Is this stupidity or no?

W

wandplus

Gawd
Joined
Jan 14, 2020
Messages
599
Who else is fed up that both Nvidia and AMD refuse to sell their 4050 or 7500 at the same time as the other cards? Wouldn't they have more sales more often with lower-priced cards or who's wrong here anyway? Seems like they're making everyone wait longer and losing out on more sales anyway. Or are they trying to make their higher-end cards look more exclusive?

Is there a business course that teaches this?
 
wandplus said:
Who else is fed up that both Nvidia and AMD refuse to sell their 4050 or 7500 at the same time as the other cards? Wouldn't they have more sales more often with lower-priced cards or who's wrong here anyway? Seems like they're making everyone wait longer and losing out on more sales anyway. Or are they trying to make their higher-end cards look more exclusive?

Is there a business course that teaches this?
Click to expand...
Yeah, any business course.

They do have more sales more often with the lower end stuff. But they are always based on the high end design, which has higher margins, and had to be designed first.

To do what you're wanting, they would have to wait to release their finished high end designs until they had completed the low end designs. This would be stupid
 
The longer they make you wait, the more chances of you caving and buying a higher end (higher priced) option. This is a common practice. Maximizing profit margins.
 
AFAIK, Nvidia and AMD are still supply constrained on chip fabrication. High tier product has better margins, so it doesn't make sense to make low tier products with the most recent fab. I don't like it, but it is what it is.

Used to be they'd release a low tier model in the same number scheme, but it was still a last gen part. I prefer the honesty of just not filling in those numbers until they're ready.
 
At this point, low end add-in GPUs seem to be making less and less sense considering they aren't $100 or less on release. Considering wafer costs, although down, it probably makes little sense to even make a 4050.

If you think about how the tiering has been this gen compared to prior generations precedence, the 4060 is already kinda of occupying that 50-class tiering spec wise. The 4060 is already on AD107 and is 8GB over 128-bits. The 3050 was 128-bit 8GB on GA106. GTX 1650 was 128-bit card on TU117. GTX 1050/1050 Ti was 128-bit on GP107. GTX 950 was 128-bit on GM206. Seems to be a pattern here.

As for AMD, I'm sure they will just replacing the low-end with APUs. Considering the 6500 was worse than the 5500, yet they wouldn't even price low enough to make sense (like $100 or less), really how does it make sense to make a 7500 that will be either the same or even worse than the 6500 and cost too much for what it is?
 
toast0 said:
AFAIK, Nvidia and AMD are still supply constrained on chip fabrication. High tier product has better margins, so it doesn't make sense to make low tier products with the most recent fab. I don't like it, but it is what it is.

Used to be they'd release a low tier model in the same number scheme, but it was still a last gen part. I prefer the honesty of just not filling in those numbers until they're ready.
Click to expand...
Yeah, it is partly this, but also dictated by the yields.

If they are getting really good yields, might only be a few pieces of silicon per wafer bad with enough defects to only be usable for use in the lowest end parts. It will take time to get sufficient supply to bother making those into cards.

They are certainly not going to take a chip good enough to be made into a 4090 and disable it on purpose to instead make a 4050.

They have done this many generations ago when fab costs were far less expensive and far less constrained, but with TSMC the only fab that can currently make these highest end chips, and the demand for the fab time also the highest it has ever been, purposely nerfing chips into low end parts is likely a thing of the past. Eventually they will have enough that those cards get made, but they will not be released at the same time as the better silicon parts launch. And some low end cards will be previous gen mid to upper parts re-branded, if there is stock left to allow it. This is a smart way to go, keeps 2 generations of silicon moving.
 
Geez, maybe I should consider myself lucky I got a GTX 1650 while they still existed o_O
 
GoodBoy said:
They have done this many generations ago when fab costs were far less expensive and far less constrained, but with TSMC the only fab that can currently make these highest end chips, and the demand for the fab time also the highest it has ever been, purposely nerfing chips into low end parts is likely a thing of the past. Eventually they will have enough that those cards get made, but they will not be released at the same time as the better silicon parts launch. And some low end cards will be previous gen mid to upper parts re-branded, if there is stock left to allow it. This is a smart way to go, keeps 2 generations of silicon moving.
Click to expand...
They don't cut them down that far - each of the last 3 generations has eventually spawned 5 distinct dies. This time around the 4060 and 4060Ti each have their own die. They do still cut them down and sometimes use more than one die in the same model, but not at the 4090->4060 level. Like GA-102 was all the 3080 and 3090 variants, plus some partial duds ended up in 3070Tis. Lately NV hasn't rebadged earlier series as something new, though they've done that in the past.

I'm thinking that RT is what's really getting in the way of a decent budget card. If the card isn't powerful enough to make use of RT it's just a waste of cost and die space.

NV's been having trouble with low end chips since Turing. I suspect a lot of the reason is ray tracing. RTX cards are supposed to do ray tracing, at least from NV's point of view. Adding ray tracing resulted in a massive increase in die size and cost. A 1060 had a 200mm^2 die. Launch price was $200 for the 3GB model and $250 for the 6GB. The 6GB 2060 has a 454mm^2 die (bigger than a 1080 and almost as large as a 1080Ti/Titan X) and launched at $350. That was as low as they could go and make RT usable, and it still wasn't good for much. RT on an RTX 2050 wouldn't have been useful so rather than making such a cruel joke of a card they ripped out the ray tracing and DLSS support and made the GTX 1600 cards.

Since then the "6" card die sizes and launch MSRPs have actually been coming down - $330 for the 3060 and $300 for the 4060 - but they still can't seem to get usable ray tracing much under $300. Sure they made a 3050, but it's not really a ray tracing card and costs too much for what you get in raster performance. 3060s are still around, but only slightly cheaper than 4060s. Now that they have a usable RTX card down to $300 perhaps they can make a "5" with usable RT. Blackwell leaks say a 5050 is coming. 128-bit bus, probably 8GB GDDR6.

AMD made the low end cards last generation, and they all sucked. Anything below the RX 6600 ended up being worse than the previous generation. RT strikes again? The RX 6500XT and RX 6400 have RT support, not that anyone would think of using it. It's harder to tell how much adding RT affected AMD's cards as they launched in the middle of Covid and the chip shortage when prices for any GPU were going berserk if you could even get one without paying a scalper. On the up side, at least the RX 6600 8GB is down to $200. I don't expect a "5" at $150 anymore, and $200 for a card that's ballpark 80% of a $250 RX 7600 or $300 RTX 4600 seems ok.

Last gen sticking around may be where we're going with the lower end of the GPU market if the CPU business is any indication. AMD has kept AM4 around for a while to provide a cheaper option, and they're rumored to be backporting Zen 4 to TSMC 7nm to make low end chips for AM5 on a cheaper process. Think Athlon, Ryzen 3, etc. here. Then Intel has split their CPU product lineup into "Core Ultra" and "Core". So far the laptop chips are new stuff (Meteor Lake, Intel 4) for Core Ultra and another Raptor Lake refresh on Intel 7 for Core and the leaks say the desktop chips are going to go the same way. If Intel and AMD are doing that for CPUs I really wouldn't find it surprising if AMD and NV basically pulled an "AM4" for budget vid cards.
 
zandor said:
They don't cut them down that far - each of the last 3 generations has eventually spawned 5 distinct dies. This time around the 4060 and 4060Ti each have their own die. They do still cut them down and sometimes use more than one die in the same model, but not at the 4090->4060 level. Like GA-102 was all the 3080 and 3090 variants, plus some partial duds ended up in 3070Tis. Lately NV hasn't rebadged earlier series as something new, though they've done that in the past.

I'm thinking that RT is what's really getting in the way of a decent budget card. If the card isn't powerful enough to make use of RT it's just a waste of cost and die space.

NV's been having trouble with low end chips since Turing. I suspect a lot of the reason is ray tracing. RTX cards are supposed to do ray tracing, at least from NV's point of view. Adding ray tracing resulted in a massive increase in die size and cost. A 1060 had a 200mm^2 die. Launch price was $200 for the 3GB model and $250 for the 6GB. The 6GB 2060 has a 454mm^2 die (bigger than a 1080 and almost as large as a 1080Ti/Titan X) and launched at $350. That was as low as they could go and make RT usable, and it still wasn't good for much. RT on an RTX 2050 wouldn't have been useful so rather than making such a cruel joke of a card they ripped out the ray tracing and DLSS support and made the GTX 1600 cards.

Since then the "6" card die sizes and launch MSRPs have actually been coming down - $330 for the 3060 and $300 for the 4060 - but they still can't seem to get usable ray tracing much under $300. Sure they made a 3050, but it's not really a ray tracing card and costs too much for what you get in raster performance. 3060s are still around, but only slightly cheaper than 4060s. Now that they have a usable RTX card down to $300 perhaps they can make a "5" with usable RT. Blackwell leaks say a 5050 is coming. 128-bit bus, probably 8GB GDDR6.

AMD made the low end cards last generation, and they all sucked. Anything below the RX 6600 ended up being worse than the previous generation. RT strikes again? The RX 6500XT and RX 6400 have RT support, not that anyone would think of using it. It's harder to tell how much adding RT affected AMD's cards as they launched in the middle of Covid and the chip shortage when prices for any GPU were going berserk if you could even get one without paying a scalper. On the up side, at least the RX 6600 8GB is down to $200. I don't expect a "5" at $150 anymore, and $200 for a card that's ballpark 80% of a $250 RX 7600 or $300 RTX 4600 seems ok.

Last gen sticking around may be where we're going with the lower end of the GPU market if the CPU business is any indication. AMD has kept AM4 around for a while to provide a cheaper option, and they're rumored to be backporting Zen 4 to TSMC 7nm to make low end chips for AM5 on a cheaper process. Think Athlon, Ryzen 3, etc. here. Then Intel has split their CPU product lineup into "Core Ultra" and "Core". So far the laptop chips are new stuff (Meteor Lake, Intel 4) for Core Ultra and another Raptor Lake refresh on Intel 7 for Core and the leaks say the desktop chips are going to go the same way. If Intel and AMD are doing that for CPUs I really wouldn't find it surprising if AMD and NV basically pulled an "AM4" for budget vid cards.
Click to expand...
AMD being limited to Navi 42 and 43 this coming generation, no high Navi 41 makes them compete against their last gen parts from the start pushing down those prices much more than if they first released a high-end card and use last generation as the lower end cards until supplies are exhausted. Which to me indicates Navi 41 was a complete failure. First or ideally, you release a card that beats everything you had before and sell it at a premium. Now, even their Navi 42 most likely won't beat the 7900XTX and even the 7900XT. So AMD is forced to sell it lower which then conflicts with what is already in the channels.

Low end cards are still being sold, just that their asking prices has more than doubled from previous generations. A low end card today is over $250 and probably will be more next generation. Talking about current generation and beyond. More budget orientated folks may just stick with APU's in the future.
 
You must log in or register to reply here.
Back
Top