When will gpu prices return to normal.

[ARF]

This is exactly why, barring a > $50 price difference in AMD vs Nvidia, I'd personally still go with Nvidia.

And nothing in the midrange is priced reasonably, at all.

Given this is a 2 year product cycle, this represents the last 4 years of the midrange, and it is not impressive at all. With a 2 year product cycle, we should see 40% uplifts the way the high end GPUs are doing.

But for Halo products, they just keep making higher and higher end models, while the mid range kind of languishes. What do we get next, the 4950XT Ti Super Titan?

And yes I mixed up AMD/Nvidia naming convention intentionally.


View attachment 261197

The reason for this difference between Turing and Ampere is that all Turing dies were very large:

RTX 2080 Ti - TU102 - 754 sq. mm.
RTX 2080 - TU104 - 545 sq. mm.
RTX 2060 - TU106 - 445 sq. mm.

While with the RTX 3060, nvidia tried to return to the more normal die sizes:

RTX 3060 - GA106 - 276 sq. mm.

GA106 is a virtual die shrink of the TU106.
When there are direct shrinks, there is no high-performance improvement.

 

[Count von Schwalbe]

FHD vs QHD vs 4K (or UHD) makes the most sense.

$700 for a GPU should get you into 2160p territory just fine. Just don't set your settings to Ultra - that's just stupid anyhow. The next setting down is typically visually indistinguishable yet performs far, far better. Reviews tend to test at Ultra because they want to test worst-case scenarios for proper stress testing, but gaming at Ultra makes no sense.

Yes!

Also, you would be better off getting a better monitor at a lower resolution if buying a midrange GPU.

GA106 is a virtual die shrink of the TU106.
When there are direct shrinks, there is no high-performance improvement

Small performance bump from clock speeds, usually. However, GA106 has 3840 shading units, vs the 2304 of TU106. The 3060 has around 50% more than the 2060.

What I can't understand is why the GPU manufacturers are not more focused on improvements to the highest volume markets - the 60-70 class units. Mindshare/press coverage maybe?

 

[ARF]

Small performance bump from clock speeds, usually. However, GA106 has 3840 shading units, vs the 2304 of TU106. The 3060 has around 50% more than the 2060.

What I can't understand is why the GPU manufacturers are not more focused on improvements to the highest volume markets - the 60-70 class units. Mindshare/press coverage maybe?

If I'm not mistaken two Ampere shaders are equal to one Turing shader. They were made to be smaller and less performance.

 

[80-watt Hamster]

FHD vs QHD vs 4K (or UHD) makes the most sense.

Yes!

Also, you would be better off getting a better monitor at a lower resolution if buying a midrange GPU.



Small performance bump from clock speeds, usually. However, GA106 has 3840 shading units, vs the 2304 of TU106. The 3060 has around 50% more than the 2060.

What I can't understand is why the GPU manufacturers are not more focused on improvements to the highest volume markets - the 60-70 class units. Mindshare/press coverage maybe?

My hypothesis, generated on the fly during a conversation with a friend yesterday, is the confluence of high die yields with strong demand for high performance cards. Back in The Day(tm), the number of customers willing to pay top dollar for the top cards was limited. Chipmakers would resort (allegedly?) to fusing off fully-functioning dice to have chips for cards lower in the stack. Then the PC gaming resurgence came around, coupled with the crypto boom. All of a sudden there was functionally unlimited demand for the highest performing chip either manufacturer could muster, even beyond the traditional one-upsmanship. From AMD/Nvidia's perspective, that doesn't leave much incentive to drive value in the midrange. AMD was the champion there in recent years because they didn't have anything that could compete with Nvidia in the high end, so needed to lean on the value play. Then the above happened, coupled with legit advancements of their own, and that wasn't necessary anymore.

 

[ARF]

My hypothesis, generated on the fly during a conversation with a friend yesterday, is the confluence of high die yields with strong demand for high performance cards. Back in The Day(tm), the number of customers willing to pay top dollar for the top cards was limited. Chipmakers would resort (allegedly?) to fusing off fully-functioning dice to have chips for cards lower in the stack. Then the PC gaming resurgence came around, coupled with the crypto boom. All of a sudden there was functionally unlimited demand for the highest performing chip either manufacturer could muster, even beyond the traditional one-upsmanship. From AMD/Nvidia's perspective, that doesn't leave much incentive to drive value in the midrange. AMD was the champion there in recent years because they didn't have anything that could compete with Nvidia in the high end, so needed to lean on the value play. Then the above happened, coupled with legit advancements of their own, and that wasn't necessary anymore.

I disagree. We have a duopoly and AMD didn't master anything at all. We see the same old performance from AMD and no improvement at all.
I mean look at the 5500 XT (2019) -> 6500 XT (2022) transition. 0% progress. None!
It even looks that AMD intentionally stagnates the market, so that no "average joe" to ever be able to upgrade from the ancient 1080p monitor.
The AMD market share is still low 20% and there is absolutely no signs that to be improving.

Please, show the cards sales - how many low-end cards were sold, how many mid-range and how many high-end?

 

[AnotherReader]

If I'm not mistaken two Ampere shaders are equal to one Turing shader. They were made to be smaller and less performance.

That isn't the case. Turing introduced separate int32 execution units. Pascal executed int32 operations on the fp32 units. Ampere added 1 fp32 unit for each int32 unit. So the resources per SMX look like this:

	Pascal	Turing	Ampere
Units	64 fp32	64 fp32 + 64 int32	128 fp32 + 64 int32
Max ops per clock	64 fp32 or 64 int32	64 fp32 or 32 fp32 and 32 int32	128 fp32 or 64 fp32 and 64 int32

Keep in mind that each SMX in Pascal and Turing is limited to 64 operations per clock. However, each SMX in Ampere can execute 128 operations per clock. In practice, the difference isn't that great. Let's use the example of a game that has a mix of 2 fp32 ops to 1 int32 op. I'm using this example, because this is what Nvidia used to illustrate Turing's improvements over Pascal.

Over the course of 6 instructions, the 3 architectures will execute them like this:

Cycle	Pascal	Turing	Ampere
1	1 int32	1 fp32 + 1 int32	1 fp32 + 1 int32
2	1 fp32	1 fp32 + 1 int32	1 fp32 + 1 int32
3	1 fp32	1 fp32	2 fp32
4	1 int32	1 fp32
5	1 fp32
6	1 fp32

Thus, Ampere, despite having twice the fp32 throughput of Turing, will be limited to a 4/3 or 33% increase in performance per SMX. On top of that, shader throughput isn't the only factor affecting frame times. Still, due to Nvidia increasing fixed function resources as well (96 ROP vs 64 for instance), this corresponds pretty closely to the relative performance of the RTX 2080 and 3070 which have the same SMX count.

 

[80-watt Hamster]

I disagree. We have a duopoly and AMD didn't master anything at all. We see the same old performance from AMD and no improvement at all.
I mean look at the 5500 XT (2019) -> 6500 XT (2022) transition. 0% progress. None!

The 6500 XT is a joke for sure. That's well-established. By contrast, the 6600 is faster than the 5700 (at ~50W less power) and only a bit behind the 5700 XT, while the 66[0,5]0 XT is faster. The 5700 XT was the top AMD card of its generation, and they've got 5 more cards above it now.

It even looks that AMD intentionally stagnates the market, so that no "average joe" to ever be able to upgrade from the ancient 1080p monitor.

That's an... interesting claim. What reason would they have to do this? If anything, I'd think it'd be the opposite; driving higher resolutions so consumers feel the need to buy the more capable, more profitable cards. After all, they have products in that segment now.

The AMD market share is still low 20% and there is absolutely no signs that to be improving.

Yeah, they've struggled with this for over a decade. Nvidia's in a very entrenched position of strength.

Please, show the cards sales - how many low-end cards were sold, how many mid-range and how many high-end?

That is data I don't have, but is it relevant to this discussion?

Rewinding a bit, I didn't claim that AMD mastered anything, but that they were "champions" of value-for-money vs. Nvidia's midrange in recent history. R9 380 vs. GTX 960, RX 480/580 against GTX 1060. In both those cases, one generally got more perf for less $ (but more watts) with the AMD solution. I wasn't paying attention to the 5000 series, but right now we have the RX 6600 (yes, I'm bringing it up yet again) conclusively beating the RTX 3050 for significantly less money. The caveat there is that those aren't supposed to be competing against each other, but they are because the 3050 is way, way overpriced.

So, I understand your frustration with the lack of progress in the low-to-midrange. I am, too (more with pricing than anything), but I also don't think it's as bad (6400/6500 XT excepted) as you're making it out to be.

 

[Mikael Andersson]

$700 for a GPU should get you into 2160p territory just fine. Just don't set your settings to Ultra - that's just stupid anyhow. The next setting down is typically visually indistinguishable yet performs far, far better. Reviews tend to test at Ultra because they want to test worst-case scenarios for proper stress testing, but gaming at Ultra makes no sense.

Also, this is a bit pedantic, but please stop calling 1440p "2K". A bit of a pet peeve, but this really drives me up the wall. The only resolution called 2K is DCI 2K, a cinema resolution that's 2048x1080 pixels (essentially a slightly wider 1080p). 1440p is 1440p, and there is no "XK" numbering for it, seeing how that numbering comes from the realm of TV sales and marketing, where 1440p has never existed.

1440p is 2.5K.

 

[RJARRRPCGP]

Now that we're getting closer to the game changing announcement on September 20th


View attachment 260703

¡con las cucarachas ajajaja!

 

[Count von Schwalbe]

My hypothesis, generated on the fly during a conversation with a friend yesterday, is the confluence of high die yields with strong demand for high performance cards. Back in The Day(tm), the number of customers willing to pay top dollar for the top cards was limited. Chipmakers would resort (allegedly?) to fusing off fully-functioning dice to have chips for cards lower in the stack. Then the PC gaming resurgence came around, coupled with the crypto boom. All of a sudden there was functionally unlimited demand for the highest performing chip either manufacturer could muster, even beyond the traditional one-upsmanship. From AMD/Nvidia's perspective, that doesn't leave much incentive to drive value in the midrange. AMD was the champion there in recent years because they didn't have anything that could compete with Nvidia in the high end, so needed to lean on the value play. Then the above happened, coupled with legit advancements of their own, and that wasn't necessary anymore.

Yes, but remember, not all cards use the same die. So, if both cost and profit are based on die size, at least one mid-range card could be focused on (full navi 23 or GA106, or even smaller).

 

[80-watt Hamster]

Yes, but remember, not all cards use the same die. So, if both cost and profit are based on die size, at least one mid-range card could be focused on (full navi 23 or GA106, or even smaller).

This is true. I was remembering cases like the GTX 970 and 980. Both were GM204, but the 980 cost over two hundred more bones. Did Nvidia actually gimp fully-functional GM204 dice that could have been used for 980s because there wasn't enough demand for those, but plenty for the 970? I don't know. It wouldn't surprise me.

 

[Count von Schwalbe]

This is true. I was remembering cases like the GTX 970 and 980. Both were GM204, but the 980 cost over two hundred more bones. Did Nvidia actually gimp fully-functional GM204 dice that could have been used for 980s because there wasn't enough demand for those, but plenty for the 970? I don't know. It wouldn't surprise me.

These thoughts make me want to make tables and charts.

 

[budget_Optiplex]

As a note of interest, I've been watching this Sapphire Pulse RX6600 price as a friend needs an upgrade from an RX470 that isn't quite cutting it in newer games on his 5760x1080 Eyefinity setup. This card went to 259.99 with free shipping in early August, then was raised to an insane 279.99 + 9.99 shipping in early September, then dropped back to 259.99 with free shipping a few days ago. Looking at sold history, these are barely moving even at 259.99, with last one sold on August 24th, and finally another one sold on September 9th. This is from newegg in the USA.

SAPPHIRE (11310-01-20G) AMD Radeon™ RX 6600 Graphics Card for sale online | eBay

Find many great new & used options and get the best deals for SAPPHIRE (11310-01-20G) AMD Radeon™ RX 6600 Graphics Card at the best online prices at eBay! Free shipping for many products!

www.ebay.com

If watching this particular card is any indication: 1. The midrange cards are still too expensive, especially in these deteriorating economic conditions. 2. People really are voting with their wallets and not buying at these prices.

I keep reading articles around the internet about GPU prices being lowered because AMD and Nvidia want to move product, but I sure can't say I'm seeing any of this watching prices on midrange cards like the RX6600 and even worse the RTX 3050 which is selling for $300+ still for a card that's slower then the RX6600. I wonder when these midrange cards will get real discounts that will move them off the shelves????

 

[Appalachian]

As a note of interest, I've been watching this Sapphire Pulse RX6600 price as a friend needs an upgrade from an RX470 that isn't quite cutting it in newer games on his 5760x1080 Eyefinity setup. This card went to 259.99 with free shipping in early August, then was raised to an insane 279.99 + 9.99 shipping in early September, then dropped back to 259.99 with free shipping a few days ago. Looking at sold history, these are barely moving even at 259.99, with last one sold on August 24th, and finally another one sold on September 9th. This is from newegg in the USA.

SAPPHIRE (11310-01-20G) AMD Radeon™ RX 6600 Graphics Card for sale online | eBay

Find many great new & used options and get the best deals for SAPPHIRE (11310-01-20G) AMD Radeon™ RX 6600 Graphics Card at the best online prices at eBay! Free shipping for many products!

www.ebay.com

If watching this particular card is any indication: 1. The midrange cards are still too expensive, especially in these deteriorating economic conditions. 2. People really are voting with their wallets and not buying at these prices.

I keep reading articles around the internet about GPU prices being lowered because AMD and Nvidia want to move product, but I sure can't say I'm seeing any of this watching prices on midrange cards like the RX6600 and even worse the RTX 3050 which is selling for $300+ still for a card that's slower then the RX6600. I wonder when these midrange cards will get real discounts that will move them off the shelves????

I had read speculation that the higher tier cards are the ones getting the better discounts because those are the ones coming out first. Ive been watching 3070's, 80's, 90's and 6900 XT. There have been some pretty modest deals the last few weeks on all of them. Ive also been watching the 3060 ti and it hasn't budged so it seems this may be the case. Of course, the higher tier cards price was already so bloated, they have a lot more room to give in pricng.

 

[RJARRRPCGP]

People really are voting with their wallets

Like I voted with my wallet against the RX 6500 XT! I skipped right to RX 6600 XT early this year.

 

[ARF]

RTX 4060 (AD106, PCIe 4.0?? x8) will be ~70% faster than RTX 3060.
RTX 4070 will be ~50% faster than RTX 3070.

NVIDIA GeForce RTX 4060 Rumored To Be Slower Than RTX 3070 Ti, AD106 & AD107 GPUs Utilize PCIe x8 Interface (wccftech.com)

 

[Mikael Andersson]

This is true. I was remembering cases like the GTX 970 and 980. Both were GM204, but the 980 cost over two hundred more bones. Did Nvidia actually gimp fully-functional GM204 dice that could have been used for 980s because there wasn't enough demand for those, but plenty for the 970? I don't know. It wouldn't surprise me.

The 980 Ti sold better than the 980 i.e. the consumer isn't fooled that easily and the 5700 XT 50th Anniversary Edition sold miserably as well with a 0.04% market share at it's peak.

 

[ARF]

RX 5700 XT was the worst. Designed during a difficult period for AMD in which the company faced the real threat of going under, to bankrupt, its share stock price was $1, and AMD had no money to design a big Navi 1 chip to fill the empty high-end niche.

 

[Mikael Andersson]

RX 5700 XT was the worst. Designed during a difficult period for AMD in which the company faced the real threat of going under, to bankrupt, its share stock price was $1, and AMD had no money to design a big Navi 1 chip to fill the empty high-end niche.

Wasn't the Zen 1 CPU's making them money?

 

[ARF]

Wasn't the Zen 1 CPU's making them money?

Yes but later, RX 5700 XT was launched in late 2019, while the design stages last 4 years at least, so its design began in 2015, two years before the Zen.

 

[Mikael Andersson]

Yes but later, RX 5700 XT was launched in late 2019, while the design stages last 4 years at least, so its design began in 2015, two years before the Zen.

Ah, I see.

 

[Valantar]

People really are voting with their wallets and not buying at these prices.

That is such a wildly myopic analysis. People aren't "voting with their wallets" - six months ago, hordes of people were throwing whatever money they had at any available GPU. Calling the current reversal "voting with their wallets" is in such blatant direct conflict with recent history that it just makes no sense. People are either already using a relatively new GPU, or they have been objecting to GPU pricing for years and abstaining from upgrading. (Or, like the vast majority most likely, they just don't care that much about computers, and just want to play games.) But that just goes to show how utterly useless the concept of "voting with your wallet is" - 'cause if that's the case, then this same group has likely been doing the same for the entire GPU shortage. And not a single person has noticed or cared, because every single GPU made, no matter the price, would sell. The only exception to this is people who previously had the means to buy a new GPU, couldn't due to the shortage, and have in the meantime lost the economic security enabling them to do so. Even with current economic insecurity, that is not a large group.

What we are seeing is a combination of oversupply, a saturated market, a well stocked used market, and economic insecurity. Reducing that to "people are voting with their wallets" is incredibly reductive.

I disagree. We have a duopoly and AMD didn't master anything at all. We see the same old performance from AMD and no improvement at all.
I mean look at the 5500 XT (2019) -> 6500 XT (2022) transition. 0% progress. None!
It even looks that AMD intentionally stagnates the market, so that no "average joe" to ever be able to upgrade from the ancient 1080p monitor.
The AMD market share is still low 20% and there is absolutely no signs that to be improving.

You always have some rather interesting responses and analyses, and this is definitely one of them. First off, did anyone say AMD had "mastered" anything? Second, the RX 6500 is well established to be a very weirdly specced and wholly underwhelming GPU that makes zero sense. It definitely shows that someone at AMD has some rather bad ideas - but crucially, it is not whatsoever representative of the RDNA2 generation at all - not whatsoever. It is a clear and remarkable outlier. It does indeed show no scaling from the RX 5500 XT - but on the other hand, the RX 6600 (non-XT) outperforms the RX 5700 at 1080p; the 6600 XT outperforms the 5700 XT at 1080p and the 5700 at 1440p; and the RX 6700 XT delivers a very significant 25-32% (dependent on resolution) jump over the 5700 XT.

Also, you're very selectively only focusing on naming tiers rather than pricing - which is after all what matters more to most buyers. It doesn't matter if the GPU you can afford is in a 4, 5, 6, 7, or 8 tier - if it's what you can afford, that's the fastest GPU you're getting. Of course, in a way, you're right - there has been massive stagnation this generation. However, singling out AMD for this makes no sense. Nvidia has been identical to AMD in this regard, delivering like-for-like price and performance increases over previous generations. Blaming the current value stagnation in GPUs on AMD in that context is just nonsensical - both AMD and Nvidia are perfectly equal in this regard for this generation. And, IMO, Nvidia as a market leader has a lot more clout and a lot more of an impact with their decisions, and thus had far more of a say of where price/performance has shaken out in the end. Yes, AMD could absolutely have undercut them as they have preivously, and that is absolutely something one could argue that AMD should have done. However, that doesn't absolve Nvidia from its responsibility in consistently driving GPU prices upwards and relative value (perf/$ gen-over-gen) downwards each generation.

AMD's market share and its failure to grow is an extremely complex topic, and one that is strongly affected by Nvidia's massive mindshare advantage and their status as a decades-long incumbent market leader. Trying to gain market share against that kind of opposition is incredibly difficult, no matter your product quality, marketing, etc. It took AMD more than three years of consistently delivering better-value and in the end fundamentally superior products to Intel to erase Intel's mindshare advantage in CPUs, and that was far less entrenched than Nvidia's mindshare in GPUs.

The GPU market is a duopoly in that there are only two significant actors, but that is also a gross misrepresentation of reality, as the term implies that they are comparable in market strength. In reality, the GPU market is a quasi-monopoly with a moderately strong runner-up that somewhat threatens the status of the incumbent, but is by no means in a position to beat them - not in finances, not in multi-year R&D funding, not in mindshare.

 

[ARF]

I think there are two problems:

1. Inflation which in the graphics cards segment is anywhere between 50% and infinity.
2. Macro-economic pressures (political?) to make the lives of the rich easier, and the lives of the poor people much more difficult.

Why I am saying this.

Because the RX 6500 XT is probably not a replacement of the old Radeon RX 5500 XT, but rather a tier below it.
The true RX 5500 XT replacement is the cut down Radeon RX 6600.
Prices - MSRP at launch in the first case - $169, in the second case - $329, or 95% inflation.

Also, let's look at idealo.de and how the prices have changed in the last 10-day period between 30 August and 9 September:

Radeon RX 6400 - 171.69 euro - 172.00 euro +0.1%
Radeon RX 6500 XT - 187.90 euro - 187.90 euro +0
Radeon RX 6600 - 278.00 euro - 278.00 euro +0
Radeon RX 6600 XT - 388.96 euro - 399.00 euro +2.6%
Radeon RX 6650 XT - 403.00 euro - 401.27 euro -0.4%
Radeon RX 6700 XT - 431.10 euro - 479.90 euro +11.3%
Radeon RX 6750 XT - 540.48 euro - 547.90 euro +1.4%
Radeon RX 6800 - 629.00 euro - 629.00 euro +0%
Radeon RX 6800 XT - 769.00 euro - 789.90 euro +2.7%
Radeon RX 6900 XT - 924.92 euro - 914.92 euro -1%
Radeon RX 6950 XT - 1173.98 euro - 999.00 euro -17.5%

For a comparison:

Prices of the low-end cards as of 16 August 2022:

Radeon RX 6400 - 168.82 euro -2%
Radeon RX 6500 XT - 169.00 euro -11%

 

[Valantar]

I think there are two problems:

1. Inflation which in the graphics cards segment is anywhere between 50% and infinity.
2. Macro-economic pressures (political?) to make the lives of the rich easier, and the lives of the poor people much more difficult.

Why I am saying this.

Because the RX 6500 XT is probably not a replacement of the old Radeon RX 5500 XT, but rather a tier below it.
The true RX 5500 XT replacement is the cut down Radeon RX 6600.
Prices - MSRP at launch in the first case - $169, in the second case - $329, or 95% inflation.

Also, let's look at idealo.de and how the prices have changed in the last 10-day period between 30 August and 9 September:

Radeon RX 6400 - 171.69 euro - 172.00 euro +0.1%
Radeon RX 6500 XT - 187.90 euro - 187.90 euro +0
Radeon RX 6600 - 278.00 euro - 278.00 euro +0
Radeon RX 6600 XT - 388.96 euro - 399.00 euro +2.6%
Radeon RX 6650 XT - 403.00 euro - 401.27 euro -0.4%
Radeon RX 6700 XT - 431.10 euro - 479.90 euro +11.3%
Radeon RX 6750 XT - 540.48 euro - 547.90 euro +1.4%
Radeon RX 6800 - 629.00 euro - 629.00 euro +0%
Radeon RX 6800 XT - 769.00 euro - 789.90 euro +2.7%
Radeon RX 6900 XT - 924.92 euro - 914.92 euro -1%
Radeon RX 6950 XT - 1173.98 euro - 999.00 euro -17.5%

For a comparison:

Prices of the low-end cards as of 16 August 2022:

Radeon RX 6400 - 168.82 euro -2%
Radeon RX 6500 XT - 169.00 euro -11%

I think your second point has some merit, though calling the first point "inflation" is adopting a very problematic term (which mainly serves to hand-wave away the actual willful politics behind the economic processes causing price hikes and value drops) and making it mean something it doesn't.

Also, your generational succesion story doesn't add up to me. Given that the RX 6600 is significantly more powerful than the RX 5600 XT, it makes no sense for it to be a successor to the 5500 XT. The RX 6500 XT is undoubtedly the successor to the 5500 XT - it's just clear that whoever designed that card at AMD didn't care about dGPUs. It is very obviously an "oh, crap, I guess we have to make a desktop dGPU as well" version of a strictly designed-for-mobile, low-power, low-cost GPU. I do agree that the RX 6500 should have been called the 6400 - at which point it would have been decent for its tier! - and there should have been a ~20-24CU RX 6500 XT, not the 16CU one we got. The 6600 isn't the GPU you're missing - the one you're missing doesn't exist, as it was never made. This is a major flaw in AMD's RDNA2 die designs, as it leaves a massive performance gap between cut-down Navi 23 and full-sized Navi 24 - the 6600 is essentially twice as fast as a 6500 XT, after all.

What AMD did was clearly bet on Navi 24 being a volume seller for low-end gaming notebooks, and they optimized the design to the extreme for that use case alone. This doesn't seem to have actually led to it getting a lot of mobile design wins, so this seems like a major strategic blunder from AMD's side, as it also made their dGPU versions of the die - clearly an afterthought - look even more silly. What they should have done IMO, is make Navi 24 either 20 or 24 CUs with a 96-bit VRAM bus (and possibly a slightly larger Infinity Cache). This would have made the die slightly more expensive, and thus a harder sell for those $6-700 entry gaming laptops, but all the more flexible and attractive as an entry gaming GPU in its own right, regardless of the target market. And, of course, it would have led to a far less stupid-looking RX 6500 XT, as it wouldn't have been the least efficient GPU of its whole generation as they wouldn't have had to OC the snot out of it to make it not look terrible, etc. I just hope AMD treats the low end and lower mid-range with some more respect for the 7000 series. Hopefully the current demand slump is enough to remind them of just how many units the RX 570/580 and GTX 1060 sold, and how lucrative the ~$150-250 market is as long as you go for volume rather than massive margins.

 

[64K]

What we are seeing is a combination of oversupply, a saturated market, a well stocked used market, and economic insecurity.

That sums it up very nicely.