AMD’s HD 6950 vs. RX 6900 XT: What Does Adding 50 Do? | Hacker News Reader

AMD’s HD 6950 vs. RX 6900 XT: What Does Adding 50 Do?

chipsandcheese.com

75 pointsgautamcgoel29 comments

Note: AMD Radeon HD 6950 was released in year 2010 with 2.64 billion transistors. AMD Radeon RX 6900 XT was released in year 2020 with 26.80 billion transistors.

https://www.techpowerup.com/gpu-specs/radeon-hd-6950.c405 , https://www.techpowerup.com/gpu-specs/radeon-rx-6900-xt.c348...

> AMD should not continue to increase model numbers. Doing so can decrease performance across a variety of metrics, as data from the HD 6950 and RX 6900 XT shows. Instead, they should hold the model number in place or even decrease it to ensure future generations perform optimally.

Lmao. This was a masterpiece of an article.

yeah. labeling stuff will make your future products perform good or bad, depending how you label your current stuff, everything else being equal

I've run over 120k AMD GPUs. They are ALL snowflakes. Even within the same model number, no two are the same and they can vary quite a bit in performance. It is actually kind of wild to see this at scale. We called it the 'silicon lottery'.

I started to write code to gather data on location on the wafer that it was printed on to see if I could do a correlation there (as I heard a rumor that that the edges aren't as good), but sadly never got a chance to finish it up.

The ones from around the edge (you can tell by the distintive curve on one side) are sold as 6900 AF's

The AF (as f**k?) suffix is really powerful: it gives you 1 generation of performance increase.

Depends on the customer I guess. For me I'll always choose AMD for its bulletproof Linux support.

What does that have to do with their performance variance?

My comment wasn't intended as negative against AMD, just an interesting note about GPU performance measurements.

I've noticed this to some extent with their CPUs as well.

From what I can tell this is, in part, because AMD run their chips much closer to their physical limits than say, Intel or Nvidia do. They're all redlining so to speak.

As such this means that any differences in the silicon are much more likely to have an impact on performance.

I can't make sense of this. CPUs microarchitecture is run at their clock speed. If they're running close to their physical limits, either they don't fail in which case all's well, or they fail in which case it's a wipeout such as a blue screen, reboot, freezes, corruption or whatever. AFAICS they can't run correctly but somehow slower. Be good if you could explain where I'm getting it wrong.

> run at their clock speed

The clock speed for a CPU is dynamic. Both Intel and AMD CPUs will boost their clocks to as high as possible given the current conditions.

Some will boost higher than others, even in the same batch. They are even labelled "up to" like a cable internet connection.

You're right, thanks.

It's really interesting to note how much they kneecapped 64-bit performance on consumer cards to stop people from using them to run simulations.

This was a very funny piece!

>Radeon RX 7900 Ti XTX SE Pro VIVO Founders Edition

I love the absurdity of this so hard.

“Some of this can be attributed to the process node too. On that note, TSMC’s 40 nm process has a bigger number than their 7 nm process. But again, bigger numbers don’t help.”

AMD's numbering scheme is probably the biggest reason I've bought a lot more Nvidia cards. It's just less work to figure out what I'm getting.

Nvidia has you covered too with the A6000.

The 6950 was my first powerful graphics card. It was a great card

I bought it used on eBay & it had already been abused. One day it died, and the company that made it told me they had no replacements, so they sent me a brand new 7950 no questions asked.

That company was MSI. I went on to recommend MSI to others for years because of that incident & one other good reason.

Then in 2016 MSI dicked me on a $2200 laptop that had an ungodly amount of issues straight from the factory any decent company would RMA. The amount of reports on the internet about those specific models having issues were insane, & the company just did not care in the slightest.

I’ve since denounced MSI after that. I presume MBAs had at some time taken over & ran customer service into the ground.

I still have the old 6950 in storage somewhere.

Some point about how to both gain and lose a customer in this text, I think. I mean, over-tightening a laptops hinges in the factory, to the point they one day suddenly & catastrophically fail, dropping your 1440p 120hz screen - knowing of the issue & not informing customers, let alone an RMA - is just fukt. and that was only one of the many, many issues (coil whine & completely inadequate cooling being some others)

That sucks. I bet it was stressful as heck having issues with $2200 worth of kit.

I've started to treat all purchases as a bit of a lottery. Things vary so much between product cycles and support quality.

Sadly, I resigned to buying tech that is cheap enough I could replace without losing sleep.

That's how I buy tech too, anymore.

As cheap as possible while not frustrating to use doing the job.

No matter what I pay, there's still a chance the thing falls apart in 3 months and catches fire anyway.

Knew I was in for a ride when I saw "VLIW". Fun to see how a card that old still has competitive compute throughput for certain workloads (obviously, it falls short in many other areas)

And for 6800 series you can compare to the GeForce one too.

AMD's naming is rough. I upgraded from an FX 6300 to a Ryzen 3600 a few years ago and it still hurts my brain keeping which one is which straight.

Interestingly on paper the two chips look surprisingly similar - they're both 6 cores at ~3.5Ghz. Multi threading and modern IPC make a big difference!

LOL how the hell did they make this URL work?

https://www.amazon.com/set-your-computer-on-fire/dp/B01DV1Z3...

If I had to guess, the number at the end is the ASIN (Amazon Stock Indicator Number), this is unique among all items listed on the site.

I suspect that as long as that number is in the correct location, the page will load the right item.

Amazon only checks the id after the /dp/. Stackoverflow does a similar where only-the-id matters url scheme so that posters can change the title.

"Normally we don’t look at first level cache bandwidth scaling, because it’s boring"

An informative, well written article to start the year.

Chips and Cheese, tip of the hat on 1 April 2023. Took me a few seconds to figure this one out.

Reply on news.ycombinator.com