News Tiger Lake Tested: We Benchmark Intel’s Latest With Iris Xe Graphics and 10nm SuperFin

[rgd1101]

nope. dislike synthetic test. real world app would be better.
like the mathlab test. or game if that what matter.

on the power thing
anandtech did it better https://www.anandtech.com/show/16084/intel-tiger-lake-review-deep-dive-core-11th-gen/7

 

[st379]

I suppose you think that Cinebench is way better, right? In any case, regardless of what you think of Geekbench you should see how much the score of the 4800U drops when unplugged relative to its own score when plugged. Unplugged it drops to 59.2% of its plugged score. A drop in performance of over 40%. Here comes AMD’s shenanigans. You see when unplugged AMD’s frequency plummets in order for their laptops to look good in battery duration benchmarks. But when plugged they draw way way more power in order to look good on performance benchmarks (which AMD knows reviewers are only conducting when plugged in).

Now about that TDP of AMD’s cpus. That’s a complete farce. You see AMD coined a different term to refer to their real TDP: Package Power Tracking (PPT). In the desktop processors this was set to 88W for the “65W TDP” cpus and 142W for the “105W TDP” cpus. That’s a bit over 35% in each case compared to the advertised TDP. This PPT is a sustained power limit for an indefinite amount of time by the way (provided the cooling is not a limit). This is like Intel’s PL1. In laptops AMD went a step further and coined PPT_long and PPT_short. The former is like Intel’s PL1 and the latter is like Intel’s PL2. PPT_short is sustained for a short period of time similar to how Intel’s PL2 is sustained for a short period of τ seconds. For the 4800U there are two “TDP” modes: “15W” and “25W”. When on “15W” the PPT_long (i.e. real tdp) is 25W. The PPT_short is something higher (~45W). When on 25W they have a PPT_long (i.e. real tdp) of 35W. The PPT_short is 63W.

So all the AMD fbs saying that this article may have tested the 4800U only in 15W mode should know that even in that case that would be the equivalent of Intel’s 25W mode and the fair thing to do. And when the 4800U is set to performance mode it is equivalent to an Intel’s 35W mode.

Also watch this where it clearly states that the 4800U is a 35W tdp cpu.

View: https://www.youtube.com/watch?v=2yl_ziFcOA4&feature=youtu.be&ab_channel=HardwareCanucks

"Running Prime95 reveals more notable differences between the Ryzen 7 4800U and the competing Core i7-1065G7 or Core i7-10710U as shown by our graphs below. Power consumption would spike to 57.5 W on our Lenovo before falling and stabilizing at 49.6 W compared to 64.4 W and 42.8 W on the Ice Lake-powered XPS 13 2-in-1. " (notebookcheck.net)

64.4=15. Did you finish elementary school?

Since when 64.4(Intel)=15? Where can I learn that kind of equation?

 

[rgd1101]

"Running Prime95 reveals more notable differences between the Ryzen 7 4800U and the competing Core i7-1065G7 or Core i7-10710U as shown by our graphs below. Power consumption would spike to 57.5 W on our Lenovo before falling and stabilizing at 49.6 W compared to 64.4 W and 42.8 W on the Ice Lake-powered XPS 13 2-in-1. " (notebbokcheck.net)

64.4=15. Did you finish elementary school?

Since when 64.4(Intel)=15? Where can I learn that kind of equation?

AMD does it too.
Tiger lake is better on the power consumption compare to Ice lake .

 

[st379]

AMD does it too.
Tiger lake is better on the power consumption compare to Ice lake .

I did not say they did not. Why not if it does not throttle?

 

[rgd1101]

That why it hard to compare even on the "same" wattage.

Maybe they should start doing what anandtech did.
have the time/score with joules, average w, and maybe temp too.

Did anyone see a review with them open this ref laptop and checkout the cooler inside?

 

[st379]

That why it hard to compare even on the "same" wattage.

Maybe they should start doing what anandtech did.
have the time/score with joules, average w, and maybe temp too.

Did anyone see a review with them open this ref laptop and checkout the cooler inside?

Intel had lots of strict rules on this "review".
What you can and can't do.
Would love to see the ddr speed and timing from a realiable app.

Anyway we can learn a couple of things abouth this CPU:

1. With the right ram the gpu could be quite powerful.
2. The cpu is ~100% slower in certain tasks because of 50% less cores.

The truth an 8 cores of this would give Amd quite a knockout.

Typical Intel being greedy and not releasing 8 cores because they have massive fanboys base.

 

[GetSmart]

Can find memory timings from one the review URLs I've posted earlier. Like this one from Notebookcheck 🆒

 

[st379]

It looks like the ram downclock itself? and where is the Amd?
So not answering my question but thanks.

 

[spongiemaster]

Intel had lots of strict rules on this "review".

No, they didn't. There were only 2 conditions. Preproductions units were sent out, so reviewers were not allowed to open the chassis and show what was inside. The 2nd was that reviewers couldn't test battery life because the preproduction unit had a smaller battery than the final production unit would have.

 

[st379]

No, they didn't. There were only 2 conditions. Preproductions units were sent out, so reviewers were not allowed to open the chassis and show what was inside. The 2nd was that reviewers couldn't test battery life because the preproduction unit had a smaller battery than the final production unit would have.

So why so many bad reviews?
No ram informatin, what is dynamic tuning some kind of PBO?
What the clocks like? Boost time?
Really disappointing to see the review on this site and Anandtech.
No productivity when finally Intel get rid of Skylake and so many more.......

 

[GetSmart]

It looks like the ram downclock itself? and where is the Amd?

So not answering my question but thanks.

Did you even bother to check the links? Everything is there, here is the link again https://www.notebookcheck.net/Intel...iew-First-Tiger-Lake-Benchmarks.494462.0.html The compared devices are clickable and will redirect to the original review of the device itself.

 

[spongiemaster]

So why so many bad reviews?
No ram informatin, what is dynamic tuning some kind of PBO?
What the clocks like? Boost time?
Really disappointing to see the review on this site and Anandtech.
No productivity when finally Intel get rid of Skylake and so many more.......

None of the major sites specialize in laptop reviews. You need to look a mobile specific sites for more detailed reviews. Almost everything you listed I saw in the few reviews I browsed through. If something was missed, it wasn't because of Intel, outside of the 2 restrictions I mentioned. Sorry if this goes against your everything Intel does sucks message.

 

[PCWarrior]

Funny that unplugged, the Ryzen 4800U has a lower single core score (675) than a 2019 phone! Even the Samsung Galaxy S10 outperforms it (691)! Yet some try to present this sorry piece of AMD s...ilicon as the best thing ever.

Oh boy, it gets worse. It turns out that even the 2018 Samsung Galaxy S9 with an Exynos 9810 beats the unplugged Ryzen 4800U! Who knew that a 2.5 year phone could beat AMD's flagship 2020 U-series laptop.

 

[GetSmart]

Hot Hardware also had a preview https://hothardware.com/reviews/intel-tiger-lake-benchmarks-first-look and they are comparing against the fastest AMD Renoir based laptop (with AMD Ryzen 9 4900HS) they have tested from their previous review here https://hothardware.com/reviews/amd-ryzen-4000-asus-rog-zephyrus-g14-notebook-review (this link is within the Intel Tiger Lake preview itself) :

 

[TerryLaze]

I actually like what I see form the other review. but hold off for the suppose 8 cores that is coming. but some intel fanboy would pick them without reading review(s) and just cherry picking data.

And look like all these review are from intel reference laptop. wonder what cooling like and what the laptop weight

Cherry picking is the whole reason that benchmarks exists, you look at all of them and decide what is most probable to benefit you.
You buy a thin and light, what is your most probable use case:

1. running cinema 4D all day on the go on the battery...
2. being an youtuber/influencer what have you and having to apply tons of filters on videos and pictures...
3. being a businessman having to run a portable office from anywhere.

You buy accordingly.
Intel is giving the new generation of internet peoples exactly the things they need right now, nobody is going to run anything multicore on a thin and light, people want something that doesn't weigh a lot and that can do what they need, they don't care if it can or can't do something they couldn't care less about.

 

[rgd1101]

Oh boy, it gets worse. It turns out that even the 2018 Samsung Galaxy S9 with an Exynos 9810 beats the unplugged Ryzen 4800U! Who knew that a 2.5 year phone could beat AMD's flagship 2020 U-series laptop.

You do know they don't test the same suit because they don't run the same.
Now, I'm thinking you are just trolling

 

[PCWarrior]

You do know they don't test the same suit because they don't run the same.
Now, I'm thinking you are just trolling

What do you mean they don’t test the same suit? They do test the same suit! Geekbench provides a full breakdown of the several tests it performs with individual scores per test. Here is a Geekbench page for a test on the 1185G7. And here is for a test on the Galaxy S10 5G. It is the same 21 tests - optimised and compiled for the different platforms (ARM/Android, x86/Windows). The final score is a weighted average of the several sub tests and is calibrated against the score of the 8100 having a score of 1000. The Geekbench tests on the S9 and S10 were performed yesterday on the latest version of Geekbench. Unplugged. The Exynos 9810 chip on the S9 scored 688 and the Exynos 9820 on the S10 5G scored 818. Meanwhile the unplugged 4800U scored 675…I am just highlighting this because I find it remarkable that, when unplugged, the overhyped 4800U which is being presented by some as the" king of 15W” laptop cpus loses to 2.5year old mobile phone cpu. It goes on to show how much AMD is cheating on the battery life tests and performance benchmarks.

Here is what happens when plugged - truly smoke and mirrors.

As for energy tests. Funny how Ian wants to bring them now after several years of bushing Intel on power. For reference below is how the 9900K/KS and the 10900 fair on energy tests from Techpower up. Didn't see Ian calling these chips efficient then.

 

[hotaru.hino]

Here is what happens when plugged - truly smoke and mirrors.

As for energy tests. Funny how Ian wants to bring them now after several years of bushing Intel on power. For reference below is how the 9900K/KS and the 10900 fair on energy tests from Techpower up. Didn't see Ian calling these chips efficient then.

To appreciate the first graph correctly, you have to take the area "under the curve" for all three of them. Yes the 4800U (I'm assuming it's a 4800U) has a higher energy usage, but it also completes the task in almost half the time. Eye balling it, Ice/Tiger Lake may have small advantage here, but they're both within spitting distance of the 4800U.

The last two graphs I don't find meaningful unless there's a time component with it.

 

[PCWarrior]

To appreciate the first graph correctly, you have to take the area "under the curve" for all three of them. Yes the 4800U (I'm assuming it's a 4800U) has a higher energy usage, but it also completes the task in almost half the time. Eye balling it, Ice/Tiger Lake may have small advantage here, but they're both within spitting distance of the 4800U.

The last two graphs I don't find meaningful unless there's a time component with it.

Yes of course there is a time component in those Techpowerup charts. It is energy consumption test. Energy is Powertime. It is measured in Joules (J) and 1 Joule = 1 watt x 1 second. And 1kJ=1KW1s.

The point I was making is that bringing up the energy consumption argument as a counter to the power consumption is at least hypocritical. The same people that now defend AMD with energy considerations by taking the integral of the power curve to calculate energy are the same people who just looked at the power figure (and mind you when running an unrealistic stress test such as Prime 95) and bushed Intel about the high power consumption. But as shown above Intel’s CPU that were bushed by the media (including Anandtech) for being too power hungry, in fact consumed less energy than their AMD counterparts in order to complete the same tests (the single core one is Super Pi and the multicore one is Cinebench). Or in the worst cases they were in the same ballbark as those of AMD. So you wouldn’t pay more electricity to run the same tasks on the Intel cpus over AMD cpus as it was the common myth.

Now of course there is value to power consumption considerations alone, however. And that’s thermals. That’s why power is tied with TDP (thermal design profile). You see when you consider heat transfer you can draw an equivalent thermal circuit with the following equivalencies:

Power (P) acts like current (I). The cpu acts like an ideal current source with current P.

Temperature difference (ΔT) acts like voltage (V)

Thermal impedance (Zth) acts like electrical impedance (R/Z) and can be modelled as an RC network (of equivalent thermal resistances and thermal capacitances).

In the simplest model where we just consider the thermal resistance only you get ΔΤ=P*Rth (in the same way Voltage=Current * Resistance in electrical circuits). For a given cooler design and thermal interface materials, Rth is fixed. Then power (not energy) is what determines operating temperatures. Working it in reverse is how TDP is calculated. You set limits on the operating temperature, assume an ambient temperature and work back to find the maximum power consumption envelope your cpu needs to stay in. And knowing that your cpu will consme that much power you design your cooling solution so that your cpu stays below a certain maximum temperature. AMD is using shenanigans however. Since tdp=P=ΔΤ/Rth they massage the numbers. They set ΔΤ and Rth to whatever values would make their quotient be whatever marketing value they want for tdp.

 

[TerryLaze]

What do you mean they don’t test the same suit? They do test the same suit! Geekbench provides a full breakdown of the several tests it performs with individual scores per test.

They are not running the x86 version is, I think, what he meant.
Running something that is specifically made for arm is always going to run better but try to run something x86 specific on arm and it will bomb hard while running arm on x86 has virtually no slow down.

 

[GetSmart]

Like to add that, Geekbench favors ARM architecture hence often the strange and higher results. Some of those sub-tests runs extremely well on ARM cores..

 

[PCWarrior]

They are not running the x86 version is, I think, what he meant.
Running something that is specifically made for arm is always going to run better but try to run something x86 specific on arm and it will bomb hard while running arm on x86 has virtually no slow down.

You are probably referring to running Windows .exe applications on Windows for ARM. But I don’t see the relevancy here. There is absolutely no reason to use the Geekbench binary compiled for x86/Windows and run it on ARM/Android through some sort of emulation or binary translation. The Geekbench source code is cross compiled for the various operating systems and instruction set architectures and there is a native binary for each platfrom. It is literally the same code. That’s the whole freaking point of using a cross-platform benchmark in the first place. To remove artificial handicaps and fairly compare the performance of cpus for apps that were programmed to run on them natively. what this tells me for example is is that if I use a Samsung S10 phone to browse the web (native apps obviously) it will offer me a better experience than an unplugged Ryzen 4800U (native apps). Isn't that a shameful showing for AMD?

 

[rgd1101]

compiled

that the different. on the same cpu. linux usually score better than windows. and it back to the same synthetic test.

I read the it best to use geekbench compare the same architect, not cross system.

 

[TJ Hooker]

In any case, regardless of what you think of Geekbench you should see how much the score of the 4800U drops when unplugged relative to its own score when plugged. Unplugged it drops to 59.2% of its plugged score. A drop in performance of over 40%. Here comes AMD’s shenanigans. You see when unplugged AMD’s frequency plummets in order for their laptops to look good in battery duration benchmarks. But when plugged they draw way way more power in order to look good on performance benchmarks (which AMD knows reviewers are only conducting when plugged in).

That's based on the power management implemented for the particular laptop, it's not inherent to the CPU.

It goes on to show how much AMD is cheating on the battery life tests and performance benchmarks.

Here is what happens when plugged - truly smoke and mirrors.

Both Intel and AMD exceed their stated TDP while boosting, that's not a secret. Where's the "smoke and mirrors"?

Here are the energy results to go along with that power graph:

• 15 W Renoir: 2842 seconds for 62660 joules
• 15 W Tiger Lake: 4311 seconds for 64854 joules

So the 4800U has lower peak power and used slightly less total energy, while completing the task in 33% less time.

But again, for either CPU those numbers can vary based on the cooling and power management implemented in the given laptop. If the CPU is allowed to draw more power it'll score better but also likely have worse efficiency.

As for energy tests. Funny how Ian wants to bring them now after several years of bushing Intel on power. For reference below is how the 9900K/KS and the 10900 fair on energy tests from Techpower up. Didn't see Ian calling these chips efficient then.

Posting those results in isolation is just as misleading here as the last time you posted them. My response is the same

If you're going to look at performance, power, and energy usage, look at all three from the same benchmark (if possible). The way you're bouncing around citing individual metrics from different tests is just cherry picking. Especially including efficiency results from desktop CPUs in a discussion of mobile chips.

 

[TJ Hooker]

what this tells me for example is is that if I use a Samsung S10 phone to browse the web (native apps obviously) it will offer me a better experience than an unplugged Ryzen 4800U (native apps). Isn't that a shameful showing for AMD?

No, what it tells you is that you run a bunch of largely synthetic tests and combine all the scores in an arbitrary way to arrive at a single number, that number will be lower for a unplugged Ideapad Slim 7 w/ 4800U than an S10.