RTX Sales Reported as Disappointing

[linkgoron]

I mean--this could have been a game-ending throat punch to AMD if nVidia had simply given their customers exactly what they expected, with a rosier look into the future of development and the honest representation of that tech.

Nvidia could have throat punched AMD a long time ago if they wanted to. At this point, it's clear that they don't really want to. Nvidia could've made the Vega 64 cards match their 2050/40 tier if they wanted to. Assuming 2080ti->2080, 2080->2060 and 2060->2050/40. They could've made the Polaris cards match the 1050 (assuming 1080ti->1080, 1080->1060, 1060->1050). They've decided to go for margins instead. I don't think they really care about AMD at this point.

I just think that they overreached with Turing. Nobody really has a reason to upgrade, unless they're willing to pay significantly more than their last purchase or they have a very old (4+ years) card.

 

[DrMrLordX]

Dat Supply and Demand! There is a clear thirst for AMD products!

I know, right?

On a slightly more-serious note, maybe eventually we'll see the 2080Ti come down in price, especially if sales continue to stay stagnant and Intel shows up earlier than later with their product.

 

[Guru]

Nvidia could have throat punched AMD a long time ago if they wanted to. At this point, it's clear that they don't really want to. Nvidia could've made the Vega 64 cards match their 2050/40 tier if they wanted to. Assuming 2080ti->2080, 2080->2060 and 2060->2050/40. They could've made the Polaris cards match the 1050 (assuming 1080ti->1080, 1080->1060, 1060->1050). They've decided to go for margins instead. I don't think they really care about AMD at this point.

I just think that they overreached with Turing. Nobody really has a reason to upgrade, unless they're willing to pay significantly more than their last purchase or they have a very old (4+ years) card.

LOL. Nice fantasy land you live in buddy. Facts are they miscalculated the need/want/penetration of RTX features, especially at such a high cost and performance "premium".

They probably put these "features" in the first place because they couldn't increase performance enough, so they essentially put psyx 2.0 in there.

 

[linkgoron]

LOL. Nice fantasy land you live in buddy. Facts are they miscalculated the need/want/penetration of RTX features, especially at such a high cost and performance "premium".

They probably put these "features" in the first place because they couldn't increase performance enough, so they essentially put psyx 2.0 in there.

You can look at my posting history, I'm not exactly a huge Nvidia fan. However, the only reason people hate the 20 series is because of pricing. If the 2080ti was released @ $700 - it would've been received extremely well. It was released at $500 more, though. Nvidia have decided to let price/perf stagnate and increased prices across the board. I assume they expected people to actually pay more in order to participate in beta testing for DLSS and RTX. Given Nvidia's margins, I believe they could have crushed AMD. AMD's cards have been extremely complex (HBM/HBM2 , interposer etc.) large and inefficient compared to Nvidia's.

The FuryX was just competitive with the stock 980 ti (recall that AIBs has 20%-25% better cards), and it needed a complex architecture with an interposer and HBM. The FuryX was still less efficient than the 980 ti. Thanks to HBM, it also only had 4 GB of memory.

Vega 64 (a 486mm^2 card with HBM2) is barely competitive with a stock 1080 (a 314mm^2 card with GDDR5x) while, again, being less efficient, costing a lot more and having exotic memory and using an interposer. It was also released more than a year later. Vega also has a few features that were supposed to work but never panned out. the smaller 1080 ti was way faster.

The Radeon 7, a 7nm 330mm^2 with 16GB HBM2, is just competitive with the stock 1080ti (471mm^2 14nm). It's also, again, less efficient by a wide margin. It was released almost two years after the 1080ti. The 2080's perf/watt is 37% better than the Radeon 7, according to TPU.

For the last four years, AMD's "high-end" designs have been barely competitive with Nvidia, and essentially never really out-performed them while always being more expensive to build, having very expensive memory, and consistently having worse perf/watt than Nvidia's card at the same price point.

AMD have also been unable to release a top to bottom architecture in years. We're still seeing the third Polaris re-release for the "low end"/midrange. Nvidia have released a top to bottom Maxwell series, Pascal series and are close to releasing their "low end"/midrange Turing cards.

I'm planning on replacing my aging system with an all AMD system later this year (assuming Navi is not a turd), but IMO the graphics landscape hasn't been extremely competitive in the last few years. AMD have managed to right the ship in the CPU side (also thanks to Intel's 14nm and 10nm troubles), let's hope they manage to right the GPU ship.

 

[Arkaign]

Good stuff. Agreed @zinfamous my 2006 brother

And @linkgoron yes a good summary. The original/FE models of the 9 and 10 series really shouldn't be even compared in today's market. I always thought it was extremely disingenuous to compare Vega 56 and 64 exclusively to FE models that were already obsolete by the time Vega dropped. Ditto with comparisons of FE 10 series to RTX and Vega 7. Just very misleading considering how massively better some of those AIBs turned out to be. Even just settling on a fairly pedestrian EVGA FTW or Strix would have been perfect. It was especially off considering the 11Gbps upgrade and AIB 'refresh' 10 series (again, launched BEFORE Vega) were so dramatically improved over launch FEs.

I mean I have no problem with a review simply having the old FEs present, buy never as the ONLY point of comparison, when they probably make up 5-10% of the 10 series entire production history. Thus, they don't represent an accurate reference for the majority of potential buyers.

 

[DrMrLordX]

The Radeon 7, a 7nm 330mm^2 with 16GB HBM2, is just competitive with the stock 1080ti (471mm^2 14nm).

It's also competitive with a 2080. If that tells you anything.

AMD also didn't do themselves any favors by overvolting the damn things. Now people think that they're noisy, inefficient cards. Nearly every one will run at 1V or lower without dropping any performance. They sip power and run faster compared to overclocked Vega64. You won't see most of that in reviews. Again, AMD's fault there. They'll never knock the 2080 off its pedestal shipping products with troublesome default configurations like that. Users shouldn't have to use Wattman to tame the product on day 1.

So if RTX sales continue to flag, it won't be on account of AMD. I don't think.

 

[pauldun170]

 

[IntelUser2000]

AMD also didn't do themselves any favors by overvolting the damn things.

Maybe its not they knew about it and purposely overvolting it. Being able to characterize the semi device and have it work reliably over 10 million devices require time, resources, and planning.

Also, remember we are speculating what AMD could have done based on very limited personal experiences. Whether it be from reading reviews of the card, or playing around with it.

What you can do when you are mass producing a product is severely limited when compared to individuals that buy them. That's why headroom always exists.

 

[Mopetar]

Maybe its not they knew about it and purposely overvolting it. Being able to characterize the semi device and have it work reliably over 10 million devices require time, resources, and planning.

Also, remember we are speculating what AMD could have done based on very limited personal experiences. Whether it be from reading reviews of the card, or playing around with it.

What you can do when you are mass producing a product is severely limited when compared to individuals that buy them. That's why headroom always exists.

We've known that AMD has been on a very limited budget in the lead up to Ryzen. It's quite likely that they set voltage levels at a place where almost all of their silicon would be able to hit a particular clock speed. They didn't have the resources to bin more aggressively or create specific parts that would remove the worst 15% of performers and allow for more sane voltages in the remaining chips.

The other problem is that GCN has pretty much hit a wall. They need a new architecture that's been designed from the ground up. Supposedly we'll get that after Navi based on old roadmaps. Hopefully now that AMD has competitive CPU designs, they'll be able to afford more R&D for their graphics division.

 

[tviceman]

We've known that AMD has been on a very limited budget in the lead up to Ryzen. It's quite likely that they set voltage levels at a place where almost all of their silicon would be able to hit a particular clock speed. They didn't have the resources to bin more aggressively or create specific parts that would remove the worst 15% of performers and allow for more sane voltages in the remaining chips.

The other problem is that GCN has pretty much hit a wall. They need a new architecture that's been designed from the ground up. Supposedly we'll get that after Navi based on old roadmaps. Hopefully now that AMD has competitive CPU designs, they'll be able to afford more R&D for their graphics division.

This excuse being spewed endlessly about AMD overvolting all their chips to meet bins is old and tired. If they could get any appreciable yields with lower set voltages, they'd do it because, as it stands now, they've basically forfeited the entire discrete laptop market. You're saying that AMD gave up an entire market of millions of discrete GPU's because they don't have any way to bin their GPU's like they do their CPU's.

"Hey guys, we don't have the $150,000 it would take to bin all of our chips so therefore we are just going to forfeit $100,000,00 in laptop GPU sales this year. Just set voltages to max and oh well."

Ridiculous.

 

[DrMrLordX]

That's why headroom always exists.

That's a hell of a lot of headroom. Mine can overclock by 140 MHz and still run at below stock voltage. It's nuts. Or I can drop voltage by .13V and run it with stock settings. You don't generally hear about that with NV cards.

At least with RTX, what you see is more-or-less what you get. You can pay a bit extra for better cooling and/or a custom board with better power delivery if you fancy overclocking the thing. But for the typical nVidia user, they can just drop the card into the system and get the performance the reviews told them they'd get without tuning anything. Those reviews sell a lot of cards for NV. They win all those benchmarks, and they're still winning.

"Hey guys, we don't have the $150,000 it would take to bin all of our chips so therefore we are just going to forfeit $100,000,00 in laptop GPU sales this year. Just set voltages to max and oh well."

Ridiculous.

We're talking about a card that is a competitor to RTX cards, specifically Radeon VII. Nobody mentioned their mobile offerings. One would hope those were better tested and binned.

 

[Leadbox]

This excuse being spewed endlessly about AMD overvolting all their chips to meet bins is old and tired. If they could get any appreciable yields with lower set voltages, they'd do it because, as it stands now, they've basically forfeited the entire discrete laptop market. You're saying that AMD gave up an entire market of millions of discrete GPU's because they don't have any way to bin their GPU's like they do their CPU's.

"Hey guys, we don't have the $150,000 it would take to bin all of our chips so therefore we are just going to forfeit $100,000,00 in laptop GPU sales this year. Just set voltages to max and oh well."

Ridiculous.

If you could just magically get into laptops because you binned all your chips, then sure it's ridiculous. But we both know it's not that simple.

 

[arandomguy]

A more attractive 2xxx release from a raw perf/$ standpoint is not without it's own issues. For one they'd have to do a massive write off of existing 1xxx over stock due to the mining collapse. The bubble collapsed and the timing is what it is for Nvidia. The main reason for this release is almost certainly that they wanted to avoid any type of write off of 1xxx cards. Even when asked they refused to consider direct rebates down the chain to lower the price.

If RTX features were purely limited to the Titan than what would you expect the developer adoption to be? If it's already considered low now than you're going to be looking at a complete 0 adoption rate.

 

[DrMrLordX]

For one they'd have to do a massive write off of existing 1xxx over stock due to the mining collapse.

nVidia already wrote off over $100 million in inventory in Q4 2018.

 

[arandomguy]

nVidia already wrote off over $100 million in inventory in Q4 2018.

Then you'd be looking at an even larger write off.

We know 1xxx series never ended up heavily discounted towards EOL. Even Q4 sales were quite meek considering how long they've been on the market. They were also better in the US compared to other regions, the "discounts" were even worse there.

If you had 2xxx series that offered typical ~50% perf/$ how would those have ever been cleared without corresponding price slashing or just completely written off?

If we want to look back to the first mining bubble that only affected AMD. AMD's R9 290x ended up crashing into the low $2xx price range down from $500. It looks like both sides do not want that type of situation this time around.

 

[coercitiv]

A more attractive 2xxx release from a raw perf/$ standpoint is not without it's own issues.

If you had 2xxx series that offered typical ~50% perf/$ how would those have ever been cleared without corresponding price slashing or just completely written off?

When you have huge unsold inventory you're already screwed, you're gonna have to eat the loss no matter what, no magic adjustment to the new gen cards will mitigate that. Whether you choose to slash prices or simply make the chips disappear in the desert to preserve product market value is another matter entirely.

 

[beginner99]

They probably put these "features" in the first place because they couldn't increase performance enough, so they essentially put psyx 2.0 in there.

They put it in there because they knew they have no competition and can take the hit. Neglecting traditional performance while competition is strong would be a huge issue. But NV could add the feature and still keep the performance crown while on a worse process. shows you how far ahead they are. It's the ideal timing to add these features.

Yet I'm still of the opinion it's the worst launch in a very long time. There is no to little incentive to upgrade even from an old GPU. We now in 1 years time we will have Navi and possibly also intel dGPU for more competition.

 

[arandomguy]

When you have huge unsold inventory you're already screwed, you're gonna have to eat the loss no matter what, no magic adjustment to the new gen cards will mitigate that. Whether you choose to slash prices or simply make the chips disappear in the desert to preserve product market value is another matter entirely.

Existing 1xxx inventory judging by retail prices ended up being cleared with minimal discount. Nvida was even asked by analysts whether they would push rebates to help 1xxx down the channel to clear them faster and they declined that option. This is the route the they choose. On the consumer side the obvious preference would be a repeat of the AMD 2xx series price collapse the last time around but that seems to be the route both sides want to avoid this time around.

I suspect other timing issues ended up at play here as well for this release. GDDR6 pricing and availability likely ended up on the worse side of the spectrum. There is also the timing issue of effectively being put at the tail end of a node yet too soon for the next one now.

I also want to circle back to the perf/$ discussion and that driving sales. How much of a stimulus would that really be for sales? How many times have you heard of the argument that existing cards are already good enough? We even have people still saying Radeon 7970s are good enough.

 

[coercitiv]

Existing 1xxx inventory judging by retail prices ended up being cleared with minimal discount.

Minimal discount yes, but not with mitigated loss. The dichotomy should be obvious here: they delayed new gen launch to allow for old inventory clear, then they blamed poor financial outcome on delayed 2060 launch.

I also want to circle back to the perf/$ discussion and that driving sales. How much of a stimulus would that really be for sales? How many times have you heard of the argument that existing cards are already good enough? We even have people still saying Radeon 7970s are good enough.

Are we going by what people say or by what people do? Historically products with better perf/dollar sell well. How do products with stagnating perf/dollar do?

The day we start saying games look good enough is the day I start saying chips are fast enough.

 

[Hitman928]

Nvidia inventories are continuing to grow. I would have to imagine that it is by and large gaming GPUs since that is the highest volume market and obviously the market that crashed the hardest after the crypto bubble pop.

(Note: The year in the table is FY which Nvidia labels a year ahead of calendar year)

https://seekingalpha.com/article/4241860-nvidia-incongruous-guidance

Nvidia CEO recently said that 10XX series cards were almost cleared out but I don't see how that aligns with the inventory build and the write off they reported in the last quarterly earnings. Maybe they'll have another write off to report next earnings and that's built into his comments. Either that or they have a huge stockpile of 20xx and 16xx cards waiting to be sold.

 

[Rifter]

Anyone who didnt see this coming is as blind as ray charles

 

[IntelUser2000]

That's a hell of a lot of headroom. Mine can overclock by 140 MHz and still run at below stock voltage. It's nuts. Or I can drop voltage by .13V and run it with stock settings. You don't generally hear about that with NV cards.

Why are you surprised? We had bigger headroom just a few years ago. Now we practically expect stock chips to ship at the levels just an inch away from DIY overclocked setups.

The things that seem simple and obvious are very much the opposite, because these guys are working at the limits of semiconductor technology.

The big company with more dollar and human resources can indeed do certain things better. Even when they are stagnating because the managers are full of arrogance due to years of unencumbered success. Things like Day 1 launch drivers, or getting it work on millions of systems without crashes or reboots are all hidden work most people don't appreciate and take for granted.

 

[Mopetar]

This excuse being spewed endlessly about AMD overvolting all their chips to meet bins is old and tired. If they could get any appreciable yields with lower set voltages, they'd do it because, as it stands now, they've basically forfeited the entire discrete laptop market. You're saying that AMD gave up an entire market of millions of discrete GPU's because they don't have any way to bin their GPU's like they do their CPU's.

Then why can almost anyone who wants to buy an AMD chip and undervolt it to get much better performance? What reason do you have to offer for AMD's behavior that provides a better explanation?

Also, they didn't give up the discrete laptop market, they're just not as big of a player, in part because they have APUs to sell that mean that the get to sell both a CPU and GPU as opposed to only one, but also because their architecture just isn't as efficient as NVidia's (even once you do undervolt the AMD GPU) so you're almost always better off going with an NVidia GPU do to power constraints that aren't as important in a desktop. AMD does provide discrete chips for Apple's laptops, but that's almost a completely separate market.

 

[DrMrLordX]

Why are you surprised?

I'm not surprised, actually. I anticipated it, provided the launch drivers allowed proper adjustments for me to test my expectations (they didn't, but the immediate driver update after launch did). NV hasn't had products like that in a long time, if ever. It's one of the reasons why AMD hasn't been competitive in the consumer market in awhile.

 

[LTC8K6]

Why wouldn't AMD bin and name the chips themselves? Vega VII / Vega VII XT or whatever for the better performing chips.