[MBCook]

I kind of like what Nvidia has been doing in GPUs. They have series. 1000s, 2000s, 3000s. Each generate is a new series.

Then the products inside seem to be in performance order. A 3080 is better than a 3060 is better than a 3020 (if that exists).

I know it gets messier than that. The fact that they add the titanium versions kind of screws with it. And in the last year they had that card they had to renumber because a change to its memory configuration (?) made it perform wrong for its moniker.

But at least you know a 3080 is two generations newer than a 1080. Compared to the seemingly random numbers on CPUs it’s simple.

[p1necone]

You've touched on the main problem (imo) that seems to happen with large hardware companies adopting consistent naming for performance segments in your comment. Once you've got a sane naming scheme and you've used it for a few generations the incentive to use it to mislead consumers is very high.

Why sell this generations X070 equivalent as an x070 when you can call it an x080 instead and know that a very large chunk of your customer base wont look at the actual specs before paying an extra ~$100 for it? Probably not good in the long term, but who cares about the long term when you can boost sales numbers on a new product now and make shareholders happy.

[gxt]

Let's just adopt an open-sourced hardware naming scheme then.

[gxt]

Not sure I understand all the downvotes and replies.. 5 secs of gpt4 later:

A naming scheme:

Manufacturer Code: A short abbreviation or code for the manufacturer, e.g., 'IN' for Intel, 'AMD' for AMD, 'NV' for NVIDIA, etc.

Component Type: A short abbreviation for the component type, e.g., 'CPU' for Central Processing Unit and 'GPU' for Graphics Processing Unit.

Generation: A two-digit number representing the generation of the component. For instance, '01' for the first generation, '02' for the second generation, and so on.

Performance Tier: A letter representing the performance tier of the component, with 'A' being the highest tier and 'E' being the lowest tier. This can be expanded to include more tiers as needed.

Sub-Tier: A two-digit number representing the sub-tier within the performance tier, with '01' being the highest sub-tier and '99' being the lowest sub-tier.

Examples:

IN-CPU-03-A05: This would represent an Intel third-generation CPU, in the highest performance tier (A), and positioned in the fifth sub-tier within that tier.

NV-GPU-02-C15: This would represent an NVIDIA second-generation GPU, in the middle performance tier (C), and positioned in the fifteenth sub-tier within that tier.

[RealStickman_]

An 01-A part should have the performance of a 02-B part while being less efficient.

[red-iron-pine]

I'm all about the FOSS but I can't see what an "open source naming system" would due for us or how that would work.

How do we account for nuances on AMD, Intel, or Qualcomm chips? Does that apply to other ARM chips?

And then there is the issue of trademark, copyright, and branding, which could get into legal struggles.

[jdiff]

That's kind of a nonsensical suggestion. You can't define this in a way significantly different than it already is that makes sense across devices. You can use metrics like "% of performance vs flagship product," which a lot of people have used to point out that lately the xx80-class-turned-xx70-class cards of today are looking more like xx60 Ti-class cards of yesterday in relative performance, but performance isn't always that simple and there's no way any company will conform to it. Even if they use it, there's no incentive not to break from it the same way they do now.

[Infernal]

It sounds like the real issue here with Intel is they got it backwards. Nvidia leads with the generation (2xxx, 3xxx) and then differentiates within the generation (xx60, xx80). Intel leads with the intra-gen differentiation (i5, i7 etc.) which is most often included in marketing, with the generation buried in the particular CPU model number that often isn't included (10xxx, 11xxx).

It doesn't help that there's also a random mix of river codenames and I can never keep straight which number they belong to.

[toast0]

Intel gets more confusing when you include the Pentium and Celerons. There, they have different leading numbers for generations compared to the iX. And if you get mobile chips, you've got trailing suffixes like U and H? that indicate a 'core series' cpu, and leading prefixes like N that indicate a 'atom series' cpu.

The codenames are hard to keep straight, especially when they started making everything a Lake, but at least you can see then list of products formerly known as X and see the whole family.

The basic process for CPU selection is pick the architecture, pick the number of cores, pick the speed tier. But none of that is clearly communicated with the model number.

[NBJack]

Let's not forget what they started calling 'gold', which in many cases is just Celeron+. But don't confuse that with 'silver', designed for ultralight computing.

Honestly, I'm somewhat convinced that Intel did this scheme intentionally to sell older hardware (hey this has an i5/'gold' too and it's slightly cheaper!), but perhaps it backfired: folks who already have an iX think that new fancy iX laptop isn't worth it.

Here's hoping there's a tell-all book in a few years with some insight.

In the grand scheme of things, it's still better than the Microsoft Xbox naming convention.

[toast0]

> Let's not forget what they started calling 'gold', which in many cases is just Celeron+. But don't confuse that with 'silver', designed for ultralight computing.

Both Gold and Celeron don't really mean anything. Celeron is often available with the same architecture as the other brands, often with less cores, maybe less speed, maybe some features disabled, but especially if you're comparing across architectures, the brand isn't important, the details are.

[userbinator]

In other words, "middle endianness".

[masklinn]

Meh. The I is the segment, the is the generation, then the actual part with the generation.

Nobody is surprised that a 3-series can refer to different cars, and the exact model (E46, G20) is not mentioned in marketing.

> It doesn't help that there's also a random mix of river codenames and I can never keep straight which number they belong to.

I’ve never known codenames to do that. Plus they don’t necessarily match generations.

[klodolph]

The problem with Nvidia is that I blinked and the numbers wrapped around, so a GeForce RTX 4070 is really new, but a GeForce FX 5200 is really old.

[p1necone]

Same! I still remember the 8800gtx being "current" high end and the 2XX cards being the recently announced space-age new ones.

Then I stuck with AMD for a while (the HD 4870 followed by the R9 390, both were very compelling options from a $/perf standpoint when they were current from what I remember) and by the time NVidia came back on my radar again they'd gone through another whole ~10 generation cycle of 3 digit numbers and were back on the four digit ones again.

[iforgotpassword]

Ahem, the same happened with ATI, they just were lucky to have been renamed to AMD on the way.

[p1necone]

I wasn't saying they didn't, just recounting my personal trajectory in the GPU model space.

[iforgotpassword]

Yeah, I think I actually wanted to reply to your parent comment and messed up

[gambiting]

That's a pretty long blink to be fair - there's 20 years(!!!!) gap between the two.

[NBJack]

At least the letters changed.

Would you like to buy my Xbox One X Box?

[modeless]

If you think Nvidia's product naming makes any kind of sense, try a few rounds of my GPU quiz: https://james.darpinian.com/gpu-quiz.html

[mhh__]

I did it for 12 questions or so and didn't get any wrong.

The naming schemes are completely, completely, retarded but they also aren't that hard to guess

[andromeduck]

Took ~30 for me to get one wrong.

[userbinator]

But at least you know a 3080 is two generations newer than a 1080.

That's similar to what Intel was doing with the numbers following the i{3,5,7,9}:

https://en.wikipedia.org/wiki/List_of_Intel_processors

[marak830]

I believe his point is, that the I(3/5/7/9) moniker represents the chip power relation, but which generation of the chip is hidden, unlike with say the GeForce cards with the 10xx/20xx/30xx.

[leetcrew]

> But at least you know a 3080 is two generations newer than a 1080. Compared to the seemingly random numbers on CPUs it’s simple.

CPU SKUs work exactly the same way? intel has gone longer than nvidia without restarting the numbering, so they need two digits instead of one.

[alpaca128]

"i7-6700" could also be "6700" without any loss of information if the numbering follows a consistent system. Instead it's needlessly verbose so everyone just shortens it to "i7" which carries little information.

> intel has gone longer than nvidia without restarting the numbering, so they need two digits instead of one.

Nvidia started the current numbering with 200 almost at the same time, with the 10XX series they added a fourth digit.

[abricot]

It's not just the Titanium, they have also launched 1650 and 1660 in between the 20s and 30s generation, and then you have the SUPER, TI and Foundation models.

AMD could also do better.

[tracker1]

They've (AMD) made their laptop CPU numbering even worse.