lexluthermiesterOz did a comparison with different PCIe gen speeds..

While still promising, a $200 6400XT/6500 SFF low profile GPU is needed.

very good aport meanwhile still waiting for review from da house

:)

Bomby569this is a card that gets beaten by a 6 year old 1050ti that now should cost way less, about half or even less i would say, and be bette value if it weren't the crazy market. A 6 year old low end card. And it launchs at a higher price point 6 years latter.

Where are you getting that from? While I haven't seen any good reviews with broad comparisons, the consensus seems to be that it's ~30% slower than the 6500 XT, which beats the ~40% slower 1050 Ti. On gen4 it roughly matches the 1650, in comparison to which the 1050 Ti is 20% slower, with a slight reduction in performance on gen3.

lexluthermiesterThere's no way for you(or anyone else outside of AMD) to know if they're 64bit/4x by design or whether they're just artificially crippled higher end parts.

Sure, except that this die is rather well known at this point, and no iteration of it in either mobile or desktop form factors has any more than an x4 PCIe link and 64-bit wide memory bus, including the workstation Radeon Pro W6400. Factoring in that TSMC's 7nm and 6nm yields are borderline perfect, yields are not an argument for cutting these things down from a theoretically higher bandwidth interface. There are other possible reasons, such as power savings or cost savings, but neither of those would apply to the RX 6500 XT - a high priced and ridiculously clock-pushed SKU. There is zero evidence to suggest that Navi 23 has any more I/O than what we've seen in products to date.

lexluthermiesterYou can't and don't know that.

.... sigh. No, I can't. But we can apply some basic logic on top of known facts. Some facts:
- The Navi 24 RX 6500 XT exists.
- TSMC's 7nm yields are so good that AMD's cut-down SKUs have consistently been in short supply for the past two years, including CPUs. (Also note the complete absence of <6-core Ryzen CPUs that aren't based on APUs - this tells us that from the millions of Zen3 dice produced, the number with 3 or more broken cores is so low as to not warrant making a product out of them.
- TSMC's 6nm process is a refinement on 7nm, and yields are likely comparable.
- It is extremely rare for chipmakers to bin a higher end chip for a SKU below one using a lower end chip. It has happened - Nvidia's 2060 KO was an example - but it's exceedingly rare. And when it happens, there are typically well documented reasons. I can't think of a single case where this has been done on anything but a very large high-end die (such as the TU104), but I might be wrong. Still, we know of no such reasons for this.

The aforementioned logic:
We have so far seen zero evidence to suggest Navi 24 has any more I/O than what we know of. It is thus safer to assume that it has no such I/O, even if this isn't technically known. There's also a fab supply crunch going on, and AMD really needs a contender for budget gaming laptops, which sell in massive volumes. That incentivises tiny die sizes to increase per-wafer die counts. Cutting I/O directly impacts die sizes, and Navi 24 is indeed quite tiny. Narrow interfaces also save power, which supports another perspective on this die: Its lack of a media engine strongly suggests that this is designed specifically for pairing with an APU, in a mobile device, which would make such hardware redundant - and cutting it again brings down die size. It is thus extremely likely that in terms of design goals, all non-mobile implementations of Navi 24 are secondary. Note that the 6500M and 6300M were launched well before the 6500 XT. Both have 64b+4x interfaces.

So: while we don't know that Navi 24 is strictly 64b+4x, all evidence available points this way, with zero evidence suggesting otherwise. All signs point towards this being a small-size-and-low-power-above-all-else, mobile-first design, with the ridiculously clocked 6500 XT being where all less efficient dice are shunted to. The 25W 6300M and 35-50W 6500M are the point of this die, and everything else is just a way to sell off remaining or non-qualifying stock. Which means there is no reason to expect it to have yet-to-be-enabled I/O.

lexluthermiesterAnyone buying this card(or a card in this price bracket) is not going to care about that feature.

I agree, but it's still a weakness. You might note that I'm not one of the people shouting "this lacks a media engine, it's trash!", I'm simply pointing it out as one of its weak points. I wouldn't care, but if you're building a low-cost media center off an old CPU, it would obviously be a deal-breaker. But for the type of gaming it's aimed at? Not at all.

Valantarthis is designed specifically for pairing with an APU, in a mobile device, which would make such hardware redundant

I believe this was later confirmed by AMD to be true

ValantarNote that the 6500M and 6300M were launched well before the 6500 XT

Were they? I think they all launched during CES at the same time but it's besides the point ofc.

Valantarthis lacks a media engine

It does have basic decode (h264, h265, sadly no av1 but that's a new thing for this generation anyway), so fine for htpc

www.amd.com/en/products/graphics/amd-radeon-rx-6400#product-specs

trsttteI believe this was later confirmed by AMD to be true

I think so too (in an interview at CES I think?) but I couldn't find the source with 30 seconds of googling so I gave up.

trsttteWere they? I think they all launched during CES at the same time but it's besides the point ofc.

The 6500 XT was teased at CES but had its launch event January 19th. 6500m+6300m launched at CES.

this card has still drivers for windows 7 which is very diy-friendly :)

ValantarI think so too (in an interview at CES I think?) but I couldn't find the source with 30 seconds of googling so I gave up.

Google search gave the below result, so at 6500XT disastrous launch day when AMD saw all the negative press, an AMD employee posted a comment to justify the encoding & PCI-E design decisions, saying that it was primarily a laptop companion dgpu:

wccftech.com/amd-navi-24-rdna-2-gpu-for-radeon-rx-6500-xt-designed-primarily-for-laptops/amp/

videocardz.com/newz/amd-employee-confirms-navi-24s-primary-use-is-in-laptops-with-rembrandt-apus