Hacker News Reader

Comment by glimshe | Hacker News Reader

It feels like the people doing the benchmarks are stuck in time. POV-Ray benchmarks but not AI? Most people today, including enthusiasts, don't even know what POV-Ray is.

View on HN

It's a gaming chip, literally a lower-cost downgrade of an AI chip meant for gamers instead of AI.

Why would it benchmarked for AI in these circumstances?

Beside, the second page look like those unnecessary AI benchmark to me. I wouldn't recognise such for certain or be aible to estimate their quality, but it's what their summary say.

Actually it would be very interesting to know how it performs for AI and whether it has a better value.

I understand where you are coming from.

But as a gamer who hasn't been able to make a significant upgrade to my rig in 6 years due to memory price and redirected processor engineering funding (all of it due to AI), it really bother me.

Weren’t the prices normal until about a year ago?

It's not a gaming chip, it's a desktop CPU chip.

Imagine thinking that all people use PCs for is games...

True, I suppose it is also useful for video editing and other uses.

The article describe it as a gaming chip, though. Might be because of the site's target audience? I don't know them.

> Might be because of the site's target audience?

Yes, but also because of official AMD marketing:

https://www.amd.com/en/products/processors/desktops/ryzen/70...

> with elite gaming performance from a powerful 8‑core AMD gaming processor

I would also call it a gaming CPU, because it is what it excels at. ¯\(ツ)/¯

It is indeed a desktop CPU chip, but the 3dV cache is real good for gaming, thus it's also marked in that region.

From the official AMD website:

> Whatever the setting or resolution, lead your team to victory with elite gaming performance from a powerful 8‑core AMD gaming processor featuring AMD 3D V‑Cache™ technology.

I assume they meant that gaming is the main market segment it serves, which from my understanding is correct. The X3D chips have great performance in games, which are generally constrained by memory latency. It's, of course, still a general purpose CPU that can do many things.

X3D is marketed as a gaming feature and benchmarks bear that out. Productivity workloads (video editing, compiling code, etc) benefit very little from the extra cache. You're better off buying a chip without it and a higher core count.

I'm usually one to say the same thing, but unless you're doing something that specifically hits instruction cache hard, for most productivity workloads (compiling, video editing, etc) you're better off foregoing the cache and getting a CPU with more threads.

Why would they benchmark it in POV-Ray if gaming is all that matters?

People who buy computers may have a primary use for them but it's useful to know performance in their secondary workloads.

What kind of AI benchmarks would you like to see?

I don't know because I'm not a specialist in benchmarks. The article authors are supposedly specialists, so I'd expect them to identify appropriate benchmarks to make statements such as "while this is a good value for games and general purpose use, it fails for local AI as these benchmarks show", for example.

Reply on news.ycombinator.com