My T430s died. I was holding out for TGL-H45 (8C, 45W) to get a replacement, but the old machine decided to give up the ghost about half a year (or more) too early for that to be available.
Since I really needed something semi-respectable (my i3-3217U NUC really isn't at this day and age) I decided to hunt down the cheapest laptop that I could find running a Tiger Lake i7.
The result of this search is the ThinkBook 14s Yoga ITL, the latest Intel version of the ThinkBook line. At 908$ for my configuration (16GB of RAM, 512GB NVMe SSD), this machine is cheap, well-built, well-cooled and good looking. Also, did I mention this an i7-1165G7 for 900$? Yeah. The only cheaper machine I could find was an Acer, but it only had 8GB of non-upgrade-able RAM, which made it a poor choice.
You can see some better pics than I could ever take on the notebookcheck review for this laptop (I even managed to get the same "limited edition" version in the "abyss blue" color, which is rather pretty). Overall, about the only bad thing to be said about this system is that while the RAM is upgrade-able, it is not fully upgrade-able. 8GB is soldered, and another memory slot is available. Also, due to using upgrade-able RAM, the machine uses DDR4-3200 instead of LPDDRX4266, which hurts its Intel Iris Xe GPU performance wise. For a system that will mostly be used with a docked eGPU, however, this doesn't matter much, and for the price, it is hard to beat.
I am happy to report that this system has good cooling, which keeps temps in check. At the stock fan settings and settling down to its 25-ish W TDP after the initial boost period (notebookcheck says 20W, but the CPU sits at around 25W in my testing, perhaps because I set the BIOS to "Extreme Performance" mode - I need to test this further), the CPU does not cross 75c under a prime95 blend load across all 8 threads, which is phenomenal for a system so thin sitting right on a desk. In fact, I'd argue that the TDP is too low. Once XTU supports Tiger Lake CPUs, I will try to raise the TDP limit (if possible) to 30 or 35W to allow for higher sustained clocks.
As it stands, the system is capable of fluctuating between 3500Mhz to 4000Mhz under flat-out loads within its TDP limit, which considering the 11th Gen IPC advantage is very fast for a U-class CPU (my 7th Gen i7-7820HQ in my former ZBook G4 would boost at 3500Mhz across all cores, but at 45W).
As side note, the system has two m.2 slots, and both support NVMe drives. This means that it also makes a good candidate for an m.2 eGPU setup. The bottom should be easy to mod to allow the cable out, or to make the slot externally accessible.
eGPU Setup:
The eGPU setup is painless and plug-and-play. Just plug it in, install drivers, and you are good to go. As an unexpected bonus, this system displays the BIOS via the eGPU (!), so the laptop can be placed with the lid closed, but you can still get into the BIOS (and use anything else plugged into the eGPU, such as a mouse/keyboard).
I've done a set of benchmarks between the Iris Xe iGPU and my GTX1080Ti, with the results below. I also plan to repeat my Total War benchmarks from a while ago to compare the advances in eGPU tech on a modern gaming title since my ZBook G4 test runs. The results have been normalized so that the iGPU result is 100%, with the 1080Ti result as a scaling of it. I tried to use benchmark settings which were "reasonable" for the Iris Xe and would yield framerates in the 20-60 range on it (with the exception of the FireStrike and TimeSpy benchmarks, which run much slower as I wanted to use the default settings) to gauge the eGPU performance boost more reliably because it would encapsulate the high-FPS performance hit into the results.
As promised, here is the Tiger Lake U (~25W) vs Kaby Lake H (45W) comparison in Total War: Three Kingdoms.
I no longer have my ZBook G4, which means I used the results from my previous testing. This is not quite an apples to apples comparison, so take it with a grain of salt:
The eGPU enclosure is the same, an AKiTiO Node Pro, and so is the GPU itself in it, a Palit GTX1080Ti GameRock.
If someone has Three Kingdoms and a 7th gen or 8th gen system coupled with a GTX1080Ti, we can try to replicate some of this benchmark session to see if the results are reliable.
And now, the results, and some analysis. First, the simpler percentage-based graphs.
The average FPS graph shows that with the exception of three specific combinations of settings and resolutions, the TGL U system beats the old KBL H. The outliers are the low preset at 2160p, and the ultra preset at 1440p and 2160p. These are odd outliers, but the results are repeatable. Averaging across everything, the average FPS is 5.5% higher with the TGL U. Not bad for about half the TDP.
If we look at the average differences across the three resolutions, we get:
1080p - TGL U is 15.9% faster.
1440p - TGL U is 4.5% faster.
2160p - TGL U is 2.1% slower.
Since we know that at higher resolutions the CPU and the Thunderbolt subsystem impact results less and less, since the GPU has to do more work per frame, the 1080p result is the one that showcases the difference between these systems the best and lends credence to the hypothesis that TGL is better for higher-FPS gaming, such as eSports titles.
The 2160p result is truly odd - I have no hypothesis for why the TGL U would be slower in this scenario.
The minimum FPS graph shows a more convincing picture. With a single outlier (ultra preset at 2160p), the TGL U is faster than the KBL H. Averaging across all scenarios, the minimum FPS with the TGL U is an impressive 19.4% higher. Since minimums are arguably more important than averages for the gameplay experience, this is a hefty boost.
Splitting by resolutions again, we get:
1080p - TGL U is 37.6% faster.
1440p - TGL U is 20.8% faster.
2160p - TGL U is 3.7% faster.
Despite the outliers in the average FPS, in the minimum FPS scenario the situation is clear: TGL U is the clear winner, even managing to eke out a few more percentage points at 2160p, and providing a whopping 37.6% higher minimum framerates at 1080p.
The next two graphs show the raw average and minimum FPS values for both systems:
Confirming: "Extreme Performance Mode" BIOS setting is what gives me 25W of TDP for the CPU. This laptop is not a 20W unit after all.
The cooling system is great, and I really need some way to throw more TDP at the CPU, because it is ridiculous seeing the system sit at around 70C while it has another 20C of thermal headroom under load! There is literally no reason why it cannot run at 4Ghz for as long as it takes besides the lack of additional TDP to play with.
Since some of my fellow redditors rightfully complained that the graphs are a mindfuck, here are some better readable options:
Confirming: "Extreme Performance Mode" BIOS setting is what gives me 25W of TDP for the CPU. This laptop is not a 20W unit after all.
The cooling system is great, and I really need some way to throw more TDP at the CPU, because it is ridiculous seeing the system sit at around 70C while it has another 20C of thermal headroom under load! There is literally no reason why it cannot run at 4Ghz for as long as it takes besides the lack of additional TDP to play with.
Might also depend on what the power delivery is capable of in a sustained scenario.
Would it be possible for you to confirm if one of the m.2 slots is gen 4? From doing a bit of googling it seems like one slot is gen 4 and the other is gen 3.
With Tiger Lake there are a couple of laptops with gen 4 slots (though no mention if it's x2 or x4) but it would be interesting to see how much performance will be lost by using the internal display when you have double the bandwidth with 4.0 x4.
Do you know what the performance loss of the 1080Ti is when in eGPU form compared to when it's in a desktop?
To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.
@mqyrk - Since my drives are all Gen3, I am not sure how I can confirm this.
With respect to the performance loss, you can take a look at my i7-7820HQ post (here) for some desktop comparisons I dug up online. Then just apply the performance increase of TGL U from this post, and you have a good sense of how the two stack up.
For some more benchmarks, a friend of mine enlisted his RTX3070, LG Gram (i7-8565U), as well as a pair of desktops (Intel Pentium G4560 and Ryzen 5 3600X) to do some comparisons vs my Tiger Lake setup.
We ran the built-in benchmark in Assassin's Creed: Odyssey on the same settings, and compared the systems.
The desktops obviously run the GPU in a regular full-length PCIe slot. The LG Gram runs its eGPU via a Gen3 m.2 slot. My i7-1165G7 uses the Node Pro via TB4.
The LG Gram and the dual-core Pentium manage similar framerates. The Yoga 14s manages a similar framerate on average despite having a much slower GPU, and beats them both in minimum FPS by a significant margin. The desktop Ryzen of course runs away from the pack, managing 86% more FPS than the Yoga 14s (keep in mind, the two GPUs in use are not equal).
Looking at the desktop results from the Techpowerup benchmarks, the Yoga 14s takes a 28% performance hit compared to their benchmark system (a 5Ghz i9-9900K) with a GTX1080Ti. Considering the mobile CPU and TB4 eGPU, this is actually pretty good.
@yukikaze, Thank you for this excellent build guide! Intel is definitely feeling the heat. Your ThinkBook Yoga 14s is a great value. The MSI Prestige 14 I tested recently is another good value. How is the noise level when you run with BIOS setting "Extreme Performance Mode"?
@itsage - It is audible, but not obnoxious when under stress test. It is quieter when in an actual game. I spend most of my time with a headset so I barely notice it.
Confirming: "Extreme Performance Mode" BIOS setting is what gives me 25W of TDP for the CPU. This laptop is not a 20W unit after all.
The cooling system is great, and I really need some way to throw more TDP at the CPU, because it is ridiculous seeing the system sit at around 70C while it has another 20C of thermal headroom under load! There is literally no reason why it cannot run at 4Ghz for as long as it takes besides the lack of additional TDP to play with.
Since some of my fellow redditors rightfully complained that the graphs are a mindfuck, here are some better readable options:
Thank you for the detailed review You could give Throttlestop a try. Managed to elevate the power limit from 14W short/11W long to 28W short/long in my HP Elitebook x360 1030 G3 with an i5-8250U. I'm just not sure if the TPL tool in Throttlestop already works with Tiger Lake...
You could give Throttlestop a try. Managed to elevate the power limit from 14W short/11W long to 28W short/long in my HP Elitebook x360 1030 G3 with an i5-8250U. I'm just not sure if the TPL tool in Throttlestop already works with Tiger Lake...