Discussion Intel current and future Lakes & Rapids thread

[LightningZ71]

Ice Lake was 4 cores and a modest iGPU increase due to the 10nm process being somewhat scaled back to improve yields, and Intel still having issues with power/performance. To make an 8 core Ice Lake, apparently, it would have yielded horribly and consumed too much power to be usable in mobile. Also, keep in mind for Tiger Lake, they are going to vastly scale back the iGPU in it to keep die size and power draw reasonable. Granted, this isn't really an issue for the target market, as TGL-H is aimed at the higher end laptop market where dGPUs are very common, and when not included, gpu performance wasn't a focus for the market sub-segment.

I suspect that TGL-H is going to be a beast as compared to the R 4800/4900H, but maybe it'll be delayed enough to have to face off against Cezanne based systems instead.

 

[clemsyn]

Intel said 8 cores TGL are coming, it is likely they did not expected AMD to move to 8C in notebooks, but i would not expect those to be on 15W.

Looking at their 4c/8t scores, don't you think they should do a 6c/12t instead? Looks like a better move than jumping to 8c/16t IMO.

 

[Adonisds]

So why did they increase the die size of cores to make them have lower IPC on average than Ice Lake?

 

[A///]

At least it appears to be significantly better than previous-gen 4c Intel mobile SoCs, which isn't something that could be said for IceLake. It's a different market now, though, so people are going to have to decide for themselves whether they want more clockspeed plus better iGPU or more cores than what they could get in the past. Honestly, when it was confirmed that TigerLake wasn't going to launch with more than 4c, did you really think it was going to beat Renoir in everything? I didn't.

I'd heard higher core parts would show up later. I'd also read something about a 30% uplift which I knew was bull right off the bat. I buy a laptop once a decade. I'm not the target buyer.

 

[Hitman928]

I know but this is not the point. You said it's running at 38W until it drops and as you can see there is no drop in this game test from Notebookcheck. Depending on how long one game bench lasted it uses much more than Anandtech claims. However it seems they blindly assumed it runs on default 15W PL1 all the time.

Haven't they blindly assumed TGL runs at 15W/28W as well then? The Notebookcheck test runs Witcher 3 for a very short period of time so it shows much higher power use because it's still in the max turbo or all core turbo zone. Long term it won't stay there. For battery life, we have tests for Renoir laptops showing very strong battery life performance. We'll have to wait for TGL because no OEM machines have been allowed to be reviewed yet and Intel didn't allow battery life tests of their reference platform.

 

[Zucker2k]

Long term it won't stay there.

And then what'll the performance be like? If anything, opportunistic boosting during gaming is much more important than power consumption, for the best results. Won't you agree?

 

[Hitman928]

And then what'll the performance be like? If anything, opportunistic boosting during gaming is much more important than power consumption, for the best results. Won't you agree?

Performance would obviously be worse than what they show which is why sites doing laptop reviews should not do such short gaming benchmark runs like NBC seems to do but instead wait until the boost states run out to show what performance you'd actually get during the vast majority of your gaming experience.

I would say that if you are playing something like Witcher 3 or and not Bejewled, then long term sustained performance would be more important than boosting performance unless you want to take frequent breaks to let your laptop cool down and reset the power timers.

 

[eek2121]

IMO Intel didn't wait, they just f.d up with 10nm
if I remember correctly, there was a leak of cannonlake 8C
well we know now that they can deliver, how much that we don't

IMO tigerlake 8C is ready, but if released now it will compete with 10700K in MT and crush it in ST, so they can't bery their own 2 months old chips

Actually, an 8 core Tiger Lake chip would crush the 10900K. Oops.

 

[DrMrLordX]

I'd heard higher core parts would show up later. I'd also read something about a 30% uplift which I knew was bull right off the bat. I buy a laptop once a decade. I'm not the target buyer.

Heh well Intel themselves sort of hinted that TigerLake wasn't going to offer much over IceLake IPC-wise (all they really said was that it had restructured cache).

Actually, an 8 core Tiger Lake chip would crush the 10900K. Oops.

You and @TheGiant may well be right, but . . . why not crush it? They could do away with Comet Lake and Rocket Lake in one decisive blow, and buy themselves some time to delay Alder Lake-S/P if necessary. That is assuming that 8c TigerLake would clock high enough to perform as well as you indicate. Who knows, maybe it would?

 

[mikk]

Haven't they blindly assumed TGL runs at 15W/28W as well then? The Notebookcheck test runs Witcher 3 for a very short period of time so it shows much higher power use because it's still in the max turbo or all core turbo zone. Long term it won't stay there. For battery life, we have tests for Renoir laptops showing very strong battery life performance. We'll have to wait for TGL because no OEM machines have been allowed to be reviewed yet and Intel didn't allow battery life tests of their reference platform.

It runs for a short period of time, the entire 80 seconds test no power drop. Intels PL2 is a 28s boost until it falls back to the sustained load power, so it's shorter.

 

[Hitman928]

It runs for a short period of time, the entire 80 seconds test no power drop. Intels PL2 is a 28s boost until it falls back to the sustained load power, so it's shorter.

It's hard to compare AMD and Intel boost durations because according to Anandtech's testing, Renoir follows a softer roll-off algorithm instead of hard drops like what Intel has introduced with TGL and the Adaptix technology.



Second, the 80s Witcher 3 test from Notebook check for the Yoga slim 7 shows the same power consumption as their average load test (1.45% difference) which is lower than their max load test and is the same behavior seen in the Dell laptop they tested with Icelake.



Lastly, the system consumption graphs for Witcher 3 and Prime95 show them both using the same amount of power at the end of the graph where Prime95 was running for 5 minutes but Witcher 3 for 90 seconds. According to their Prime95 graph, the 4800u consumes ~15W after about 90 seconds or so of Prime95 run time.






So the system power consumption numbers between Prime95 and Witcher3 seem to indicate they both use the same amount of power long term and that the SOC is using 15W of that system power consumption. With that said, I'll repeat, testing laptops is not easy and the best way to do it is compare like for like models for performance and battery life as in the end, that's what really matters anyway.

The Ryzen 7 4800U is an Absolute Monster: Lenovo Yoga Slim 7 14 Laptop Review

After facing a few delays, AMD's premier Ryzen 7 4800U is finally starting to hit store shelves. Results are insane for a 14-inch subnotebook to the point where it'd be tough to recommend an Intel counterpart when given the same price.

www.notebookcheck.net

 

[ondma]

So why did they increase the die size of cores to make them have lower IPC on average than Ice Lake?

What cores are you talking about? Tiger Lake certainly does not have lower IPC than IL. In fact it should be higher. How much we can only tell after direct comparisons.

 

[tamz_msc]

What cores are you talking about? Tiger Lake certainly does not have lower IPC than IL. In fact it should be higher. How much we can only tell after direct comparisons.

On average TGL has slightly lower IPC than ICL according to Anandtech. There are gains in workloads which take advantage of the redesigned cache, but there are regressions in other workloads. Bottom line is that TGL does not improve IPC over ICL, just like Intel said it wouldn't.

 

[LightningZ71]

[To the question of why not just do 6 cores from above]
For the H market, it now expects to have 8 cores available. I don't think that they could get away with just six cores. It looks like they don't want to make the die much larger than it currently is (probably for yield reasons more than anything else) so its going to require cutting down the iGPU anyway. If they are cutting it down, it won't be a performance leadership selling point, so they will need to have 8 cores no matter what. They can still offer a six core for reject recovery parts.

 

[jpiniero]

For the H market, it now expects to have 8 cores available. I don't think that they could get away with just six cores. It looks like they don't want to make the die much larger than it currently is (probably for yield reasons more than anything else) so its going to require cutting down the iGPU anyway. If they are cutting it down, it won't be a performance leadership selling point, so they will need to cores no matter what. They can still offer a six core for reject recovery parts.

I think the core counts were always what Intel intended. The H 4C+96 EU doesn't really make much sense to me though, doesn't really seem like it'd be smaller or meaningfully better than the U parts.

 

[mikk]

Here is another test I found:

Handbrake video conversion doesn't make sense to me.

 

[LightningZ71]

H 4C+96EU may make sense in the slightly larger than an ultrabook market. If you need maximum portable AVX-512 performance without the bulk and draw of a dGPU, it would cover your needs. If they can take the same silicon and get an additional 25% ASP for it, why wouldn't they?

 

[IntelUser2000]

On average TGL has slightly lower IPC than ICL according to Anandtech. There are gains in workloads which take advantage of the redesigned cache, but there are regressions in other workloads. Bottom line is that TGL does not improve IPC over ICL, just like Intel said it wouldn't.

No it doesn't. They tested using SPEC workload, where even the Integer test is fairly demanding.

It only looks to have lower perf/clock because its clocked nearly 25% higher. This isn't Dhrystone or Geekbench we're talking about here.

Nothing uniformly increases performance. Especially when things change the fundamentals, like the L2/L3 where it hasn't changed since Nehalem in 2010. That's why overall with many different applications are needed.

@Hitman928 There's something definitely going on with the GPU power figure, but the Slim 7 is not at 15W. It uses more than Intel systems that have 25W settings, and significantly higher than AMD systems at 15W.

 

[itsmydamnation]

@Hitman928 There's something definitely going on with the GPU power figure, but the Slim 7 is not at 15W. It uses more than Intel systems that have 25W settings, and significantly higher than AMD systems at 15W.

i have a slim 5 with a 4700u so not quite the same , but i can check power draw on some specific tests if you want.

 

[jpiniero]

H 4C+96EU may make sense in the slightly larger than an ultrabook market. If you need maximum portable AVX-512 performance without the bulk and draw of a dGPU, it would cover your needs. If they can take the same silicon and get an additional 25% ASP for it, why wouldn't they?

But that's pretty much what U is. As you can see from AT's review, Tiger U could easily be set to PL1 of 45 W. It may just simply be the U silicon in H packaging (so OEMs can easily put the 8 C + 32 EUs die later if they get any). It makes for weird marketing unless all the H models are 32 EUs max regardless of what's available on the die.

 

[IntelUser2000]

i have a slim 5 with a 4700u so not quite the same , but i can check power draw on some specific tests if you want.

That's cool.

Ideally, you'd do this.

Monitor power usage when plugged in using a power meter. Report the ballpark power figure in the first 15 seconds of the Prime95 benchmark. Then, report the numbers again after 30 mins of looping the same bench.

You'd also do it under certain power settings, or tell us at least. Whether its Extreme Performance Mode, or Balanced, or Power Saver.

Even better, do it also while gaming. This is probably going to be a lot harder unless it has a built-in benchmark.

You could run DC power comparisons while having HWInfo up and monitor the system power drain, but without comparable Icelake system on hand, there's not much point.

 

[Adonisds]

What cores are you talking about? Tiger Lake certainly does not have lower IPC than IL. In fact it should be higher. How much we can only tell after direct comparisons.

From the Anandtech review:
IPC improvements of Willow Cove are quite mixed. In some rare workloads which can fully take advantage of the cache increases we’re seeing 9-10% improvements, but these are more of an exception rather than the rule. In other workloads we saw some quite odd performance regressions, especially in tests with high memory pressure where the design saw ~5-12% regressions. As a geometric mean across all the SPEC workloads and normalised for frequency, Tiger Lake showed 97% of the performance per clock of Ice Lake.

 

[Markfw]

From the Anandtech review:
IPC improvements of Willow Cove are quite mixed. In some rare workloads which can fully take advantage of the cache increases we’re seeing 9-10% improvements, but these are more of an exception rather than the rule. In other workloads we saw some quite odd performance regressions, especially in tests with high memory pressure where the design saw ~5-12% regressions. As a geometric mean across all the SPEC workloads and normalised for frequency, Tiger Lake showed 97% of the performance per clock of Ice Lake.

And what are the 2 clocks ? Ice Lake and Tiger Lake that you are using before normalization ?

 

[itsmydamnation]

That's cool.

Ideally, you'd do this.

Monitor power usage when plugged in using a power meter. Report the ballpark power figure in the first 15 seconds of the Prime95 benchmark. Then, report the numbers again after 30 mins of looping the same bench.

You'd also do it under certain power settings, or tell us at least. Whether its Extreme Performance Mode, or Balanced, or Power Saver.

Even better, do it also while gaming. This is probably going to be a lot harder unless it has a built-in benchmark.

You could run DC power comparisons while having HWInfo up and monitor the system power drain, but without comparable Icelake system on hand, there's not much point.

this is what i have done so far

screen brightness set to max , 2d monitor via usb-c
I would clear data in hwmon wait 5 seconds and record the median max value to determine the 1T core, max vlaue of package etc
I think this method is over stating average power usage in cinebench a bit because unlike prime small FFT its power usage values bounch around alot on the active core (4-9 watt range)

wall plugged	time (sec)	power mode	wall measure (w)	hwmon
				Package	Core	1T core	clock
idle		intel cooling	15​	2.5​	0.4​
prime small FFT 1T	15​	intel cooling	37​	19​	18.2​	16.2​	4.2ghz
prime small FFT 1T	60​	intel cooling	37​	19​	18.4​	16.2​	4.2ghz
prime small FFT 1T	180​	intel cooling	30​	12.9​	11.4​	10​	4.0ghz
prime small FFT 1T	780​	intel cooling	30​	12.98​	11.4​	10​	4.0ghz
idle		extreme performance	15​	2.5​	0.3​
prime small FFT 1T	15​	extreme performance	38​	24​	19​	17.7​	4.2ghz
prime small FFT 1T	60​	extreme performance	38​	19​	18.7​	16.2​	4.2ghz
prime small FFT 1T	420​	extreme performance	38​	18.8​	18.7​	16​	4.2ghz
prime small FFT 1T	900​	extreme performance	38​	18.8​	17.7​	17.8​	4.15ghz
prime small FFT 1T	1200​	extreme performance	39​	19​	19​	17.7​	4.15ghz
idle		intel cooling	15​	2.5​	0.4​
Cinebench R20 1T	15​	intel cooling	29​	12​	10​	9.0​	4.2ghz
Cinebench R20 1T	60​	intel cooling	30.8​	12​	10​	9.0​	4.2ghz
Cinebench R20 1T	180​	intel cooling	29.4​	11.5​	10.9​	9.0​	4.2ghz
Cinebench R20 1T	900​	intel cooling	28.8​	11.5​	10.9​	9.0​	4.2ghz

 

[IntelUser2000]

@itsmydamnation Thanks for the test that's great.

Those system power numbers are in line with Notebookcheck systems that are set at 15W. Of course its running on one core so its to be expected.