're also still just duals and quads

I think going forward I will be purchasing my CPUs based on the base clock spec and not turbo at all. I've just finished troubleshooting a Macbook Pro with a hexacore Intel in it... apparently 'normal use' is now having a laptop you can't place on your lap without burning a hole in your pants (and not just from the 3K price tag!). 'In spec'....

That together with the recent Throttlestop issues / undervolt lockdown on recent CPUs does not bode well.

Plundervolt does not affect the new "big core" uarches, I think. Ice Lake users have been able to UV just fine, it's just Comet Lake and earlier 14nm thermonuclear weapons that needed to BIOS rollback or edit their BIOSes.

Intel's straight up dug themselves back into the hole. The higher they push Skylake, the bigger the frequency hit that comes with the inevitable 10nm transition. The more cores Intel piles onto Comet Lake, the Rocket Lake transition becomes that much more nonsensical. If all goes according to rumors, Ice Lake to Tiger Lake handoff is going to be an entertaining show; I don't know out of which orifice Intel is going to pull a 4C/8T that's supposed to best a 8C/16T 10885H after they get rid of the entire moar-cores half of their lineup without a real successor. Sunny Cove hasn't got that revolutionary innovation in it, and isn't going to become that either with just a few cache changes in Willow Cove.

Renoir is great, because laptop manufacturers can never be trusted to engineer proper cooling solutions for ultrabooks. So instead of headbutting, retreating, then headbutting again into that 100° PROCHOT or artificial 15W PL1, much like Matisse, Renoir just regulates itself.

The problem isn't the single core max boost; when the load is on that core, it gets a little warmer than the rest and that's about it. Somewhere along the line, Intel started slapping these mind-boggling turbo tables on their -U SKUs where all-core turbo is just 200MHz below 1-core turbo. Remember when even full-fat desktop chips lost a few hundred MHz all-core to try and satisfy their TDPs? Ice Lake doesn't even reach 4GHz, and it already runs no cooler than the Whiskey Lake-U SKUs because despite the big cores's big IPC promises, it's only got so many threads and just messing around in Windows heavily loads 75% of them. Then the turbo bins kick in and there you are, sitting with Reactor No.4 in your lap.

Rather ironic that desktop -T and -S SKUs have not changed their power management configurations in 8 years, yet 15W -U mobile SKUs threw their turbo tables out the window and devolved into insanity. Speaking of turbo tables, Intel doesn't even publish turbo tables anymore, something they've done consistently for years. Now, you have to buy the product, open up Throttlestop and see for yourself what the bins are.

 

[londiste]

The point of 4.7 Ghz turbo with a TDP ceiling of 15W completely eludes me. So they can do burst loads... great, what's new? They're still stuck with terrible performance after that. Sub or near 1 Ghz is just shit no matter how you slice this. Those are frequencies we used to idle at... and IPC has NOT improved substantially over the past five years.

You are completely missing the point it seems.
- What is wrong with 4.7 GHz turbo in 15W? It means good single/low thread performance if it works.
- The entire point of this particular result is that Reported Stock Core Clock is 2.8GHz. This is the base clock. For comparison, Ice Lake 1065G7 has a base clock of 1.3GHz. If it turns out to be a true result for a 15W Tiger Lake, this is a significant improvement. 1.3 > 2.8 is a very big difference.
- Idle is completely irrelevant for this discussion, idle and low usage scenario clock and power management has been figured out for a while now.
- IPC has improved substantially over the past five years. Ice Lake is +18%, Tiger Lake so far is reported to get another 10+% improvement.

Speaking of turbo tables, Intel doesn't even publish turbo tables anymore, something they've done consistently for years. Now, you have to buy the product, open up Throttlestop and see for yourself what the bins are.

There is some merit in the claim that this is too dynamic to put in the table ever since core count went over 4 or so, along with the improvements in dynamic clocks speeds, temperature dependent boost and rising importance of power limits. By the way, AMD does not publish turbo tables either, never has and gives the same reasoning.

 

[Vayra86]

You are completely missing the point it seems.
- What is wrong with 4.7 GHz turbo in 15W? It means good single/low thread performance if it works.
- The entire point of this particular result is that Reported Stock Core Clock is 2.8GHz. This is the base clock. For comparison, Ice Lake 1065G7 has a base clock of 1.3GHz. If it turns out to be a true result for a 15W Tiger Lake, this is a significant improvement. 1.3 > 2.8 is a very big difference.
- Idle is completely irrelevant for this discussion, idle and low usage scenario clock and power management has been figures out for a while now.
- IPC has improved substantially over the past five years. Ice Lake is +18%, Tiger Lake so far is reported to get another 10+% improvement.

Ok, time to put on my reading glasses, because somehow I concluded these bases were stuck in 1 Ghz land.. thx lol

 

[tabascosauz]

There is some merit in the claim that this is too dynamic to put in the table ever since core count went over 4 or so, along with the improvements in dynamic clocks speeds, temperature dependent boost and rising importance of power limits. By the way, AMD does not publish turbo tables either, never has and gives the same reasoning.

AMD can hide behind that excuse because PB2 has a complex relationship (from user observations over the past 9 months) with temperature, which comes into play in regulating performance even when the chip is far away from its actual Tjunction. Renoir just takes thermal management one step more aggressively than Matisse, which only really started scaling back performance at 80C.

Intel does the same C-state and deep idle tricks as AMD, albeit slower and less finely tuned, but that's where the similarities end. Current -U SKUs aren't nearly as smart as they sound. They are hard limited by the PL1 and PL2 wattage limits (and sometimes a third long-term power limit) that are often arbitrarily set by the OEM. Aside from extremely poor VRMs that can't deliver enough power in some ultrabooks, they aren't limited by much.

Turbo Boost isn't dependent on temperatures. Not unless you count running into 95C PROCHOT then throttling, then proceeding to do exactly the same thing all over again after temps drop a bit, as "temperature dependent". That, and Thermal Velocity Boost, but that [thankfully] doesn't exist for 15W -U parts. If Intel -U SKUs were smart and knew how best to keep themselves from throttling, Throttlestop wouldn't need to exist.

The trend of all-core turbo bins creeping up to 1-core turbo bins in 15W parts isn't helping the fact that almost all Intel-powered ultraportables have throttling problems. Ice Lake solved none of those problems, despite its efficiency. Higher all-core and base clock is the reason for 45W parts' existence, and has always been. And base clock doesn't mean a whole lot to these -U chips; they'll boost all day until they run into PROCHOT, unless they're in Lenovo laptops, in which case they're tempered by the weird extra power limit.

Now, if Tiger Lake pulls out some smart power/temperature management controls on the level of PB2, then we might be in for an interesting treat. Otherwise, it's just going to do what the rest of the -Us have been doing for years - running facefirst into the same brick wall, repeatedly.

 

[londiste]

Thermal Velocity Boost absolutely exists for 15W U parts, specifically in Whiskey Lake U and now Comet Lake U.

 

[tabascosauz]

Thermal Velocity Boost absolutely exists for 15W U parts, specifically in Whiskey Lake U and now Comet Lake U.

I stand corrected. -U from those two families allow +100MHz when under 70C.

But that hardly changes anything in terms of thermal management. You can more or less treat TVB as an extra +1 turbo bin that's only going to be in effect for half a second if anything more than 1 core is used, thus something that only really works on 1 core. Which is fine, as lightly threaded boost is generally sustainable and is what drives user experience on ultraportables, but that doesn't change anything else as to how these processors behave.

 

[londiste]

That was not my point anyway. Thermal is but one of the variables. There are multiple and even if we only look at a basically table-based frequency, there is a power limit, there are different types of loads we have not even brought into discussion, like AVX2 for Intel. Basically if a table of frequencies would be published, there would be many many exceptions to that with any of the today's CPUs.

 

[watzupken]

You are completely missing the point it seems.
- What is wrong with 4.7 GHz turbo in 15W? It means good single/low thread performance if it works.
- The entire point of this particular result is that Reported Stock Core Clock is 2.8GHz. This is the base clock. For comparison, Ice Lake 1065G7 has a base clock of 1.3GHz. If it turns out to be a true result for a 15W Tiger Lake, this is a significant improvement. 1.3 > 2.8 is a very big difference.
- Idle is completely irrelevant for this discussion, idle and low usage scenario clock and power management has been figured out for a while now.
- IPC has improved substantially over the past five years. Ice Lake is +18%, Tiger Lake so far is reported to get another 10+% improvement.

There is some merit in the claim that this is too dynamic to put in the table ever since core count went over 4 or so, along with the improvements in dynamic clocks speeds, temperature dependent boost and rising importance of power limits. By the way, AMD does not publish turbo tables either, never has and gives the same reasoning.

There is nothing wrong to have a faster turbo clock. Only problem is how long will it sustain this speed in order for us to observe tangible performance improvement. Based on my current Ice Lake U based laptop, the boost speed is at most a second or less, and it depends on how often it happens. As to the base clockspeed, compared to the listed specs between Ice Lake and Tiger Lake, the former has a significantly lower base clockspeed. Again based on my logs, the Ice Lake U in my laptop is running between 3 to 3.3Ghz 68% of the time when under gaming load which I ran for round about 22 mins. But again, this is at the mercy of the cooling solution implemented by the PC maker.

In terms of IPC improvement, I am not sure if you mean +18% over 5 years? Ice Lake as compared to Whiskey Lake may have improved somewhat due to increased cache, but +18% don't sound right to me. Unless under specific conditions and not general performance, then perhaps.

 

[TheGuruStud]

There is nothing wrong to have a faster turbo clock. Only problem is how long will it sustain this speed in order for us to observe tangible performance improvement. Based on my current Ice Lake U based laptop, the boost speed is at most a second or less, and it depends on how often it happens. As to the base clockspeed, compared to the listed specs between Ice Lake and Tiger Lake, the former has a significantly lower base clockspeed. Again based on my logs, the Ice Lake U in my laptop is running between 3 to 3.3Ghz 68% of the time when under gaming load which I ran for round about 22 mins. But again, this is at the mercy of the cooling solution implemented by the PC maker.

In terms of IPC improvement, I am not sure if you mean +18% over 5 years? Ice Lake as compared to Whiskey Lake may have improved somewhat due to increased cache, but +18% don't sound right to me. Unless under specific conditions and not general performance, then perhaps.

What OEM is your laptop? That's a lot of watts for them to allow (or maybe just 25 b/c it's gaming workload).

 

[efikkan]

In terms of IPC improvement, I am not sure if you mean +18% over 5 years? Ice Lake as compared to Whiskey Lake may have improved somewhat due to increased cache, but +18% don't sound right to me. Unless under specific conditions and not general performance, then perhaps.

IPC is "workload agnostic", an average across many benchmarks, as you can see here.
Ice Lake offers 18% IPC gains over the Skylake family, but Ice Lake-U runs at lower clock speeds than Whiskey Lake-U, so the real world difference may be less.
I don't know if the claims of IPC gains for Tiger Lake is true, but it will at least achieve much higher clock speeds.
Also be aware that U-series chips can be configured as 15/25/35W by the OEM, as well as the delay of the power limit.

If your objection to the IPC claim is due to the computer not "feeling" significantly faster, then there is a good reason for that. As low power CPUs try to squeeze in more cores and higher boost clocks (faster than the nodes advances), they have to sacrifice sustained all core speed. This increases the variance of the performance of the CPU, which becomes more and more noticeable for the end user. Using laptops today feels like driving a car with a tiny engine and a huge turbo; sometimes it performs and sometimes it doesn't…

 

[RandallFlagg]

Why speculate so much? You all have the perfect comparison mark via the MacBook Air Ice Lake vs Skylake (Y) processors. In single core it is showing 35% bumps in CPU performance (even more in multi-core but that is b/c of moar cores). This is the result of a combination of higher IPC + Intel 10nm process node (which is similar in density to TSMC 7nm).

Those Ice lake 9-10W CPUs (replacing the Skylake Y series) are actually more powerful than the Skylake U series 15-25W CPUs. There is every reason to indicate that these new Tiger Lake CPUs will heavily encroach on the performance of the bottom half of the Skylake 9th/10th gen 35-45W+ H-series CPU performance.

It's too bad Intel could not get 10nm out in even higher power parts, but that does not in any way negate the performance these chips are demonstrating in their respective segments.

 

[TheGuruStud]

Why speculate so much? You all have the perfect comparison mark via the MacBook Air Ice Lake vs Skylake (Y) processors. In single core it is showing 35% bumps in CPU performance (even more in multi-core but that is b/c of moar cores). This is the result of a combination of higher IPC + Intel 10nm process node (which is similar in density to TSMC 7nm).

Those Ice lake 9-10W CPUs (replacing the Skylake Y series) are actually more powerful than the Skylake U series 15-25W CPUs. There is every reason to indicate that these new Tiger Lake CPUs will heavily encroach on the performance of the bottom half of the Skylake 9th/10th gen 35-45W+ H-series CPU performance.

It's too bad Intel could not get 10nm out in even higher power parts, but that does not in any way negate the performance these chips are demonstrating in their respective segments.

No, apple finally kinda upgraded cooling so they're not stuck at full throttle 24/7.

 

[RandallFlagg]

No, apple finally kinda upgraded cooling so they're not stuck at full throttle 24/7.

Do you always just make up straw man fake arguments, or only when it’s about tech?

Apple did not improve cooling in the mba, they merely made the fans modular along with a few other components for easier service replacement.

 

[TheGuruStud]

Do you always just make up straw man fake arguments, or only when it’s about tech?

Apple did not improve cooling in the mba, they merely made the fans modular along with a few other components for easier service replacement.

I was thinking of the pro.

But I do guarantee it's almost all in single core boost.

 

[RandallFlagg]

I was thinking of the pro.

But I do guarantee it's almost all in single core boost.

Yeah mbpro is still skylakeish.

2020 mba has double the skylake version multi thread score in geekbench.

There aren’t many ways you can slice it, 10nm ice lake is a huge improvement.

Now there are some slices. The best SKUs are only for apple for instance, equivalent non apple skus were removed.