6 ms·

From this article: https://wccftech.com/intel-launches-tiger-lake-h-tgl-h-mobility-cpus-at-ces-2021-fastest-laptop-processor-single-thread-on-the-10nm-superfin

by Jonnax 6y ago

From this article: https://wccftech.com/intel-launches-tiger-lake-h-tgl-h-mobility-cpus-at-ces-2021-fastest-laptop-processor-single-thread-on-the-10nm-superfin-process/ https://wccftech.com/intel-launches-tiger-lake-h-tgl-h-mobil... Intel say they have the fastest single threaded performance of any laptop processor. They show a graph of a benchmark of something called SPECRATE * 2017_int_base compiled with the Intel compiler. Does the Intel compiler still hamstring AMD? Also is this benchmark useful in real world usage?

reader_mode 6y ago
Does it really matter in a 45W power range ? I have a 2018 15 MBP with an i9-8950HK that has the same power range - this thing is unusable at full load for anything but burst - and the way it taxes the thermals makes it a terrible laptop (people in the office start turning their heads). I'm not convinced any portable laptop can handle that TDP efficiently - so maybe this is interesting for gaming laptops - but to me this looks like Intel is still making shitty power-hungry chips for laptops They are boasting performance comparisons to AMD Navi 2, Navi 3 also brings a ~20% IPC improvement which they claim to have over it. I'd speculate that Ryzen 7 5800U should outperform this chip in most relevant use cases (eg. sustained load like running a simulator while developing/running build watchers or running VM/docker and IDE/compiler) while being in 25W range which can be handled reasonably at this device size. Unfortunately laptop manufacturers are so slow at adopting AMD in premium line laptops (I preorder a surface pro with an AMD 5xxx CPU). I don't see myself buying an Intel based laptop until they improve the thermals.
- AmVess 6y ago
  Apple's M1 has shown us how antiquated and far behind Intel has become. My Macbook is not only much faster than my XPS unit, but it doesn't get hot, nor does it use up the battery in 4 hours.
- bserge 6y ago
  It's the heatsink. And Apple isn't the only offender. All major manufacturers use terrible cooling systems, even on thick workstation class laptops. Their testing is seriously flawed, they design the heatsink for "normal workload" which means jack shit. Everyone's got different uses for their machines, so just test the heatsink at maximum load/TDP for several hours at least, make sure it doesn't overheat. Is that too much to ask? And if you design your laptops to be thin and light, don't shove a 6 core 45W processor in it... just don't. It will perform slower than a 4 core 35W chip, so what's the point? Having it on the spec sheet is not worth the negative feedback from buyers. Apple went the other way with the M1, it doesn't overheat because it just uses much less power, while having the same performance. Really impressive. You can fix the overheating problem by buying some good thermal paste and applying it yourself and/or by using software like Throttlestop to limit the maximum TDP (and even undervolt it to claw back some of that performance) and NotebookFanControl for the fans. Not sure if there are alternatives for the MacBook, though.
  - reader_mode 6y ago
    You can't under-volt on latest machines (mine included). I need to limit TDP to get the machine usable. I've thought about opening it up and reapplying paste but I don't want to risk bricking my work device - I don't have the time to deal with it. I will have a few weeks off soon so might consider it if the expected gains are significant but from what I saw back when I looked into it (this[1] mostly) it's not worth it. https://www.reddit.com/r/macbookpro/comments/96wwyf/so_i_replaced_the_thermal_compound_on_my_mbp_2018/ https://www.reddit.com/r/macbookpro/comments/96wwyf/so_i_rep...
    - AmVess 6y ago
      It may drop the temp a few C or delay throttling, but small heatsinks and tiny fans can only dump so much heat out of the chassis. I did all kinds of things to my XPS, but the largest gain came from undervolting and it still sounded like a leafblower when it was being worked moderately hard. My M1 Macbook on the other hand, is dead quiet on top of absolutely destroying the XPS in CPU performance while having vastly superior battery life. I don't have to run anything out of spec to keep it from cooking itself.
      - reader_mode 6y ago
        Yeah no doubt M1s are amazing, I'm kind of sad Apple is the only one with it because Apple isn't big on providing options. My ideal form factor is a Yoga or Surface Pro style - and iOS is not a real option.
    - bserge 6y ago
      That's unfortunate. Is it because of Plundervolt? Intel's answer to that was basically "lol, just disable manual voltage control", even older machines got BIOS updates disabling it, which made a tiny number of people angry.
  - fomine3 6y ago
    VAIO argue that they design great cooling system (with TruePerformance brand) so their laptop with lower TDP CPU performs better than other manufacturer's higher TDP laptop with crap cooling. It looks interesting because it's also portable, but it's expensive due to VAIO's selling scale.
- jeffbee 6y ago
  The 45W rating is hardly relevant. Excluding Apple, Intel 11th Gen Core CPUs are the fastest on most single-threaded benchmarks, whether they are configured at 15W, 27W, or 45W. In fact there is barely any difference in peak performance between the 15W and the 27W config.
  - reader_mode 6y ago
    Compared to last gen AMD cores ?
    - jeffbee 6y ago
      11th Gen Core are 10-20% faster than Zen 2 “Renoir” on the workloads I’ve tried.
      - reader_mode 6y ago
        That's the single core improvement you have between Zen 2 and Zen 3, but Zen 3 also has sightly better power usage than Zen 2 and Zen 2 was already superior (allegedly - I haven't actually used a Zen 2 mobile device since nothing really premium was built with it) - Zen 3 mobile isn't out yet but I suspect it's going to be a much better experience for mobile just like Zen 2 was compared to 10th gen intel.
- kllrnohj 6y ago
  The 15 MBP has worst-in-class thermal design. MacBooks have always had worst-in-class thermal designs, as Apple prioritized looks over function on that. Hence why there's no visible vents. What a 15 MBP can and can't cool is therefore not very interesting. Thermally it's terrible. 45W is still going to be a challenge for other designs, but it can (and has) been done without being deafeningly loud like a MBP or monstrously thick. Simply having intake & exhaust vents beyond just the keyboard & hinge alone massively improves things here. But yes you will hear it under sustained load. But you're also doing whatever you're doing a lot faster as a result - it's a tradeoff here. You could always disable half the cores if you don't like the 45W, Intel will sell you quad cores all day long quite happily.
eloff 6y ago
Yeah, they did hamstring AMD last I heard. That won't be a fair benchmark. Intel up to their old games again.
ksec 6y ago
1. Hamstring might not be fair to "Intel Compiler", as it was designed for Intel in the first place. Unfair Comparison would be a more valid description. i.e The problem is not with "Intel's compiler", the problem is with the benchmarks and marketing. 2. Keep in mind fastest Single Threaded benchmark is still somewhat true considering Intel can turbo up to 5Ghz. 3. TDP comparison between AMD and Intel are meaningless these days. They have different definition of TDP. 4. So if you put 2 and 3 together, that is like saying they have the fastest Laptop processor by overlocking it to 5Ghz with a higher energy usage within limited time frame. Final Note: Generally speaking one should avoid wccftech unless you know precisely how to read into their rumours and findings. ( Unless their source are official like in this case ) Anandtech is a much better source.
- daniel-cussen 6y ago
  By "hamstring" they mean things that the Intel Compiler did get caught doing, like run the code badly if the CPUID did not say "genuine intel."
- giantrobot 6y ago
  The Intel compiler has been doing function multi-versioning for many years now. If you annotate a function for multi-versioning the compiler will build several optimized versions of it, including a generic version, running the most optimal one for the host CPU based on a CPUID check. Intel purposefully misidentifies AMD chips with the CPUID check so always runs the generic version of a function on AMD. It does this even if the host AMD chip supports 100% of the features as some Intel CPU that runs the optimized version. So for any benchmark compiled with ICC any results for AMD chips are meaningless. You can only compare Intel to Intel or AMD to AMD. Compiling that same code without FMV so each CPU runs the exact same instructions will see the difference between them narrow or reverse. Intel is free to optimize the shit out of their compiler but it's bullshit that they also actively sabotage AMD chips at the same time. It's very much a vendor lock-in tactic.
  - ksec 6y ago
    >Intel is free to optimize the shit out of their compiler but it's bullshit that they also actively sabotage AMD chips at the same time. It's very much a vendor lock-in tactic. I would argue they actively make sure their compiler optimisation works only on Intel CPU. Which is actually a fair deal. Those optimisation work isn't free. It is BS tactics? Yes. Do I Like it ? No. It is wrong to do it? That is up to debate. The point is using an Intel Compiler on Non Intel CPU comparison makes the benchmarks pretty much useless. Something like GCC / LLVM should have been used to be vendor neutral.
agloeregrets 6y ago
huh, looking at that chart.. I just can't help but think.. Is there a bar missing there? Just one bar, maybe just two letters? Perhaps even running in x86 emulation? Hardcore borders on false advertising.

by Jonnax 6y ago

reader_mode 6y ago

Does it really matter in a 45W power range ? I have a 2018 15 MBP with an i9-8950HK that has the same power range - this thing is unusable at full load for anything but burst - and the way it taxes the thermals makes it a terrible laptop (people in the office start turning their heads). I'm not convinced any portable laptop can handle that TDP efficiently - so maybe this is interesting for gaming laptops - but to me this looks like Intel is still making shitty power-hungry chips for laptops They are boasting performance comparisons to AMD Navi 2, Navi 3 also brings a ~20% IPC improvement which they claim to have over it. I'd speculate that Ryzen 7 5800U should outperform this chip in most relevant use cases (eg. sustained load like running a simulator while developing/running build watchers or running VM/docker and IDE/compiler) while being in 25W range which can be handled reasonably at this device size. Unfortunately laptop manufacturers are so slow at adopting AMD in premium line laptops (I preorder a surface pro with an AMD 5xxx CPU). I don't see myself buying an Intel based laptop until they improve the thermals.

AmVess 6y ago
Apple's M1 has shown us how antiquated and far behind Intel has become. My Macbook is not only much faster than my XPS unit, but it doesn't get hot, nor does it use up the battery in 4 hours.
bserge 6y ago
It's the heatsink. And Apple isn't the only offender. All major manufacturers use terrible cooling systems, even on thick workstation class laptops. Their testing is seriously flawed, they design the heatsink for "normal workload" which means jack shit. Everyone's got different uses for their machines, so just test the heatsink at maximum load/TDP for several hours at least, make sure it doesn't overheat. Is that too much to ask? And if you design your laptops to be thin and light, don't shove a 6 core 45W processor in it... just don't. It will perform slower than a 4 core 35W chip, so what's the point? Having it on the spec sheet is not worth the negative feedback from buyers. Apple went the other way with the M1, it doesn't overheat because it just uses much less power, while having the same performance. Really impressive. You can fix the overheating problem by buying some good thermal paste and applying it yourself and/or by using software like Throttlestop to limit the maximum TDP (and even undervolt it to claw back some of that performance) and NotebookFanControl for the fans. Not sure if there are alternatives for the MacBook, though.
- reader_mode 6y ago
  You can't under-volt on latest machines (mine included). I need to limit TDP to get the machine usable. I've thought about opening it up and reapplying paste but I don't want to risk bricking my work device - I don't have the time to deal with it. I will have a few weeks off soon so might consider it if the expected gains are significant but from what I saw back when I looked into it (this[1] mostly) it's not worth it. https://www.reddit.com/r/macbookpro/comments/96wwyf/so_i_replaced_the_thermal_compound_on_my_mbp_2018/ https://www.reddit.com/r/macbookpro/comments/96wwyf/so_i_rep...
  - AmVess 6y ago
    It may drop the temp a few C or delay throttling, but small heatsinks and tiny fans can only dump so much heat out of the chassis. I did all kinds of things to my XPS, but the largest gain came from undervolting and it still sounded like a leafblower when it was being worked moderately hard. My M1 Macbook on the other hand, is dead quiet on top of absolutely destroying the XPS in CPU performance while having vastly superior battery life. I don't have to run anything out of spec to keep it from cooking itself.
    - reader_mode 6y ago
      Yeah no doubt M1s are amazing, I'm kind of sad Apple is the only one with it because Apple isn't big on providing options. My ideal form factor is a Yoga or Surface Pro style - and iOS is not a real option.
  - bserge 6y ago
    That's unfortunate. Is it because of Plundervolt? Intel's answer to that was basically "lol, just disable manual voltage control", even older machines got BIOS updates disabling it, which made a tiny number of people angry.
- fomine3 6y ago
  VAIO argue that they design great cooling system (with TruePerformance brand) so their laptop with lower TDP CPU performs better than other manufacturer's higher TDP laptop with crap cooling. It looks interesting because it's also portable, but it's expensive due to VAIO's selling scale.
jeffbee 6y ago
The 45W rating is hardly relevant. Excluding Apple, Intel 11th Gen Core CPUs are the fastest on most single-threaded benchmarks, whether they are configured at 15W, 27W, or 45W. In fact there is barely any difference in peak performance between the 15W and the 27W config.
- reader_mode 6y ago
  Compared to last gen AMD cores ?
  - jeffbee 6y ago
    11th Gen Core are 10-20% faster than Zen 2 “Renoir” on the workloads I’ve tried.
    - reader_mode 6y ago
      That's the single core improvement you have between Zen 2 and Zen 3, but Zen 3 also has sightly better power usage than Zen 2 and Zen 2 was already superior (allegedly - I haven't actually used a Zen 2 mobile device since nothing really premium was built with it) - Zen 3 mobile isn't out yet but I suspect it's going to be a much better experience for mobile just like Zen 2 was compared to 10th gen intel.
kllrnohj 6y ago
The 15 MBP has worst-in-class thermal design. MacBooks have always had worst-in-class thermal designs, as Apple prioritized looks over function on that. Hence why there's no visible vents. What a 15 MBP can and can't cool is therefore not very interesting. Thermally it's terrible. 45W is still going to be a challenge for other designs, but it can (and has) been done without being deafeningly loud like a MBP or monstrously thick. Simply having intake & exhaust vents beyond just the keyboard & hinge alone massively improves things here. But yes you will hear it under sustained load. But you're also doing whatever you're doing a lot faster as a result - it's a tradeoff here. You could always disable half the cores if you don't like the 45W, Intel will sell you quad cores all day long quite happily.

eloff 6y ago

Yeah, they did hamstring AMD last I heard. That won't be a fair benchmark. Intel up to their old games again.

ksec 6y ago

1. Hamstring might not be fair to "Intel Compiler", as it was designed for Intel in the first place. Unfair Comparison would be a more valid description. i.e The problem is not with "Intel's compiler", the problem is with the benchmarks and marketing. 2. Keep in mind fastest Single Threaded benchmark is still somewhat true considering Intel can turbo up to 5Ghz. 3. TDP comparison between AMD and Intel are meaningless these days. They have different definition of TDP. 4. So if you put 2 and 3 together, that is like saying they have the fastest Laptop processor by overlocking it to 5Ghz with a higher energy usage within limited time frame. Final Note: Generally speaking one should avoid wccftech unless you know precisely how to read into their rumours and findings. ( Unless their source are official like in this case ) Anandtech is a much better source.

daniel-cussen 6y ago
By "hamstring" they mean things that the Intel Compiler did get caught doing, like run the code badly if the CPUID did not say "genuine intel."
giantrobot 6y ago
The Intel compiler has been doing function multi-versioning for many years now. If you annotate a function for multi-versioning the compiler will build several optimized versions of it, including a generic version, running the most optimal one for the host CPU based on a CPUID check. Intel purposefully misidentifies AMD chips with the CPUID check so always runs the generic version of a function on AMD. It does this even if the host AMD chip supports 100% of the features as some Intel CPU that runs the optimized version. So for any benchmark compiled with ICC any results for AMD chips are meaningless. You can only compare Intel to Intel or AMD to AMD. Compiling that same code without FMV so each CPU runs the exact same instructions will see the difference between them narrow or reverse. Intel is free to optimize the shit out of their compiler but it's bullshit that they also actively sabotage AMD chips at the same time. It's very much a vendor lock-in tactic.
- ksec 6y ago
  >Intel is free to optimize the shit out of their compiler but it's bullshit that they also actively sabotage AMD chips at the same time. It's very much a vendor lock-in tactic. I would argue they actively make sure their compiler optimisation works only on Intel CPU. Which is actually a fair deal. Those optimisation work isn't free. It is BS tactics? Yes. Do I Like it ? No. It is wrong to do it? That is up to debate. The point is using an Intel Compiler on Non Intel CPU comparison makes the benchmarks pretty much useless. Something like GCC / LLVM should have been used to be vendor neutral.

agloeregrets 6y ago

huh, looking at that chart.. I just can't help but think.. Is there a bar missing there? Just one bar, maybe just two letters? Perhaps even running in x86 emulation? Hardcore borders on false advertising.