11 ms·

Once again crushing the competition in single core performance, which is what matters.

by bzb3 6y ago

ido 6y ago
Matters for whom? For a lot of use cases (web servers, compiling, databases...) having lots of cores matters (and total = performance of core * number of cores), for others it doesn't.
- globular-toast 6y ago
  No, it does not matter. That's a common misconception. If you double your core count you can run double the number of threads in the same amount of time. That's it. If you double your single core speed you can still run double the number of threads in the same amount of time or the same amount of threads in half the time. Single core performance is always preferable for CPUs. The CPU manufacturers basically conned everyone into thinking multi-core was "good enough". Software people who knew what they were talking about back in the day called them out on it but nobody listened. Now here we are on Hacker News in 2020 and people still not understanding this.
  - gpm 6y ago
    We all know that a single core with twice the performance and twice the cache space is generally better than two cores (though a sibling comment does raise a good point that poorly written software can lock up an entire core regardless of it's Hz, also some VM scenarios would prefer extra cores over faster cores). We also all know that a single core with twice the performance and twice the cache space costs a lot more than two cores. It's a lot more expensive than even 3 cores. I'd rather have 3 4Ghz cores than 1 8Ghz core (except for the novelty value).
    - yjftsjthsd-h 6y ago
      We all know that a single core with twice the performance and twice the cache space is generally better than two cores I'm not sure that's actually true, even if you want performance (although it's dependent on whether your workload can use more than 1 thread, of course): A single core at a higher clock speed might be better if it can hold that clock speed. But the faster you run, the more power you use and the more heat you produce, and worse, the relationship isn't linear. So ex. 1 core at 4ghz won't stay at 4 ghz, where 2 cores at 2 ghz can hold that speed forever.
      - gpm 6y ago
        Right, when I say an xGhz core I'm pretending to live in a simpler world where they can all hold that speed forever. Consider an 8Ghz core to be a core + cooling that can consistently hold a speed twice as fast as the 4Ghz core + cooling can hold.
    - globular-toast 6y ago
      > We also all know that a single core with twice the performance and twice the cache space costs a lot more than two cores. It's a lot more expensive than even 3 cores. I'm a programmer, I don't care how much it costs the hardware manufacturer. I want a single core. Multi-threaded programming sucks. I don't care if that's implemented via multiple cores acting together and only presents itself as a single core. I just want a single core. > I'd rather have 3 4Ghz cores than 1 8Ghz core (except for the novelty value). I'd much rather avoid multi-threaded programming and get the 8GHz core. But your comment is missing the point, really, which is that single-core performance matters. If the 8GHz core existed then you'd still be able to get 3 8GHz cores. Still want the 3x4GHz?
      - gpm 6y ago
        8Ghz cores do exist if you have enough money to hire someone to deal with the overclocking for you (or enough liquid nitrogen and time to do it yourself). I, on the other hand, care about the cost of my computers a bit more than that, not having infinite money. Which is why even if 8Ghz cores were commercially available I suspect I would still be buying 3 times as many 4Ghz ones.
        blattimwind 6y ago
        A Bulldozer FX barreling down the highway at over 8 GHz is actually still slower than any modern CPU at less than half the clock speed. The latter also doesn't require LN2 cooling and doesn't draw more than 100 W for a single core.
      - zrm 6y ago
        > If the 8GHz core existed then you'd still be able to get 3 8GHz cores. Well no. Right now you can get ~5GHz cores and they draw close to 300 watts per core, which is why they can't run all cores at 5GHz on threaded programs. For the same amount of power as one core at 5GHz you can get 64 cores at almost 3GHz, e.g. ThreadRipper 3990X. An 8GHz core would presumably draw 500+ watts per core and maybe even a lot more than that. To get the same performance as the 3990X, you would either need a single 185GHz core (basically impossible) or >23 8GHz cores (expected power draw >11kW). That would cost thousands a year in electricity, but a normal 20 amp household circuit breaker would trip before you hit 2.5kW anyway. More than that, by the time you write code that can scale to 16+ threads, 64 isn't that much harder, and 256 isn't much harder than that. So get used to it, because in a few years there are going to be 256-core mobile devices and 1024+ core desktops and servers.
  - sk0g 6y ago
    So say I'm close to maxing out my 8c/16t CPU, would a better alternative be a 16*4; 64Ghz single core processor? Why? You wouldn't have true parallelism then, would you? So if you have something, for whatever reason, blocking further processing, the entire computer would lock up. Why is this a better scenario again?
    - globular-toast 6y ago
      You don't care about parallelism, you care about concurrency. Operating systems have handled that just fine for many decades now.
    - paulmd 6y ago
      > You wouldn't have true parallelism then, would you? So if you have something, for whatever reason, blocking further processing, the entire computer would lock up That doesn't occur because the OS scheduler will eventually pre-empt your thread and swap to another task. You don't really have control over that. All modern OSs use pre-emptive multitasking and there's not a ton you can do to prevent it. (read: anything in the last 30 years, last cooperative multitasking I remember was MacOS Classic (probably like OS7 or so.) Single-core processors were the norm up until 15 years ago, and people routinely ran more than one process on them. Yes, we ran Discord (we called it Roger Wilco or Teamspeak back then) and played music in the background too, without 4 dedicated cores for the background tasks. Yes, in the abstract it would be preferable to have a 64 GHz processor instead of 16 4 GHz processors, because then you can have all that speed devoted to a single task. Context switching is not that expensive (particularly if you have hyperthreads/SMT to soak up the wasted cycles). The physics of transistors make that impossible, but we absolutely would not do any of this multithreading crap if we didn't have to. Single-threaded code is easier to write and usually faster (in terms of less total cycles - multithreaded programs waste extra cycles on synchronization, overhead, spinning for locks, etc that single-threaded programs don't have to).
      - theandrewbailey 6y ago
        > (read: anything in the last 30 years, last cooperative multitasking I remember was MacOS Classic (probably like OS7 or so.) I remember reading about that a while ago. Wasn't everything up to and including Mac OS 9 cooperatively multitasked?
  - ben-schaaf 6y ago
    Turns out this has been tested, and the results are clear: Running a single core 9900k is pretty bad in comparison to 8 heavily downclocked cores on the same CPU. While single threaded tasks are obviously slower, frame rates are much more stable, windows is more usable with 8 cores. Several games simply didn't work with a single core, or took ages to load. Reference: https://youtu.be/feO54CBUJCk https://youtu.be/feO54CBUJCk
    - tylerhou 6y ago
      The video you linked is pretty weak; there was (understandably) no performance tuning done on both configurations (Windows scheduler, game scheduling). https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/36448.pdf https://static.googleusercontent.com/media/research.google.c...
    - SomeoneFromCA 6y ago
      This is because for a fair comparison you need to scale the L1/L2 cache size as well. More cores - more cache. It is also important to not scale below 2 cores, because servicing interrupts causes performance hiccups. It is enough to have the second core heavily downclocked, and you'll still have better performance then with just a single core.
  - madushan1000 6y ago
    > same amount of threads in half the time Sure, if you forget all the additional cache evictions, memory reads and writes and TLB flushes caused by all the competing threads on a single core. Not to mention that one shitty bit of code in one thread can poison the cache enough to re-read the data from memory every context switch.
SketchySeaBeast 6y ago
Depends on workload, and we've seen that with the latest round of AMD offerings. As a gamer, yes, but that's really it.
- gpm 6y ago
  Are there actually games that take advantage of a CPU this fast? I would imagine that all the game developers would be targeting the more modest hardware that most of their players actually have.
  - zamadatix 6y ago
    Generally those with the top end CPU have the top end GPU and the top end monitor. In that regard all the game has to do to "take advantage" of the CPU is support high or unlocked framerates and high resolutions.
    - leetcrew 6y ago
      > Generally those with the top end CPU have the top end GPU and the top end monitor I sort of doubt this is true, if only for the reason that most people don't buy a new CPU, GPU, and monitor at the same time. monitor and pc are probably going to be on different upgrade cycles, unless the monitor came with the pc (in which case, probably not "top end"). for DIYers, the GPU and CPU+RAM+mobo are likely also on different upgrade cycles. I bought my current 4K monitor about four years ago, but I ran most games at 1080p until I had enough spare cash to buy a 1080 ti. I finally upgraded my CPU, RAM, and motherboard this year, mainly because I was tired of having 8GB of DDR3. in any case, it's usually not a rational use of funds to buy the fastest CPU and GPU for a 4K monitor unless your budget is unlimited. with a constrained budget, it's usually better to drop a tier or two on the CPU if it gets you the next higher tier GPU.
  - InvaderFizz 6y ago
    Anything high refresh rate(120+Hz) benefits immensely from these high clock rates. The high clocks result in less pipeline stalls meaning your high frame rates are more achievable as well as significantly reducing the dips (0.1% lows is what every reviewer hates). This isnt about "average", this is about having the best experience possible, which Intel provides in this very specific scenario. Intel has three things going for it(in the consumer space), which matter a lot or not at all to different folks: - better idle/low usage power consumption than AMD (low load laptop battery life) - Intel's QuickSync is fantastic for media encode/decode (NVENC is typically better, AMD is basically non-existent) - Intel has the highest clock rates. Outside of these niches, AMD is dominating at every level of the product stack.
dralley 6y ago
The gap is pretty small, even now, and Zen 3 comes out in a few months. The value proposition of these chips is pretty bad, as well. Enormous power and cooling requirements, short-lived motherboard sockets, and sheer price/performance ratio are all not great in comparison.
mschuetz 6y ago
/s ? For my computer graphics and processing work, multithreaded tasks scale amazingly well with the number of cores so single core performance doesn't matter all that much. A single thread isn't even be able to process all the data that's streamed from an SSD. I need three threads just to process/prepare the data in the background, in addition to the main thread that keeps rendering. I suspect it's only a matter of time until multithreading is a must for games. The additional programming effort may not have paid off if players only have 2 or 4 core CPUs, but now that 8 cores with 16 threads are increasingly common, it's going to make a difference.
- ido 6y ago
  AAA games have all but required quad cores (or at least 2c/4t) for years now. I think a lot of games already scale well to 4c/8t today.
  - Uehreka 6y ago
    Do they use those cores? Sure. But when I’m playing a game (even a game that feels like it should be ripping threads like Factorio) I usually see that one core is pegged at 100% while the others are chilling around 40-60% (I have an overclocked 8700K). In my experience, single core performance is still the bottleneck (as far as CPUs are concerned) for gaming.
    - driverdan 6y ago
      Yes, recent games do. For example, Doom Eternal can max out all CPU cores.
    - skocznymroczny 6y ago
      Factorio is an indie game, I guess it might not be easy for indie developers to make use of multithreading, whereas AAA game developers have entire teams dedicated to engine development and multithreading. Anyway, traditional graphics APIs like OpenGL are practically single-threaded. Modern APIs like DX12 and Vulkan have been designed with multithreading in mind and support scaling with number of cores much better (with added overhead of having to do manual synchronization).
kevin_thibedeau 6y ago
... for gamers playing single-threaded games.
- ajross 6y ago
  To be fair, that's exactly the target market for high end desktop CPUs like this. This isn't a datacenter part, and other parallel workloads tend not to make up much share of the personal workstation market. If you want to max out gaming benchmarks right now, this is your CPU. And there are a decent number of people who do, so heavily tuning existing dies (I mean, let's be honest, this a rebinned and overclocked 10900X) makes sense.
- paulmd 6y ago
  When people say "single-threaded performance" they don't mean literally one single thread will ever be used. The single-core performance is a proxy for performance-per-core and that's a certainly a very important number. Intel's per-core performance is decently higher in gaming, between their lower latency giving them slightly higher IPC in gaming tasks[1] (as these are latency sensitive) and the much higher all-core clocks they are pushing (at least 10% higher than AMD at present). That still gives them an edge in gaming, although certainly much less of an edge against Zen2 than it was against Zen1 (which was much worse than people were willing to admit at the time, people at the time were benching it with a GTX 1080 (the fastest card available at the time) at 1440p and 4K and claiming there was "no noticeable difference", but faster GPUs showed the weakness). I think the 10700F looks good, 4.6 GHz all-core 8C16T part for under $300 looks compelling for gaming customers if you can keep it cool. The 10900K fills the "I want the top gaming performance first and foremost, with as many cores as I can get, and I'm not super concerned about budget" but is definitely on the expensive side. Other than that, nothing terribly impressive here IMO, the other chips fall into a range between "not fast enough" and "not cheap enough". [1] https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-core-i9/ https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-...
  - zozbot234 6y ago
    > The single-core performance is a proxy for performance-per-core A very bad proxy then, since performance-per-core with multiple cores in use will almost certainly be limited by constraints on (1) memory bandwidth and (2) power/thermals. "Single-core performance" only applies when you're literally pegging one core at 100% cpu.
specto 6y ago
Not really... If you look at the FPS numbers, sure AMD is slightly lower. However AMD can also perform other functions like encoding/decoding. If you want pure FPS, sure Intel will win. For general computing AND gaming, AMD wins every time per dollar.
tpmx 6y ago
It is what matters to web browsing performance. Web browsing is no longer a "light" thing to do. You need solid integer single core computing performance to have a snappy browsing experience. AFAIK, there is not yet a web browser core that can parallelize tasks in a meaningful way.
- gpm 6y ago
  Mozilla has been making pretty good progress with parallelization of firefox with their oxidation project. All browsers are splitting different tabs into different processes which means they are taking advantage of multithreading if you have more than one compute intensive tab/window open.
  - tpmx 6y ago
    Splitting tabs into separate processes is mainstream these days, but it doesn't really help that much. You only need the focused tab to be snappy.
    - gpm 6y ago
      You only need the focused tab to be snappy, but to let it be snappy you need to keep all the other work away from that core. Gmail updating emails, youtube playing a video in the background, discord pinging chat messages at you, and so and and so forth.
      - tpmx 6y ago
        With a modern javascript engine you can throttle the activity of tabs that are not focused. (And browsers do indeed do that. Pretty aggressively, too.)
        Marsymars 6y ago
        I wish there was some extension that blocked any non-focused tabs from performing any javascript execution unless I whitelist the domain.
loufe 6y ago
While I agree this post should be downvoted I do think it's worth mentioning at least two different reasons why higher clock speed would push me to intel this generation: AutoCAD and Europa Universalis IV. Those are my two principal demanding softwares where they cannot see any real means of moving past single core for the vast majority of processing.
anthony_doan 6y ago
We should all wait for real world testing. IIRC AMD and Intel test their performance differently. Intel tend to give turbo performance which can only run for so long before heat becomes a problem.
- sq_ 6y ago
  Looks like GN already has a video out, so I imagine the rest of the big hardware channels will follow along quickly. Don't have time to watch GN's video now, but I bet they'll have taken a nice in-depth look at the performance.
lawrenceyan 6y ago
I think part of the reason why you may be being downvoted has to do with the tone you're using in making your statement as well as the newness of your account. Also, though you're statement is technically correct, it fails to account for the nuance that increasingly across all applications, not just gaming, multi-threaded processing performance is becoming the far more important benchmark to look at.
- ksec 6y ago
  > crushing... crushing adjective [usually before noun] used to emphasise how bad or severe something is I had to check the dictionary just in case. And in this case neither was it technically correct, or downvoted because of the tone. I downvoted it because it was factually wrong. Zen 2 actually has better IPC than Skylake in many cases, on a sustained single thread performance they are likely to be on a similar level despite having a slightly lower clock speed disadvantage on a much lower TDP. Hardly crushing.
redisman 6y ago
Just the fact that it's yet another socket for basically the same process is insane. On AM4 I've been able to use their low-end to very-high end CPUs for the last several years and most likely will be able to use the ones coming out this year too.

Once again crushing the competition in single core performance, which is what matters.

by bzb3 6y ago

Once again crushing the competition in single core performance, which is what matters.

ido 6y ago

Matters for whom? For a lot of use cases (web servers, compiling, databases...) having lots of cores matters (and total = performance of core * number of cores), for others it doesn't.

globular-toast 6y ago
No, it does not matter. That's a common misconception. If you double your core count you can run double the number of threads in the same amount of time. That's it. If you double your single core speed you can still run double the number of threads in the same amount of time or the same amount of threads in half the time. Single core performance is always preferable for CPUs. The CPU manufacturers basically conned everyone into thinking multi-core was "good enough". Software people who knew what they were talking about back in the day called them out on it but nobody listened. Now here we are on Hacker News in 2020 and people still not understanding this.
- gpm 6y ago
  We all know that a single core with twice the performance and twice the cache space is generally better than two cores (though a sibling comment does raise a good point that poorly written software can lock up an entire core regardless of it's Hz, also some VM scenarios would prefer extra cores over faster cores). We also all know that a single core with twice the performance and twice the cache space costs a lot more than two cores. It's a lot more expensive than even 3 cores. I'd rather have 3 4Ghz cores than 1 8Ghz core (except for the novelty value).
  - yjftsjthsd-h 6y ago
    We all know that a single core with twice the performance and twice the cache space is generally better than two cores I'm not sure that's actually true, even if you want performance (although it's dependent on whether your workload can use more than 1 thread, of course): A single core at a higher clock speed might be better if it can hold that clock speed. But the faster you run, the more power you use and the more heat you produce, and worse, the relationship isn't linear. So ex. 1 core at 4ghz won't stay at 4 ghz, where 2 cores at 2 ghz can hold that speed forever.
    - gpm 6y ago
      Right, when I say an xGhz core I'm pretending to live in a simpler world where they can all hold that speed forever. Consider an 8Ghz core to be a core + cooling that can consistently hold a speed twice as fast as the 4Ghz core + cooling can hold.
  - globular-toast 6y ago
    > We also all know that a single core with twice the performance and twice the cache space costs a lot more than two cores. It's a lot more expensive than even 3 cores. I'm a programmer, I don't care how much it costs the hardware manufacturer. I want a single core. Multi-threaded programming sucks. I don't care if that's implemented via multiple cores acting together and only presents itself as a single core. I just want a single core. > I'd rather have 3 4Ghz cores than 1 8Ghz core (except for the novelty value). I'd much rather avoid multi-threaded programming and get the 8GHz core. But your comment is missing the point, really, which is that single-core performance matters. If the 8GHz core existed then you'd still be able to get 3 8GHz cores. Still want the 3x4GHz?
    - gpm 6y ago
      8Ghz cores do exist if you have enough money to hire someone to deal with the overclocking for you (or enough liquid nitrogen and time to do it yourself). I, on the other hand, care about the cost of my computers a bit more than that, not having infinite money. Which is why even if 8Ghz cores were commercially available I suspect I would still be buying 3 times as many 4Ghz ones.
      - blattimwind 6y ago
        A Bulldozer FX barreling down the highway at over 8 GHz is actually still slower than any modern CPU at less than half the clock speed. The latter also doesn't require LN2 cooling and doesn't draw more than 100 W for a single core.
    - zrm 6y ago
      > If the 8GHz core existed then you'd still be able to get 3 8GHz cores. Well no. Right now you can get ~5GHz cores and they draw close to 300 watts per core, which is why they can't run all cores at 5GHz on threaded programs. For the same amount of power as one core at 5GHz you can get 64 cores at almost 3GHz, e.g. ThreadRipper 3990X. An 8GHz core would presumably draw 500+ watts per core and maybe even a lot more than that. To get the same performance as the 3990X, you would either need a single 185GHz core (basically impossible) or >23 8GHz cores (expected power draw >11kW). That would cost thousands a year in electricity, but a normal 20 amp household circuit breaker would trip before you hit 2.5kW anyway. More than that, by the time you write code that can scale to 16+ threads, 64 isn't that much harder, and 256 isn't much harder than that. So get used to it, because in a few years there are going to be 256-core mobile devices and 1024+ core desktops and servers.
- sk0g 6y ago
  So say I'm close to maxing out my 8c/16t CPU, would a better alternative be a 16*4; 64Ghz single core processor? Why? You wouldn't have true parallelism then, would you? So if you have something, for whatever reason, blocking further processing, the entire computer would lock up. Why is this a better scenario again?
  - globular-toast 6y ago
    You don't care about parallelism, you care about concurrency. Operating systems have handled that just fine for many decades now.
  - paulmd 6y ago
    > You wouldn't have true parallelism then, would you? So if you have something, for whatever reason, blocking further processing, the entire computer would lock up That doesn't occur because the OS scheduler will eventually pre-empt your thread and swap to another task. You don't really have control over that. All modern OSs use pre-emptive multitasking and there's not a ton you can do to prevent it. (read: anything in the last 30 years, last cooperative multitasking I remember was MacOS Classic (probably like OS7 or so.) Single-core processors were the norm up until 15 years ago, and people routinely ran more than one process on them. Yes, we ran Discord (we called it Roger Wilco or Teamspeak back then) and played music in the background too, without 4 dedicated cores for the background tasks. Yes, in the abstract it would be preferable to have a 64 GHz processor instead of 16 4 GHz processors, because then you can have all that speed devoted to a single task. Context switching is not that expensive (particularly if you have hyperthreads/SMT to soak up the wasted cycles). The physics of transistors make that impossible, but we absolutely would not do any of this multithreading crap if we didn't have to. Single-threaded code is easier to write and usually faster (in terms of less total cycles - multithreaded programs waste extra cycles on synchronization, overhead, spinning for locks, etc that single-threaded programs don't have to).
    - theandrewbailey 6y ago
      > (read: anything in the last 30 years, last cooperative multitasking I remember was MacOS Classic (probably like OS7 or so.) I remember reading about that a while ago. Wasn't everything up to and including Mac OS 9 cooperatively multitasked?
- ben-schaaf 6y ago
  Turns out this has been tested, and the results are clear: Running a single core 9900k is pretty bad in comparison to 8 heavily downclocked cores on the same CPU. While single threaded tasks are obviously slower, frame rates are much more stable, windows is more usable with 8 cores. Several games simply didn't work with a single core, or took ages to load. Reference: https://youtu.be/feO54CBUJCk https://youtu.be/feO54CBUJCk
  - tylerhou 6y ago
    The video you linked is pretty weak; there was (understandably) no performance tuning done on both configurations (Windows scheduler, game scheduling). https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/36448.pdf https://static.googleusercontent.com/media/research.google.c...
  - SomeoneFromCA 6y ago
    This is because for a fair comparison you need to scale the L1/L2 cache size as well. More cores - more cache. It is also important to not scale below 2 cores, because servicing interrupts causes performance hiccups. It is enough to have the second core heavily downclocked, and you'll still have better performance then with just a single core.
- madushan1000 6y ago
  > same amount of threads in half the time Sure, if you forget all the additional cache evictions, memory reads and writes and TLB flushes caused by all the competing threads on a single core. Not to mention that one shitty bit of code in one thread can poison the cache enough to re-read the data from memory every context switch.

SketchySeaBeast 6y ago

Depends on workload, and we've seen that with the latest round of AMD offerings. As a gamer, yes, but that's really it.

gpm 6y ago
Are there actually games that take advantage of a CPU this fast? I would imagine that all the game developers would be targeting the more modest hardware that most of their players actually have.
- zamadatix 6y ago
  Generally those with the top end CPU have the top end GPU and the top end monitor. In that regard all the game has to do to "take advantage" of the CPU is support high or unlocked framerates and high resolutions.
  - leetcrew 6y ago
    > Generally those with the top end CPU have the top end GPU and the top end monitor I sort of doubt this is true, if only for the reason that most people don't buy a new CPU, GPU, and monitor at the same time. monitor and pc are probably going to be on different upgrade cycles, unless the monitor came with the pc (in which case, probably not "top end"). for DIYers, the GPU and CPU+RAM+mobo are likely also on different upgrade cycles. I bought my current 4K monitor about four years ago, but I ran most games at 1080p until I had enough spare cash to buy a 1080 ti. I finally upgraded my CPU, RAM, and motherboard this year, mainly because I was tired of having 8GB of DDR3. in any case, it's usually not a rational use of funds to buy the fastest CPU and GPU for a 4K monitor unless your budget is unlimited. with a constrained budget, it's usually better to drop a tier or two on the CPU if it gets you the next higher tier GPU.
- InvaderFizz 6y ago
  Anything high refresh rate(120+Hz) benefits immensely from these high clock rates. The high clocks result in less pipeline stalls meaning your high frame rates are more achievable as well as significantly reducing the dips (0.1% lows is what every reviewer hates). This isnt about "average", this is about having the best experience possible, which Intel provides in this very specific scenario. Intel has three things going for it(in the consumer space), which matter a lot or not at all to different folks: - better idle/low usage power consumption than AMD (low load laptop battery life) - Intel's QuickSync is fantastic for media encode/decode (NVENC is typically better, AMD is basically non-existent) - Intel has the highest clock rates. Outside of these niches, AMD is dominating at every level of the product stack.

dralley 6y ago

The gap is pretty small, even now, and Zen 3 comes out in a few months. The value proposition of these chips is pretty bad, as well. Enormous power and cooling requirements, short-lived motherboard sockets, and sheer price/performance ratio are all not great in comparison.

mschuetz 6y ago

/s ? For my computer graphics and processing work, multithreaded tasks scale amazingly well with the number of cores so single core performance doesn't matter all that much. A single thread isn't even be able to process all the data that's streamed from an SSD. I need three threads just to process/prepare the data in the background, in addition to the main thread that keeps rendering. I suspect it's only a matter of time until multithreading is a must for games. The additional programming effort may not have paid off if players only have 2 or 4 core CPUs, but now that 8 cores with 16 threads are increasingly common, it's going to make a difference.

ido 6y ago
AAA games have all but required quad cores (or at least 2c/4t) for years now. I think a lot of games already scale well to 4c/8t today.
- Uehreka 6y ago
  Do they use those cores? Sure. But when I’m playing a game (even a game that feels like it should be ripping threads like Factorio) I usually see that one core is pegged at 100% while the others are chilling around 40-60% (I have an overclocked 8700K). In my experience, single core performance is still the bottleneck (as far as CPUs are concerned) for gaming.
  - driverdan 6y ago
    Yes, recent games do. For example, Doom Eternal can max out all CPU cores.
  - skocznymroczny 6y ago
    Factorio is an indie game, I guess it might not be easy for indie developers to make use of multithreading, whereas AAA game developers have entire teams dedicated to engine development and multithreading. Anyway, traditional graphics APIs like OpenGL are practically single-threaded. Modern APIs like DX12 and Vulkan have been designed with multithreading in mind and support scaling with number of cores much better (with added overhead of having to do manual synchronization).

kevin_thibedeau 6y ago

... for gamers playing single-threaded games.

ajross 6y ago
To be fair, that's exactly the target market for high end desktop CPUs like this. This isn't a datacenter part, and other parallel workloads tend not to make up much share of the personal workstation market. If you want to max out gaming benchmarks right now, this is your CPU. And there are a decent number of people who do, so heavily tuning existing dies (I mean, let's be honest, this a rebinned and overclocked 10900X) makes sense.
paulmd 6y ago
When people say "single-threaded performance" they don't mean literally one single thread will ever be used. The single-core performance is a proxy for performance-per-core and that's a certainly a very important number. Intel's per-core performance is decently higher in gaming, between their lower latency giving them slightly higher IPC in gaming tasks[1] (as these are latency sensitive) and the much higher all-core clocks they are pushing (at least 10% higher than AMD at present). That still gives them an edge in gaming, although certainly much less of an edge against Zen2 than it was against Zen1 (which was much worse than people were willing to admit at the time, people at the time were benching it with a GTX 1080 (the fastest card available at the time) at 1440p and 4K and claiming there was "no noticeable difference", but faster GPUs showed the weakness). I think the 10700F looks good, 4.6 GHz all-core 8C16T part for under $300 looks compelling for gaming customers if you can keep it cool. The 10900K fills the "I want the top gaming performance first and foremost, with as many cores as I can get, and I'm not super concerned about budget" but is definitely on the expensive side. Other than that, nothing terribly impressive here IMO, the other chips fall into a range between "not fast enough" and "not cheap enough". [1] https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-core-i9/ https://www.techspot.com/article/1876-4ghz-ryzen-3rd-gen-vs-...
- zozbot234 6y ago
  > The single-core performance is a proxy for performance-per-core A very bad proxy then, since performance-per-core with multiple cores in use will almost certainly be limited by constraints on (1) memory bandwidth and (2) power/thermals. "Single-core performance" only applies when you're literally pegging one core at 100% cpu.

specto 6y ago

Not really... If you look at the FPS numbers, sure AMD is slightly lower. However AMD can also perform other functions like encoding/decoding. If you want pure FPS, sure Intel will win. For general computing AND gaming, AMD wins every time per dollar.

tpmx 6y ago

It is what matters to web browsing performance. Web browsing is no longer a "light" thing to do. You need solid integer single core computing performance to have a snappy browsing experience. AFAIK, there is not yet a web browser core that can parallelize tasks in a meaningful way.

gpm 6y ago
Mozilla has been making pretty good progress with parallelization of firefox with their oxidation project. All browsers are splitting different tabs into different processes which means they are taking advantage of multithreading if you have more than one compute intensive tab/window open.
- tpmx 6y ago
  Splitting tabs into separate processes is mainstream these days, but it doesn't really help that much. You only need the focused tab to be snappy.
  - gpm 6y ago
    You only need the focused tab to be snappy, but to let it be snappy you need to keep all the other work away from that core. Gmail updating emails, youtube playing a video in the background, discord pinging chat messages at you, and so and and so forth.
    - tpmx 6y ago
      With a modern javascript engine you can throttle the activity of tabs that are not focused. (And browsers do indeed do that. Pretty aggressively, too.)
      - Marsymars 6y ago
        I wish there was some extension that blocked any non-focused tabs from performing any javascript execution unless I whitelist the domain.

loufe 6y ago

While I agree this post should be downvoted I do think it's worth mentioning at least two different reasons why higher clock speed would push me to intel this generation: AutoCAD and Europa Universalis IV. Those are my two principal demanding softwares where they cannot see any real means of moving past single core for the vast majority of processing.

anthony_doan 6y ago

We should all wait for real world testing. IIRC AMD and Intel test their performance differently. Intel tend to give turbo performance which can only run for so long before heat becomes a problem.

sq_ 6y ago
Looks like GN already has a video out, so I imagine the rest of the big hardware channels will follow along quickly. Don't have time to watch GN's video now, but I bet they'll have taken a nice in-depth look at the performance.

lawrenceyan 6y ago

I think part of the reason why you may be being downvoted has to do with the tone you're using in making your statement as well as the newness of your account. Also, though you're statement is technically correct, it fails to account for the nuance that increasingly across all applications, not just gaming, multi-threaded processing performance is becoming the far more important benchmark to look at.

ksec 6y ago
> crushing... crushing adjective [usually before noun] used to emphasise how bad or severe something is I had to check the dictionary just in case. And in this case neither was it technically correct, or downvoted because of the tone. I downvoted it because it was factually wrong. Zen 2 actually has better IPC than Skylake in many cases, on a sustained single thread performance they are likely to be on a similar level despite having a slightly lower clock speed disadvantage on a much lower TDP. Hardly crushing.

redisman 6y ago

Just the fact that it's yet another socket for basically the same process is insane. On AM4 I've been able to use their low-end to very-high end CPUs for the last several years and most likely will be able to use the ones coming out this year too.