Discussion Intel current and future Lakes & Rapids thread

[B-Riz]

The TigerLake-H Review

Loool, are we supposed to be excited and or impressed by Tiger Lake H?

"In our benchmarks, both the Core i9-11980HK and the Ryzen 9 5980HS battle it out, sometimes with the Intel chip coming ahead, sometimes with the AMD chip leading the results. The issue with this comparison though is that we’re comparing a 45W chip vs a 35W chip, and more often in compute heavy workloads such as rendering or encoding, the AMD chip comes ahead even though it has a lower TDP. "

 

[DrMrLordX]

And TSMC do business, it doesn't matter where the money comes.

Of course it matters where the money comes from. If propping up Intel helps Intel roll out 2nm GAAFETs earlier than later, maybe TSMC is going to think twice about helping Intel maintain/restore their core revenue stream. TSMC's motives are a bit off-topic, mind you, but wrt anyone Intel out of their mess, it's going to take some major cash and other concessions for the pure-play fabs to want any part of it.

If all Intel does is take outside wafers for their dGPUs where they aren't an established player, then sure, maybe.

Other what Gelsinger said : Intel designs are free to use best manufacturing process available, from Intel itself or from the foundries. So if they aren't also trying to get best possible process as soon as possible they aren't executing properly.

That's an extension of the idea that Intel designs aren't somehow locked to a specific node anymore, starting with Sunny Cove. Which, after Rocket Lake, I'm beginning to think is more smoke and mirrors than truth.

 

[Zucker2k]

Loool, are we supposed to be excited and or impressed by Tiger Lake H?

"In our benchmarks, both the Core i9-11980HK and the Ryzen 9 5980HS battle it out, sometimes with the Intel chip coming ahead, sometimes with the AMD chip leading the results. The issue with this comparison though is that we’re comparing a 45W chip vs a 35W chip, and more often in compute heavy workloads such as rendering or encoding, the AMD chip comes ahead even though it has a lower TDP. "

Except that the Ryzen 9 5980HS won't run at 35w for much, if at all, of those tests. I'm not saying the 11980HK is, either, but the difference comes from throttling, from boosting, where the Intel chip boosts for only 5 seconds? That's atrocious.

 

[ondma]

Ok, if 11900K actually had good performance than power use could be excused.

Except it performs quite horribly. Zen 3 based chips are a far better deal. In scenarios where it would actually matter to save money and time is where Zen 3 whips Rocketlake completely.

Intel is forced to push it out of bounds because they are that far behind.

Zen 3 desktop chips have double the cores of Zen 3 mobile parts. Alderlake-S should have been at least 8+12, or even 8+16, and mobile getting less.

I can see Alderlake mobile potentially being very good. Desktop? Better than today is all I have to say.

Come on, Rocket Lake does not perform "horribly". If tuned properly, it comes very close to Zen 3 in gaming and productivity. If RL performs "horribly", they Zen 3 is only some percent above "horrible". I will grant that performance per watt is very poor, and Zen 3 is clearly a better processor, but RL is a very powerful chip.

 

[Zucker2k]

What I find more interesting, is that Tiger Lake-H has marginally better ipc than Cezanne, about 7%

 

[Zucker2k]

Come on, Rocket Lake does not perform "horribly". If tuned properly, it comes very close to Zen 3 in gaming and productivity. If RL performs "horribly", they Zen 3 is only some percent above "horrible". I will grant that performance per watt is very poor, and Zen 3 is clearly a better processor, but RL is a very powerful chip.

I honestly don't understand how much more performance Intel is supposed to squeeze from an aged process on a back-ported design.

 

[Hulk]

Regarding Zen 3 vs Tiger Lake power draw comparisons based on the new Tiger Lake H review.

Looks like at 65W given adequate cooling Tiger Lake H on 10SF could sustain 3800MHz at 65Watts.

Looking back at the Zen 3 5950X review we see 120 Watts at 3775MHz. So basically 120W vs. 130W for TMSC Zen 3 7nm vs TGL 10SF. Pretty darn close. I believe both of these measurements are package power. TGL also includes iGPU so we could call this even.

 

[Hitman928]

Regarding Zen 3 vs Tiger Lake power draw comparisons based on the new Tiger Lake H review.

Looks like at 65W given adequate cooling Tiger Lake H on 10SF could sustain 3800MHz at 65Watts.

Looking back at the Zen 3 5950X review we see 120 Watts at 3775MHz. So basically 120W vs. 130W for TMSC Zen 3 7nm vs TGL 10SF. Pretty darn close. I believe both of these measurements are package power. TGL also includes iGPU so we could call this even.

Are you sure the load on the CPUs is the same in both tests? You're also ignoring that Intel would have to go to a different (most likely higher power) interconnect scheme to reach the 16 cores of the 5950x which would also reduce its frequency within the same power envelope and the iGPU is going to be power gated if it's not in use so it won't be eating into the power budget of the SoC.

A 5800x limited to 65W would be more interesting but I don't really understand the point when we have directly comparable Zen 3H and TLGH processors to compare. . .

 

[Hitman928]

What I find more interesting, is that Tiger Lake-H has marginally better ipc than Cezanne, about 7%

Assuming they are at their max boost frequencies it is (according to Anandtech's 1T SPEC test):

Vermeer > TGLH > CezanneH(S)

111.7% > 103% > 100%

 

[andermans]

Assuming they are at their max boost frequencies it is (according to Anandtech's 1T SPEC test):

Vermeer > TGLH > CezanneH(S)

111.7% > 103% > 100%

I think one should indeed be careful extrapolating Cezanne performance to Vermeer. AFAICT in partice the difference between those two is bigger than just power limits (e.g. compare a 5700G vs. a 5600x on ST, should be similar powerlimit & similar top ST boost). Could be the difference in L3 cache size, could be something else.

Also I don't think we've seen clock normalized tests yet? Certainly Anandtechs SPEC tests don't seem to be so we're technically talking about +7% ST performance, not +7% IPC (unless there was a test that I've missed?).

No snide against Tigerlake meant here, but I think extrapolating results is hard, especially because for a lot of these CPUs we're talking <10% differences on average.

 

[Hitman928]

I think one should indeed be careful extrapolating Cezanne performance to Vermeer. AFAICT in partice the difference between those two is bigger than just power limits (e.g. compare a 5700G vs. a 5600x on ST, should be similar powerlimit & similar top ST boost). Could be the difference in L3 cache size, could be something else.

Also I don't think we've seen clock normalized tests yet? Certainly Anandtechs SPEC tests don't seem to be so we're technically talking about +7% ST performance, not +7% IPC (unless there was a test that I've missed?).

No snide against Tigerlake meant here, but I think extrapolating results is hard, especially because for a lot of these CPUs we're talking <10% differences on average.

I didn't extrapolate, the 5950x SPEC results are in the same review. . .

I somewhat agree on IPC comparisons are tough, especially outside of the same series of architectures or even ISAs and that talking IPC in broad or sweeping terms is silly, but with the proper scope in mind, it can be an interesting discussion.

Edit: I will say, in today's world, I think perf/W is a far more interesting discussion at given thresholds is far more interesting than perf/Hz (IPC).

 

[andermans]

I didn't extrapolate, the 5950x SPEC results are in the same review. . .

Sorry, that was meant wrt your discussion on the 120W vs. 130W above where Hulk extrapolated by doubling TGL-H. MY bad, should've been clearer.

 

[Gideon]

Except that the Ryzen 9 5980HS won't run at 35w for much, if at all, of those tests. I'm not saying the 11980HK is, either, but the difference comes from throttling, from boosting, where the Intel chip boosts for only 5 seconds? That's atrocious.

This is what Andrei himself had to say about the subject:

This is generally irrelevant to the results because our test suite runs things back-to-back, and actually handicaps the AMD systems in terms of them hitting that skin temperature dependent boost. It would only boost the first few workloads of hours-long testing or workloads following some very light workloads at below TDP which had allowed the laptop to cool down.

So looking a the power graphs from the article and what he stated, it's very clear that for the vast majority of the tests Cezanne was running at 35W and Tiger Lake at 45W.

 

[Zucker2k]

This is what Andrei himself had to say about the subject:


So looking a the power graphs from the article and what he stated, it's very clear that for the vast majority of the tests Cezanne was running at 35W and Tiger Lake at 45W.

He also said:

The Ryzen 9 has a prolonged 300s semi-turbo state where it sustains 42W power until thermal saturation of the laptop. During this period, with similar power consumption to the 11980HK and also quite similar thermal results of around 80-83°C, both platforms seem to be quite similar – except for the fact that AMD Zen3 cores able to operate at all-core boost frequencies of around 4GHz, while the Willow Cove cores of the TGL-H system operate at around 3200MHz and below. This is an important metric to note as we dive deeper in other results of our test suite.

 

[Hulk]

Are you sure the load on the CPUs is the same in both tests? You're also ignoring that Intel would have to go to a different (most likely higher power) interconnect scheme to reach the 16 cores of the 5950x which would also reduce its frequency within the same power envelope and the iGPU is going to be power gated if it's not in use so it won't be eating into the power budget of the SoC.

A 5800x limited to 65W would be more interesting but I don't really understand the point when we have directly comparable Zen 3H and TLGH processors to compare. . .

I just thought it interesting. For all of Intel's apparent process issues 10SF seems to keep up with TMSC 7nm pretty well.
Perhaps if Intel had the capacity they could have released TGL for the desktop and the resulting power/thermal envelope would have been comparable to Zen 3.

 

[uzzi38]

He also said:

Yes, Cezanne clocks vastly higher (4GHz vs 3.2GHz) at the same power consumption (well 42W vs 45W here) vs Tiger Lake. I don't get it though, what's your point?

 

[uzzi38]

Regarding Zen 3 vs Tiger Lake power draw comparisons based on the new Tiger Lake H review.

Looks like at 65W given adequate cooling Tiger Lake H on 10SF could sustain 3800MHz at 65Watts.

Looking back at the Zen 3 5950X review we see 120 Watts at 3775MHz. So basically 120W vs. 130W for TMSC Zen 3 7nm vs TGL 10SF. Pretty darn close. I believe both of these measurements are package power. TGL also includes iGPU so we could call this even.

In this workload the iGPU is not utilised. Essentially speaking it does not pull power at all.

 

[Hulk]

In this workload the iGPU is not utilised. Essentially speaking it does not pull power at all.

Yes I realize that. But even if it's only drawing 3 or 4 watts that's closes a bit the gap to Zen 3.

 

[Zucker2k]

Yes, Cezanne clocks vastly higher (4GHz vs 3.2GHz) at the same power consumption (well 42W vs 45W here) vs Tiger Lake. I don't get it though, what's your point?

If Cezanne is enjoying a 20% clock advantage in prolonged all-core parallelized tests and they're still trading blows, what does that say about Tiger Lake-H ipc?

 

[Hitman928]

Yes I realize that. But even if it's only drawing 3 or 4 watts that's closes a bit the gap to Zen 3.

It's not going to pull 3 or 4 watts. It will be effectively power gated and pulling essentially 0.

 

[Hitman928]

If Cezanne is enjoying a 20% clock advantage in prolonged all-core parallelized tests and they're still trading blows, what does that say about Tiger Lake-H ipc?

Cezanne is not enjoying the 20% clock advantage in these tests, when Cezanne is boosting beyond 35 W, Tigerlake is boosting beyond 45 W, so we're back to the same conclusion of TGLH can be competitive in performance if it's able to use a significant amount more power.

 

[uzzi38]

If Cezanne is enjoying a 20% clock advantage in prolonged all-core parallelized tests and they're still trading blows, what does that say about Tiger Lake-H ipc?

It says surprise surprise, turns out AMD lost some IPC when they cut L3 to 16MB per CCX.

Tiger Lake IPC isn't any better than desktop Zen 3. It's better than the castrated Zen 3 on mobile. Congratulations?

EDIT: Also, in "prolonged" (>5 minute) workloads the 5980HS goes back to pulling 35W where that clock advantage no longer exists. At 35W it clocks at the same 3.2GHz as Tiger Lake in P95. So uh...

 

[Zucker2k]

Let me quote this again:

The Ryzen 9 has a prolonged 300s semi-turbo state where it sustains 42W power until thermal saturation of the laptop. During this period, with similar power consumption to the 11980HK and also quite similar thermal results of around 80-83°C, both platforms seem to be quite similar – except for the fact that AMD Zen3 cores able to operate at all-core boost frequencies of around 4GHz, while the Willow Cove cores of the TGL-H system operate at around 3200MHz and below. This is an important metric to note as we dive deeper in other results of our test suite.

I've highlighted all the relevant points, I think. Basically, for 5 minutes, Zen 3 enjoys a 20% all-core clock advantage before settling to 35w. Note that this is not normal boost, but a "semi-turbo state." It's all in the quote.

 

[coercitiv]

I've highlighted all the relevant points, I think. Basically, for 5 minutes, Zen 3 enjoys a 20% all-core clock advantage before settling to 35w. Note that this is not normal boost, but a "semi-turbo state." It's all in the quote.

You need to understand that SPEC is not a 5-10 minute test, but a 6 hours one.

The test takes several hours of runtime (6 hours for this TGL-H SKU) and is under constant full load, so lower duration boost mechanisms don’t come into play here.

In terms of the overall performance, the 45W 11980HK actually ends up losing to AMD’s Ryzen 5980HS even with 10W more TDP headroom, at least in the integer suite.

The 65W TDP scores for TGL should be ignored though, as the test machine cannot handle this TDP and actual performance would be higher.

 

[uzzi38]

Let me quote this again:


I've highlighted all the relevant points, I think. Basically, for 5 minutes, Zen 3 enjoys a 20% all-core clock advantage before settling to 35w. Note that this is not normal boost, but a "semi-turbo state." It's all in the quote.

It is a "normal boost". Just because AMD boosts differently to Intel, doesn't mean it's not normal. Intel laptops just have a very different approach to boosting. While AMD slowly try and max out thermal saturation of a chassis over a 5 minute period before mating thermal dissipation of the device afterwards, Tiger Lake-H's base specification of 107W PL2 and 45W PL1 indicates that in that 56s PL2 they hit thermal saturation and then match thermal dissipation afterwards with PL1.

If you're trying to make out a point about IPC, then I already posted the SPEC2017 numbers above. And they clearly show TGL-H exhibiting about 10% worse 1T performance despite the 5GHz rated boost vs the 5950Xs 4.9GHz rated boost. Tiger Lake's IPC isn't anything special. It's right between Cezanne and Vermeer - or in other words exactly as expected. Perhaps a little worse, was expecting it to be closer to Vermeer, but oh well.