Slashdot Mirror
Ask Slashdot: Should CPU, GPU Name-Numbering Indicate Real World Performance?
dryriver writes: Anyone who has built a PC in recent years knows how confusing the letters and numbers that trail modern CPU and GPU names can be because they do not necessarily tell you how fast one electronic part is compared to another electronic part. A Zoomdaahl Core C-5 7780 is not necessarily faster than a Boomberg ElectronRipper V-6 6220 -- the number at the end, unlike a GFLOPS or TFLOPS number for example, tells you very little about the real-world performance of the part. It is not easy to create one unified, standardized performance benchmark that could change this. One part may be great for 3D gaming, a competing part may smoke the first part in a database server application, and a third part may compress 4K HEVC video 11% faster. So creating something like, say, a Standardized Real-World Application Performance Score (SRWAPS) and putting that score next to the part name, letters, or series number will probably never happen. A lot of competing companies would have to agree to a particular type of benchmark, make sure all benchmarking is done fairly and accurately, and so on and so forth.
But how are the average consumers just trying to buy the right home laptop or gaming PC for their kids supposed to cope with the "letters and numbers salad" that follows CPU, GPU and other computer part names? If you are computer literate, you can dive right into the different performance benchmarks for a certain part on a typical tech site that benchmarks parts. But what if you are "Computer Buyer Joe" or "Jane Average" and you just want to glean quickly which two products -- two budget priced laptops listed on Amazon.com for example -- have the better performance overall? Is there no way to create some kind of rough numeric indicator of real-world performance and put it into a product's specs for quick comparison?
But how are the average consumers just trying to buy the right home laptop or gaming PC for their kids supposed to cope with the "letters and numbers salad" that follows CPU, GPU and other computer part names? If you are computer literate, you can dive right into the different performance benchmarks for a certain part on a typical tech site that benchmarks parts. But what if you are "Computer Buyer Joe" or "Jane Average" and you just want to glean quickly which two products -- two budget priced laptops listed on Amazon.com for example -- have the better performance overall? Is there no way to create some kind of rough numeric indicator of real-world performance and put it into a product's specs for quick comparison?
As soon as someone gives me a definitive definition of what "real world performance" for a CPU/GPU is that doesn't change over time/software-version/user-care-ometer is, I might agree that it's feasible to use it to name models.
Passmark. You're welcome. https://www.passmark.com/
'Real world performance' according to who or what, precisely?
Clock speed doesn't tell you the whole story and to the vast majority of people (read as: non-technical people) it wouldn't mean anything to them anyway, other than maybe one number is bigger than another number.
Same goes for so-called 'benchmark' test suites, which I think can be argued as being biased in one way or another (or a processor gaming the system to make it appear it's faster on such-and-such benchmark test).
I think that for the people such information matters to, they're going to already know what's what without anyone spelling it out for them.
What exact 'performance' figure does dryriver suggest?
Raw GIPS/TFLOPS? pretty much meaningless and very easy to get an achievable peak number.
SPECINT/SPECFP? with what OS, compiler, flags, version, etc?
Anyone who knows much about cpu/gpu performance knows why this is a very very very silly 'suggestion'. It would be not more meaningful than the numbers they assign now.
The complain should be with the manufacturers - please come up with more sensible naming practices, but in the end, thats their decision.
Marketing is what is being questioned here - nothing technical - and good luck with that..
Market separation/obfuscation through complex naming schemes is seen as a feature by manufacturers, not a problem..
Cars are also complex, they don't have simple-to-understand names and variants and require you to document yourself and investigate for large amounts of time before committing to a purchase.
Don't try to dumb down complex machinery. It will never work.
...gis sdrawkcab (usually not responding to ACs; don't bother posting as AC)
1. Like EPA fuel mileage tests, manufacturers will find ways to rig their chips so they benchmark better.
2. Unlike the EPA/feds, there's nobody to punish corporations when they cheat.
It's really not hard to do a little research to see how CPUs compare. Yes, it's a PIA if you're buying spur-of-the-moment and comparing laptops at the Big Box Store. But you need to do research. Hyperthreading and multiprocs will speed up some apps and do very little for others, some standardized benchmark number printed in the specs won't really tell consumers anything very useful. Too many variables and dependencies.
Never let a lack of data get in the way of a good rant.
I'll try a novel analogy instead of the typical car thing. Imagine these chips (CPUs, GPUs, etc.) as shoes. Yes, shoes. Now there are obviously shoes of all kinds of sizes and types, and no one shoe of a certain size/type can be said to fit a particular person's requirements. Too big, too small. Great (9) for the red carpet runway, not so much (2) the tarmac kind. Perfect (10) for the alpine, chafing and sweaty (1) on the beach.
User A does spreadsheets all day, B does FPS games, C does CAD, D AI research, etc.. Some require multi-threaded performance; some, single-threaded, etc. etc.. What might seem like a good performance for one use is weak for another. It's just not possible to come up with a workable single axis performance metric when performance is determined by multiple variables, each having their own weight depending upon the user.
If you want to shop for kit that best fits your needs, you first need to come up with an understanding of the importance of each of the variables then go comparison shop the various benchmarks out there. As with most nearly everything it's best to just ignore the marketing speech and go do your own research.
Two of my imaginary friends reproduced once
I got bitten by Intel Obfuscation Syndrome when I bought a Core 2 Quad Q8200, not realizing that it was the only one of the Core 2 Quads to not have virtualization. Yeah, I should have looked before I leaped. In the end, it was a bad buy all around, as the DG43NB motherboard I bought to go with it also ended up crapping out in a surprisingly short time, but lasting long enough to be out of warranty. Needless to say, all of my later builds have been AMD (with various makes of motherboards).
Oh, no! You have walked into the slavering fangs of a lurking grue!
As "Computer Buyer Joe", I have found that the best approach is to get my computer nerd nephew to hook me up with the good shit. I tell him how much I can spend and which games I want to play and he does the rest. Then, I throw him $50, which he immediately spends on oxycontin or rap records or whatever it is that kids spend money on these days.
You are welcome on my lawn.
They already do this. Always look for the standardised number following the dollar sign.
I can dive right into the numbers because I've been immersed in tech for 25 years. Put in the time to learn it or lean on someone who already has and compensate them fairly. If it's a friend or family member, do them a favor; if not, then buy the parts from whoever you talk to, or compensate them monetarily. You can't trivialize this...you can't boil it down to some simple number to describe all types of components.
I don't think the name needs to denote real performance numbers. However, it ABSOLUTELY SHOULD denote different products which HAVE PERFORMANCE DIFFERENCES! Case and point, AMD first released a RX560 which was benchmarked and reviewed by all the media/press which used 1024 Stream Processors and 16 compute units. A few weeks/months later, AMD quietly released a new version, still calling it the RX560 (with no other indication of a change and no announcement of a change), and instead having 896 Stream Processors and 14 compute units. Nothing indicates the difference between the two versions, but one version is absolutely less powerful than the other. Nvidia is now essentially doing a similar thing with the MX150, releasing a new version that uses less power, and runs with a 36% slower base clock speed, 32% slower boost clock speed, and 17% slower memory clock speed. Benchmarks are showing it having about a 20-25% slower performance as a result. Yet, it is still called MX150, and will be advertised as such when people go and look at the part.
In both these cases, they should be clearly renamed a different product, denoting the actual product, which you can then get clear idea of the performance of the product based on reviews and benchmarks.
But how are the average consumers just trying to buy the right home laptop or gaming PC for their kids supposed to cope
They don't need to. The average user will have their needs met by any computer built in the past 10 years.
if you want high-end or specialised stuff, just let the price guide you. The more expensive (so long as you don't get suckered into paying a brand premium) a generic computer is, the better it will perform.
Most people buy to a budget, anyway - not to a specification. That is why the first question a sales-droid will ask you is "how much do you have to spend?".
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
The model names are mostly standardized now, even across different manufacturers.
Intel has, for example Core i7-4790K, and Core i7-8700K. AMD would have Ryzen 2950x, and nVidia would have GTX 1070 Ti. There is a similar pattern in all of them.
Intel (desktop) chips read like 4-7-90-K, 4th generation, i7, last iteration (highest performance variant), unlocked (non-K versions are not enabled for overclocking). Then 8-7-00-K would be 8th generation, i7, first iteration.
AMD copied this to an extend. 2-9-50X would ve second generation Ryzen, i9 counterpart, mid-level, but -X suffix seem to mean slightly improved performance (all AMD chips were unlocked for overclocking).
nVidia is similar 970 would be 9th generation GTX, second highest level (geared towards gamers with mid-to-large budgets), while 10-80 Ti would be 10th generation GTX, highest level (geared towards people with serious money), and updated (Ti) edition.
In general, generation increases add significant power reduction, allowing less running cost, and higher performance for the same price. In fact a future i3 might be better than a previous i5.
(I'm skipping Pentium/Celeron which are lower binned silicones of the same design, and Atoms, and of course Xeon server and workstation chips).
Looking at Wikipedia for the CPU/GPU generation gives sufficient detail for differences between offerings. If I'm planning to purchase a CPU to use for many years, I would benefit spending some time understanding those differences.
https://en.wikipedia.org/wiki/...
A table like the above, will explain the breakdown of the numbering. It hasn't changed much with each generation since the core series lineup came in 10 years ago.
Here is a better breakdown with more words than numbers. https://www.intel.com/content/...
But perhaps youre a casual and that's all a bit too esoteric for you?
If you want to easy it up, just go to www.cpubenchmark.net and you can easily compare all cpus and pricepoints. Look at single thread performance if that's all your application can handle (or you are a gamer..), and total performance if its multithreaded. There is a wealth of user submitted data there that i would never view processor advertisements without.
Its really not something you need to spend more than an afternoon getting acquainted with. An exercise that anyone who wants to spend $500+ on a new PC should be more than willing to do. As others have said, basic research is important when buying most things.
-
This is not specific to CPUs and GPUs, but I am sick of model names/numbers being reused for different products. I was browsing Dell's website recently and found it frustrating to find that the Inspiron 3000 series comes in 15" and New 15" varieties. The 15" variety could use either a Celeron or Pentium processor, while the New 15" ones could be either 7th or 8th generation i3, i5 or i7. Why have three 0s in the number if you are never going to change any of those digits?
And don't get me started on the Inspiron 5000 15" and New 15" range or the Inspiron 7000 15" (but no New 15") range. There is also the Nvidia GTX1060, which comes in two varieties that performs differently.
It's already bad enough that hardware manufacturers tweak and skew their drivers to eke out another dot at some artificial benchmark program, I don't want them to actually produce their hardware to fit an arbitrary metric that has nothing to do with real world problems because they have to since some illiterates want to compare numbers instead of finding out what they mean.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.