Quote from: ttlim on September 03, 2020, 15:38:34

Quote from: vertigo on September 02, 2020, 23:19:32
If I've misinterpreted or otherwise misunderstood something, please let me know, but as of right now, based on this article I'll probably be looking for an AMD laptop over TL, all else being equal.

We need to evaluate TGL as a total platform solution, rather just Cinebench multicore benchmark.

TGL laptop performance per watt, battery life, responsiveness, stability, AI, GNA, the ability to power multi 4k monitors, Wifi6.....all this has to be examine in totality.

Is Cinebench a good gauge meter for our daily Chrome browser, Word, Outlook, Python programming....usage?

True, and I mentioned that in my second post, but it also proves my point: the big picture is what matters, but Intel themselves are showcasing TL by doing benchmarks that compare raw CPU/GPU performance. It's in their best interest to put their best foot forward, and they're doing so by showing this narrow picture instead of the big picture, further exacerbated by the fact the benchmarks are comparing TL in a (likely ideal) test system to CL/IL/Ryzen in (often compromising) production systems. Yes, all the other stuff has to be examined and taken into account, too, but we don't have that right now. All we have is this, and it's not very impressive. If it turns out battery life hasn't improved or, as was sometimes the case with CL/IL, actually gets worse over the previous generation, then it will be worse than unimpressive, it'll be sad. OTOH, if they manage to improve battery life by say 10-20% (almost certainly unrealistic, but I can hope) in addition to these meager performance gains, then I'll be excited. But as it is, just going off what info is currently available, TL appears to be roughly equivalent to current-gen AMD offerings, and I'd personally bet that Cezanne will outperform it similar to, if not more than, Renoir outperforms CL/IL.

And, of course, since we have to consider the whole package, we have to take cost into account, and most people don't want to pay a couple hundred dollars more for a chip that performs the same or worse. Even if Cezanne only manages to match TL (which I would be very surprised if that's all they manage), it would still be better based on cost, especially if they have USB 4 with Thunderbolt. Of course, TL may offer certain unique extras that some people prefer, and for those people they'll go with it because of that, but most people will be better served with Ryzen.

Quote from: vertigo on September 02, 2020, 23:19:32
If I've misinterpreted or otherwise misunderstood something, please let me know, but as of right now, based on this article I'll probably be looking for an AMD laptop over TL, all else being equal.

We need to evaluate TGL as a total platform solution, rather just Cinebench multicore benchmark.

TGL laptop performance per watt, battery life, responsiveness, stability, AI, GNA, the ability to power multi 4k monitors, Wifi6.....all this has to be examine in totality.

Is Cinebench a good gauge meter for our daily Chrome browser, Word, Outlook, Python programming....usage?

Quote from: Mate on September 03, 2020, 09:04:49
I dont think we will see 8cores 15W CPUs from Intel anytime soon. You cant just pack 100 cores in chip and set TDP to 15W. All cores require energy and chip designer needs to make sure single-core performance is not crippled(which can happen if you have too many cores with too low TDP). From what I read AMD planned Renoir lineup to be max 6 cores CPUs but they added 2 cores after they realized that TSMC manufacturing process is very energy efficient.   Maybe Intel will add 2 more cores, but 4 will be clearly too much.

28W TGL can probably use another set of cores so it doesn't run in the inefficient region during the 28s 50W multithreaded turbo. When the power budget decreases to 15W, the 4 cores aren't deep into the inefficient region so adding more cores has diminishing returns.

Renoir has the same issue with 8C in 15W, but you just don't see it in short benchmark runs because they do 40W for 30s and then 25W+ for 5 minutes before finally dropping down to 15W.

I dont think we will see 8cores 15W CPUs from Intel anytime soon. You cant just pack 100 cores in chip and set TDP to 15W. All cores require energy and chip designer needs to make sure single-core performance is not crippled(which can happen if you have too many cores with too low TDP). From what I read AMD planned Renoir lineup to be max 6 cores CPUs but they added 2 cores after they realized that TSMC manufacturing process is very energy efficient.   Maybe Intel will add 2 more cores, but 4 will be clearly too much.

@Mate Yeah you make a good point on the TDP. I'm guessing Intel will come along with an 8-core equivalent at some point, but until then AMD definitely has the higher core segment down. I'm not an Intel fanboy by any means, but just pointing out that # of cores is probably the biggest factor in multithreading benchmarks.

@frisbfreek
Its  not unfair if 4cores and 8cores have similar TDP. Its not AMD fault that they use better manufacturing process. 

Anyway single core performance now makes difference only in some games on engines that put too many tasks on 1 or 2 threads.  Everything else is 100% about multicore performance.

Quote from: frisbfreek on September 02, 2020, 22:59:59Sure, no single metric tells the whole story. But it's also quite unfair to compare a 4-core CPU with an 8-core CPU using a multi-threading metric. If your workflow is multi-threading intensive, then you'll obviously go with more cores, and I don't think anyone would be telling you to get these Tiger Lakes over a 4800U.

So it's a fair comparison when Intel compares its own offerings to 8-core parts but when WE compare it, suddenly it's unfair? Huh?

You can't pick and choose when and where you want to compare. You compare across the board or not at all. Single core performance largely doesn't matter anymore. Since the launch of Ryzen at least, AMD has been within spitting distance of Intel's offerings, and yes, while Intel still has an edge on desktops with the 10-series CPUs, the price premium is staggering for a marginal increase in games and applications. Mind you that's only on a core-for-core basis too.

In a price for price battle, AMD smokes Intel and it's why The vast majority of DIY builders go AMD. They're just better value for what you get. It's no longer AMD for cheap options and Intel for performance options. No different with the mobile market, just so happens that AMD can offer better performance at vastly lower prices because of their designs. :p

I should also add that I focused my reply on performance since that's (unfortunately) what this and most articles and reviews focus on. We're at the point, though, and have been for a while, that performance is more than enough for most people, and those that really need more can, and probably will in many cases, get laptops with desktop CPUs. What the supposed mobile chips really need to be focusing on at this point is battery life, which has been pretty stagnant save for a few notable exceptions for several years. I'd rather have a CPU with last-gen performance (or even that of two generations ago) that allows a laptop to get an extra couple hours of runtime. Though, of course, what would be ideal is if they could do that while getting the current levels of performance when plugged in. I use a laptop for portability, as do a lot of people, and don't want to have to plug in every 3-6 hours.

Color me very unimpressed. Yet another modest improvement combined with Intel apparently still insisting on overly complicated SKUs, only now they seem to be making it harder for consumers to research computers based on the processor. Here's my translation of the article, which seems (intentionally?) worded to put the middling improvements in their best light:

"In Cinebench R15 (CB15), the single-thread performance gains are immediately evident with the Core i7-1165G7 bagging a 24% higher score compared to the Ice Lake Core i7-1065G7. Both the Core i7-1165G7 and the Core i5-1135G7 are faster (15% and 7%, respectively) than the Comet Lake-U Core i7-10710U, which is actually a 6C/12T part. In fact, the Core i7-1165G7 easily overtakes both the AMD Ryzen 5 4500U and the Ryzen 7 4700U with leads of up to 24% in single-core."

When comparing Intel's brand-new, yet to be released, top of the line CPU with the slower i7 of the previous gen (Ice Lake has less cores and slower CPU performance in exchange for better GPU performance), it's a whopping 24% faster. But that's not really saying much, and the next sentence is more representative of typical gen-to-gen improvements with Intel, with a more modest improvement of 15%, though, to be fair, that's actually better than usual, especially considering the Comet Lake is a hexa-core chip. It then goes on to say that this brand-new (and, once again, yet to be released) chip is up to 24% faster than the Ryzen 5 and 7 that have been out for months and, the way it's worded, I read as being up to (as in probably often less, but one test had this much difference) 24% faster than the Ryzen 5, AMD's mid-tier chip, whereas the gains over the Ryzen 7 are probably a fair bit less. I would hope their cutting edge i7 could beat a months-old Ryzen 5 by a good bit; that's nothing to really brag about. And I suspect it's only marginally better than the Ryzen 7, and that's for single-threaded performance, so multi-threaded it'll probably be pretty close.

"Understandably though, the Core i7-10710U leads in CB15 multi-core but not by a significant margin — it is only 8% faster than the Core i7-1165G7 and the latter is in turn just 3% slower than the Ryzen 7 4700U, which is an 8C/8T part. Good gains are also seen with the Core i5-1135G7, which seems to be about 14% faster than the Ice Lake Core i5-1035G7 and 22% faster than the Core i7-1065G7. "

So the brand-new i7 is 8% slower than the last-gen i7 (granted, it has two less cores, but the simple fact is "upgrading" to Tiger Lake from Comet Lake will result in a decrease in performance, regardless of why) and, as I suspected with my earlier statement, it's close to (and actually behind) the (again, *months old*) Ryzen 7. So it's essentially (maybe) slightly faster in single-threaded performance and slightly slower in multi-threaded performance vs an older chip. And I'm guessing the last statement in that paragraph is mixed up, since it makes no sense that it would be 14% faster than the Ice Lake i5 and 22% faster than the i7. Assuming it's 22% faster than the i5 and 14% faster than the i7, I do have to say that's pretty respectable, however, it's just not good enough. Improving significantly over a poorly performing product isn't anything to get excited about. If you have a low baseline, a large improvement can still result in a mediocre product, which is exactly the case here. Yes, they appear to have improved quite a bit from Comet/Ice Lake, but they're still only at the level of the current-gen Ryzen chips, not exactly newsworthy.

Things get a little better at this point, with the TL i5 outperforming the R7 by 10% in single-threaded performance, which is quite impressive. I don't know what the difference is between R15 (CB15) and R20 (CB20) tests, and why the latter seems to show much better single-threaded gains (or maybe the prior also does, but that's not shown by how it's worded, as mentioned before). But then we get to multi-threaded performance and, once again, TL actually loses (again) to Ryzen and Comet Lake. Then we get to 3DMark 11, where TL finally shows some real promise, actually besting the R7 by 18%. Of course, when you actually look at the graphic, you see that while it beats the R7 4700u, it loses to the (again, months old) R7 4800u. So still not impressed.

On the graphics side, it does look like TL supercedes Ryzen but, again, not by a significant margin, nothing groundbreaking anyways.

A couple other things to keep in mind with all of this: 1) while it's possible performance will improve as it gets closer to production, these benchmarks are comparing TL in a (probably optimized, with excellent cooling) test platform vs Comet Lake, Ice Lake, and Ryzen in production systems, most/all of which limit the full capabilities of the chips due to insufficent cooling, design (i.e. making them run at lower TDP to keep heat down or battery life up), or crappy firmware/drivers; and 2) even with possible gains over current Ryzen chips, that will likely only last a couple months before AMD takes back the crown, not to mention almost certainly continuing to be cheaper as well. It really seems to me that we are back to how things were when the Athlon 64 came out, with AMD crushing Intel and Intel struggling just to keep up. Hopefully this time AMD will be able to hold on to the lead, at least for longer than before, and hopefully if/when they do lose it, they'll at least be able to stay closer to Intel.

If I've misinterpreted or otherwise misunderstood something, please let me know, but as of right now, based on this article I'll probably be looking for an AMD laptop over TL, all else being equal.

Quote from: A on September 02, 2020, 22:23:41
24% improvement in single thread is not 24% improvement in everything. More and more apps are making use of multithreads and people have stuff running in the background as well.

It might be close to the 4700U in multithreading but it is far behind the 4800U in multithreading. Like over 50%.

Sure, no single metric tells the whole story. But it's also quite unfair to compare a 4-core CPU with an 8-core CPU using a multi-threading metric. If your workflow is multi-threading intensive, then you'll obviously go with more cores, and I don't think anyone would be telling you to get these Tiger Lakes over a 4800U.

That's pretty pathetic... so much for beating Renoir by a wide margin, and it pretty much matches my predictions.

Quote from: sigh on September 02, 2020, 21:21:37

Quote from: toven on September 02, 2020, 20:51:06
Why compare 27watt chip to 15watt? AMD 4800HS is cheaper and could be drawing similar power with this i7.

15W TDP on Renoir is probably the most successful bamboozle AMD has pulled.

Check the 4800U benchmark .The laptop is pulling 50W in stress tests until it thermal soaks and loses 30% of its performance.

What do you think Intel has been doing? There is a reason Intel's mobile CPUs are using more power than desktop cpus.

Quote from: DavidC1 on September 02, 2020, 21:06:45

Quote from: Intel is in trouble. on September 02, 2020, 20:38:20
Not impressed.

Only 24% in single threaded load compared to previous gen and renoir. But we already know now that cezanne will be 20% faster in integer.

So what? Then Alderlake comes a few months later and beats Cezanne.

Also, the 24% advantage in ST means it'll be 24% faster in everything. This is why its so close to 4700U with 8 cores without HT. Any workload that doesn't go over 8 threads or benefits fully from greater than 8 threads the 24% advantage will come into play.

24% improvement in single thread is not 24% improvement in everything. More and more apps are making use of multithreads and people have stuff running in the background as well.

It might be close to the 4700U in multithreading but it is far behind the 4800U in multithreading. Like over 50%.

Quote from: toven on September 02, 2020, 20:51:06
Why compare 27watt chip to 15watt? AMD 4800HS is cheaper and could be drawing similar power with this i7.

15W TDP on Renoir is probably the most successful bamboozle AMD has pulled.

Check the 4800U benchmark .The laptop is pulling 50W in stress tests until it thermal soaks and loses 30% of its performance.

QuoteHDMI 2.0b comes as default but support for HDMI 2.1 for 8K60 output is possible via a DP 1.4 bridge.

Does anyone know what this means exactly? Does this mean a laptop can potentially have a bandwidth-gimped (DP 1.4 bandwidth) HDMI 2.1 port?

Quote from: Intel is in trouble. on September 02, 2020, 20:38:20
Not impressed.

Only 24% in single threaded load compared to previous gen and renoir. But we already know now that cezanne will be 20% faster in integer.

So what? Then Alderlake comes a few months later and beats Cezanne.

Also, the 24% advantage in ST means it'll be 24% faster in everything. This is why its so close to 4700U with 8 cores without HT. Any workload that doesn't go over 8 threads or benefits fully from greater than 8 threads the 24% advantage will come into play.

If there won't be laptops with full (not that 2 lanes crippled trash) speed Thunderbolt outside of the premium segment then Intel could go to hell with their 4 core with fairy 28W TDP.