Tiger Lake Iris Xe G7 and GeForce MX350 are neck-to-neck in 3DMark — so why is the Intel chip still so much slower when gaming?

[Redaktion]

Unoptimized drivers, a shared TDP pool, and lack of VRAM on the Iris Xe G7 may be attributing to some steep performance deficits when compared to the GeForce MX150, MX250, or MX350 series when running certain games. A select few titles, however, show the Intel chip to be on par with Nvidia.

https://www.notebookcheck.net/Tiger-Lake-Iris-Xe-G7-and-GeForce-MX350-are-neck-to-neck-in-3DMark-so-why-is-the-Intel-chip-still-so-much-slower-when-gaming.496466.0.html

[Arthon]

So now intel fan boys + nvidia haters will say ( nvidia paid article ) ?

i think NBC now clear all the things , no need to cry now.

i will not take iGPU and waiting 2y to optimize it , just take discrete GPU with its VRam and good driver .

thank you NBC for clearing this for us .

[mik]

The Zenbook suffers from thermal throttling, PL2 goes down to 18W under sustained load. As a result Xe LP runs at 850 Mhz only which is 50% under Intels nominal specification. At 28W it shouldn't have issues matching or beat MX250/MX350 but needs proper cooling. Many of these thin and light might have troubles with cooling. And furthermore the driver they did use was outdated.

[Arthon]

The Zenbook suffers from thermal throttling, PL2 goes down to 18W under sustained load. As a result Xe LP runs at 850 Mhz only which is 50% under Intels nominal specification. At 28W it shouldn't have issues matching or beat MX250/MX350 but needs proper cooling. Many of these thin and light might have troubles with cooling. And furthermore the driver they did use was outdated.

can u te me wich model of zenbook that have this issue ?
if u have it , tel me more about it , i`m gona get one soon and all reviews say its not haveing major issues.

thanks.

[Watzupken]

I feel there is a lot of hype around how powerful Tiger Lake (including Xe graphics) is. But so far I've yet to see a review that concludes this. The iGPU is a big improvement, I give Intel the credit here. I am not at all surprised that the thermals and shared resources will hamper performance though. When both CPU and iGPU are under load in games, they generally don't boost as high or even under clock to stay within thermal and power limitations. So it is unlikely to outdo a dedicated GPU solution in most cases.

[Watzupken]

The Zenbook suffers from thermal throttling, PL2 goes down to 18W under sustained load. As a result Xe LP runs at 850 Mhz only which is 50% under Intels nominal specification. At 28W it shouldn't have issues matching or beat MX250/MX350 but needs proper cooling. Many of these thin and light might have troubles with cooling. And furthermore the driver they did use was outdated.

The problem is that poor cooling is not an uncommon problem. Just as you rightly pointed out, slim chassis is going to pose a problem when it comes to cooling. So yes it can perform better with more power, you will need to maintain a high TDP in order to do so. In fact, I don't think 28W is sufficient as well.

[DavidC1]

This is clickbait from NBC.

I suspect they have to resort to clickbait because the one advantage of their highly detailed articles is that it costs them a lot to maintain. So any ad revenue is necessary.

Look at the Swift 5. It uses the same TDP and it performs 25% better in 3DMark and 35-50% faster in games. That article makes a completely different conclusion, as how the MX350 is not needed and it gives them hope Iris Xe can perform in games as well as it does in 3DMark against the MX350.

Mind you this isn't even full performance. 25W devices like the Surface, XPS, Spectre will do even better.

[Spunjji]

Let's be honest: is anyone outside of Intel fanboys surprised that they'd drop a massive fart when it comes to drivers?

They have never, *ever* been able to extract anything like their claimed performance in actual games. Unflattering performance - along with the weird texture issues, pogo frame-rates and compatibility issues - should have been what anyone familiar with their products would be expecting.

[vertigo]

This is clickbait from NBC.

I suspect they have to resort to clickbait because the one advantage of their highly detailed articles is that it costs them a lot to maintain. So any ad revenue is necessary.

Look at the Swift 5. It uses the same TDP and it performs 25% better in 3DMark and 35-50% faster in games. That article makes a completely different conclusion, as how the MX350 is not needed and it gives them hope Iris Xe can perform in games as well as it does in 3DMark against the MX350.

Mind you this isn't even full performance. 25W devices like the Surface, XPS, Spectre will do even better.

The Swift 5 review and the (limited) gaming benchmarks in it do show that TL can do very well, but this review tests multiple TL laptops and shows that not to be the typical case. What all this says to me is that in order for it to get good performance, the computer design has to be optimal, further evidenced by the fact that in that review, even the Swift 5 is significantly behind the Intel reference computer, which just further proves my earlier concerns about that being the perfect, unrealistic scenario and real life results likely being nowhere near it.

So yes, Tiger Lake appears to have a good bit of potential, but that is far too reliant on the OEM doing a very good design job in order to provide adequate cooling, drivers, etc, and let's face it, that's not very common. Is that Intel's fault? Yes and no. It is because they built a (possibly) good CPU that is too sensitive to the computer it's installed in, whereas AMD clearly did a better job in that respect (based on the average AMD performance being better than the average Intel performance, as the average of multiple units is more important than one cherry-picked one, and you could probably find an excellent Ryzen performer that would match or exceed the TL Swift 5), and because they aren't exercising enough leverage over the OEMs to get them to use their chips in an actually good design.

Granted, Evo is potentially a step in the right direction here, but Intel would rather just "encourage" OEMs in various ways to use their chips over the competition without actually caring about the end result or the consumer experience; all they care about is that their chips are being used/sold, not that they actually perform as they should or do better than an AMD chip. They could easily require OEMs to meet certain performance requirements so any TL-equipped laptop would perform at a similar level as the Swift 5, but they don't, so you end up with a large spread of performance, with most performing significantly worse, dragging down the average and resulting in this article. Whereas AMD, despite OEMs seeming to intentionally try to handicap laptops containing their chips, does much better on average and seems to have less of a performance spread.