Comment by kup0 | Hacker News Reader

Comment by kup0 | Hacker News Reader

Parent Full thread

I'm curious if they're just catering to their most common readers. If they have a high percentage of people that use the site that use workloads where single-core performance or low-count-multi-core performance is most important (gaming, apps that take advantage of 4 cores but not 18, etc) then maybe they thought that their algorithm originally favored Ryzen too much, and this is a balancing correction, not a favoring of Intel?

Also, I don't understand the uproar over one small little overall percentage number- the rest of the numbers are there for anyone to easily weigh one against the other. Just a few lines down it's clear than the 18-core part demolishes the 4-core part in multi-core workloads (as anyone would expect without even visiting the website in the first place).

Being able to handle more programs at a time/etc really has no bearing on "effective speed"- if their workloads are not apps that take advantage of 18 cores (encoders, etc), it's possible a 4-core part could "feel" just as fast, or faster, than an 18-core part if it does their workload faster.

That said, the whole AMD/Intel war is less important to me because fanboyism in any amount is toxic and worth avoiding. People biased heavily enough towards one company will always see anything like this as some kind of conspiracy or attack, regardless of whether or not it's true (something I don't know in this case)

What do you most people use their computer for? Web browsing, word. Maybe some foto or video editing. With the exception of word all of these things MASSIVELY gain from having lots of cores. Not sure where the myth that most people have single threaded workloads comes from because it's not true at all at this time.

I don't think its true that most people do web browsing. Its just an easy generalization to make. The world has thousands of industries and each specialty domain has their set of software suites that they need for their business. In my industry (pharma), and I'll adda caveat here that I haven't done a rigorous scientific analysis of every single workload, but off hand, all of our software is heavily dependent on single-thread perf. To be sure, there is some stuff that benefits from multiple cores.. dna sequencing/alignment/processing comes to mind, primarily because we're processing hundreds of GBs of data. But by and large, I would pick a CPU that has better single-thread perf. For our accounting folks who use quickbooks a lot, are not going to benefit "massively" from a high core count. For our facilities team that uses autocad.. again same thing .. I don't find a massive benefit from 10-12+ core systems. Sure you can always on a case-by-case basis invent scenarios which could potentially benefit from more cores, but this massive gain across the board that you claim is a bit of an exaggeration. IMHO/IME/YMMV/$0.02 etc..

99% of single threaded industry software doesn't need more power. That's why all of these cheap office pc have i3 processors that are years old. Single threaded performance has only one real advantage at this time and that is gaming. I think multi threading performance is way more important in the real world.

>99% of single threaded industry software doesn't need more power.

Again, I don't know if that is true. I'd love to see stats on that. My gut tells me improved single-threaded perf will always be welcome in most software because most data transforms are not not easily parallelizable. Certainly for me, more single threaded perf is always welcome in all the software I use daily.. from office suites to compilers to IDEs to photoshop to soft-PLCs/SCADA/HMI or even when opening a super large excel file with crazy forumlas. I don't think its just gaming.

The specifications of budget/cheap computers are not really an indicator of what will or wont benefit from single-threaded perf. Whether the benefit is worth paying for will obviously depend on other factors. For me its a nobrainer.

In what world would going from 4 cores to 18 cores even remotely matter for web browsing? If you notice a difference in >4 cores for web browsing (and maybe it IS noticeable and I just don't realize it)... then the web's even in a sadder state than I thought. Anecdotally, I had about 20-30 tabs open yesterday on a four-core machine, a couple of those I was actively using (one playing a video), and CPU usage was maybe 25% or less.

Photo/video editing on the other hand will DEFINITELY see a benefit, something I agree with- but I doubt nearly as many people do this as one would think- and again, it's clear from userbenchmark's page of data that the 18-core part in these workloads would demolish the 4-core part...so I don't see a problem.

I guess everyone just can't let go of what the "effective overall speed" number/percentage means. Maybe it's better not to have an "overall number" in the first place, since it can never always be true and is too dependent on the user? I have a feeling the majority of people reading the site are going to look at more detail than a single number anyway. They're certainly more likely to be gamers and/or enthusiasts

Ever since users could open multiple tabs at once and web pages could be active in the background, along with everything else that's running on the machine in the first place.

If single-core is all that's important, why are AMD and Intel coming out with multi-core chips at all and why are people buying them if they don't see any improvement?

It's less about single-core vs. multi-core, and even then, it's really not about quad-core vs more-cores.

My thought pattern through all of this (and maybe I'm completely off base) is that, for instance, having a 4-core part with high single-core performance might be preferable to a gamer/enthusiast (depending on common workloads) than an 18-core part with less single-core performance, and that maybe this is all a reason to consider why the weighting of UserBenchmark's "overall" score changed instead of immediately attaching the change to some anti-AMD bias

Maybe my idea that the most common workloads of UserBenchmark users (I don't claim to know their demographics, but I figured one could make assumptions based on the types of individuals that would look up benchmarks in the first place- very common among gamers) would see diminishing returns after, say 4 cores, is wrong- but that was how my thought process was going.

And a related point I'll concede (I admit my thinking maybe is just too old-school, biased, too filtered through my own experience and needs updating) is maybe there are more games and programs than I thought that actually take advantage of more than four cores.

Reply on news.ycombinator.com