Tech Gaming Hardware discussion (& Hardware Sales) thread
- Thread starter Madmick
- Start date
- Joined
- Apr 18, 2007
- Messages
- 12,924
- Reaction score
- 6,108
Intel releasing a new budget GPU line where the next two generation versions are already in development:
www.pcgamer.com
Also leaks of Valve releasing a new Steam Box. Which is likely just a codename for Valve R&D'ing potential new Steam Deck designs and more importantly further developing SteamOS.
Intel just launched a brand new gaming graphics architecture but it has already designed a new generation of GPUs and it's working on the generation after that
Celestial, Druid, call them what you will, Intel still has a full gaming GPU road map. For now...
Also leaks of Valve releasing a new Steam Box. Which is likely just a codename for Valve R&D'ing potential new Steam Deck designs and more importantly further developing SteamOS.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
Difficult to be optimistic about these. They don't offer any better a value in terms of rasterization on paper than the Arc A750 which was MSRP'd at $289.Intel releasing a new budget GPU line where the next two generation versions are already in development:
Intel just launched a brand new gaming graphics architecture but it has already designed a new generation of GPUs and it's working on the generation after that
Celestial, Druid, call them what you will, Intel still has a full gaming GPU road map. For now...
If Intel was going to find a backdoor into relevance in terms of volume one is tempted to think it would be by stealing the market segment that NVIDIA has been all too happy to ignore for half a decade, now, and that's the >75W niche. Hell, they've behaved as if they're annoyed by it. These are the GPUs, of course, that gamers who repurpose office comps or other non-gaming comps may buy and slot into their computers that will run off power from the motherboard alone. This means they don't need to buy a new PSU which requires effectively rebuilding the computer (if a new PSU is even possible because often it's not with those proprietary form factor cases). It's a wonderful introduction to building, and as simple as installing RAM which many people undertake including many who aren't gamers. Simple plug-it-in job.
Icing on the cake would be to design the card specifically with two major designs, one low-profile, the other compact, so that at least one of the two variants could fit into just about any comp in existence.
There's no volume left in this segment, let alone margins to eek out. You're talking at best a low single digit percentage of gaming desktops, if even that.
Only way to grow unit TAM is by hitting the midrange segments that SI's dominate in so that Intel can hope to move up to the RTX --70 tier where margins are a little better for GPUs. That's what Intel is doing, even if their GPU group probably won't get the resources to seriously compete. If you don't see a huge wave of OEMs and SI's at CES saying they'll launch Battlemage desktops in 2024, that's pretty much an obituary on Arc's chances to break out in the next few years.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
There used to be, but I suspect you didn't know that. The GTX 1050 Ti was the most popular card on the Steam survey for multiple months, at one point, and the GTX 1050 was also up there. The GTX 750 before it was among the most popular cards, too, and so was the GTX 1650 that followed it even as it fell behind the curve in terms of performance offered relative to the current crop of GPUs as they existed upon its release. These were dominant cards with a strong volume shipped just like the midrange class you're regurgitating off your pie chart, presently.
It disappeared because the value disappeared; because NVIDIA stopped addressing it. Neither the appetite for that class, nor the underlying resale market of office computers furnishing its relevance disappeared, and that is easily observable if you look at the Amazon desktop market today. In fact, a Dell Optiplex unit is the #1 bestselling tower computer for the past 24-hours on Amazon. There are four or five of those in the Top 15 at any given time. Developing a card that could cater to those compact form factors in addition to the modest PSU wattage that lack additional power connectors doesn't strike me as an unreasonable niche to pursue when nothing else about your cards is going to entice consumers away from the alternatives.
There wasn't even a 3050. The original 3050 released, the 4GB/8GB iterations, aren't sub-75W. It was only with the release of the 3050 6GB variant, which is actually a cut-down version, not identical to the former card even outside of the VRAM difference, that they finally updated the GTX 1650 G6 from 2020. That released in February of this year.
That's how little of a shit NVIDIA gives about it. It's not because they wouldn't sell. It's because NVIDIA wants to twist the consumer's arm into spending more.
I knew that, but it's not particularly relevant given the GPU market and what gamers want has changed a lot in the past decade. As I mentioned, Steam Surveys are not statistically reliable and look very different from sell-in numbers, let alone sell-out.
Nvidia leaving the segment wasn't the result of them offering lesser value there, it was overwhelmingly because the market dried up. Since it's old enough that it doesn't really matter, here are the Circana numbers for some US budget and midrange GTX and RTX desktops from launch through last year (rounded to nearest 10K). Feel free to believe them or not since I'm a guy on a karate forum and you can't crosscheck the numbers with public data.
GTX 1650: 230K
GTX 1650 Super: 190K
GTX 1660 Super: 390K
GTX 1660 Ti: 240K
RTX 2060: 260K
RTX 2060 Super: 100K
RTX 2070: 80K
RTX 2070 Super: 120K
And for additional context, RTX 3060 (both GPU variants) vs RTX 3050 desktop volume was more than 5 to 1 in favor of 3060.
TLDR: Notice how 75W GPU sales were outstripped by midrange cards?
Amazon's top selling lists, much like Best Buy's, are not accurate. Some of those are close to correct, but they don't line up weekly or monthly sell-out numbers. There's a reason PC companies don't make decisions based off of Amazon's best seller list.
We can make this easy. What's the TAM (units) in the US for this market segment right now? If you do the napkin math you'll kind of see why it makes no business sense to go after this market.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
I see you didn't even bother to read the rest. Odd since you love proclaiming knowledge about the PC market and you're actually getting a peak into numbers that you try and replicate with a badly constructed survey where market share often doesn't even exceed the MOE.
I'll also point out your poor understanding of the English language since my original comment was that there was no volume left in that segment, which means there used to be volume there.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
Ah yes, the classic MadMick tail tuck and flee when confronted with evidence that contradicts your view. Don't forget that the next step in your delusion is to start proclaiming that you were right all along and that you are accepting concessions lol. Never change, it's practically a sherdog tradition at this point and a good laugh.
Also here's the clear flaw with your argument that you don't need actual numbers to disprove. If there's so much volume and presumably revenue in this segment, why has every major GPU manufacturer and most major board partners abandoned it over the last couple of years? I guess they hate making money or something.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
I've listened to your pretended-to-know-something-after-the-fact opinion. Your ignorant circular logic doesn't impress me. "There can't be a market for this because there isn't a market for this." Uh huh. As I said from the get-go, this was a market too happily abandoned by NVIDIA because they wanted larger margins, not because it didn't move volume. Those who watched pricing observed this on GPU resale markets from the substantial premiums users were paying to find stronger >75W cards as NVIDIA stopped feeding the demand with a more relevant GPU. And all of the fundamentals that contribute to that market existing haven't changed.
Even the Circana data you shared affirms the appetite for the 1650. It's running neck-and-neck with the second bestselling cards there behind the 1660 Super. Let me ask you a simple question. Do you understand why that's so impressive given the context of this market's niche consideration?
Once again, your reading compression is terrible. I pointed out that appetite for these kind of products dried up. You see that in the sell-out for GTX 16 and RTX 20, where 75W GPU systems were about 10% of sales. Even before that, it wasn't much more.
And yet AMD has also abandoned this segment. No one wants it because it's a dead end. You have less volume than -60 tier GPUs and you have significantly less margin.
Not to mention given how large Arc dies are still, you'd be murdering your margins even more by trying by trying to sell $200 GPU while silicon costs remain static.
Which is an insignificant data point given that a major system integrator buys more GPUs in a month than ebay sells used in a year.
The fundamental that changed is that people during Covid moved up a performance tier. Those who used to by 75W GPUs bought up a class, and it's incredibly difficult to get them to move down a performance tier after the fact, even with unfriendly pricing. There isn't sizeable hunger in the market for $200 GPUs and $500 gaming desktops, even Walmart doesn't want to do that anymore.
And Walmart is by far the biggest entry-level GPU outlet in the US.
It's not very impressive given the market at the time. It was outsold more than 4 to 1 by GTX 1660 during launch year and its strength from 2020 on was mostly Covid demand. Its pandemic bump in pricing was also worse than almost every other GPU on the market.
The -70 tier was the strongest grower during the pandemic, those people aren't hopping back down to -60, let alone 75W GPUs.
Like I said, there's a simple way to assess whether your scenario is realistic. US desktop market (this is most of GPU sales), what do you think the TAM (units)for 75W GPUs is in 2024?
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
I asked a simple question.
I answered it with no, it's not impressive compared to how other GPUs performed over the past few years.
Like I said: What do you think the TAM for a sub 75W GPU in the US is this year? You're convinced it exists to a substantial degree but can't even offer a WAG estimate.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
The reason it's impressive is because of the relatively pitiful level of performance, and its inflated price, relative to the mainstream mid-level cards upon its launch compared to past >75W. When the GTX 1050 launched it carried an MSRP of 73% the cost, despite releasing less than three years earlier, while the GTX 1050 Ti offered about 40% of the frames of the GTX 1080 which was their flagship at the time it released which made it more attractive than weaker later offerings. These were respectable gaming GPUs upon release, not aged fabrications begrudgingly bestowed on gamers too late. That very low $109 price point of the former was particularly attractive to the budget builders. I surmise the still respectably low price of the GTX 1650 is a significant reason why its sales accelerated post-2020; because GPU prices were so steeply inflated due to cryptocurrency and global supply constraints, and yet there were so many new casual gamers joining the PC gaming race trying to figure out how they could transform their computers into gaming PCs. The appetite is there.
Why would the 1650 still sell well in 2024? Why would a rare variant of the RTX 3050 (also old tech) that isn't named distinctly from the same card which wouldn't work to upgrade these sorts of comps? So nice of NVIDIA to do that. It's almost like they want to confuse the casual market to steer them away. Remind you of something? A recent launch, perhaps?
The simple truth eluding you is this: NVIDIA deliberately controls its inventory so it doesn't cannibalize its own products. All the SC corps holding the lead have always done this. Why do you think NVIDIA doesn't release the 4060, 4060 Ti, 4070 when they launch the new gen? They're constrained by how much they can produce, sure, but the reason they led with that shitshow where they tried to call two different cards the same thing-- RTX 4080-- was because they don't want the mainstream cards to cannibalize those higher-margin products. For the same reason they have eschewed catering to this budget niche within PC gaming. They throw scraps from the lesser fab lines. Sure. It makes sense for them. They're in a position of strength.
Intel is not in a position of strength. So it would seem to make sense that they might gain control of at least one portion of the market and become synonymous with being the brand of choice in that niche. Build a reputation. Gain trust and following among PC gamers.
Instead, it appears they're going to go head on at NVIDIA where NVIDIA is determined to throw its weight. Bold strategy, Cotton. Let's see how it works out for them.
GTZ 1650 sales didn't accelerate post 2020, they actually slumped during the gaming boom. And supply was healthy enough, the slow down was demand side. Again, most people bought upward during the boom period, asking them to buy a less performant (relatively) product is a really tough sell at this point.
It did not sell well in 2024, it's been EOL for quite some time. I did not include 2024 numbers.
RTX 3060 6GB also didn't sell well, albeit more due to supply constraints. None of the OEMS wanted it, so the SI's bought up all the stock rather quickly to replace 3050 desktops.
No disagreement here.
They aren't going to persuade gamers who have a RTX 3060 or RTX 4070 to buy a 75W card that costs $200 less. Nor are there gamers waiting to purchase $500 systems/$200 GPUs. It's a mature market, only way to sell GPUs is by taking share from someone else.
It's not a portion of the market worth controlling. Let's be real optimistic and assume that TAM for 75W GPUs is 20% of the US market, that gives about 3 million pre-builds and maybe another 1 million DIYs. If they win every single one of those customers, that gives Intel revenue of $400 million (optimist partner sale value of $100 per unit). That's a terrible slice of the market to chase, and global sales would maybe double it at best.
TLDR: If Intel won every single sale at a best case scenario price, they still wouldn't even cover the sunken costs from Alchemist, let alone Battlemage's development and ramp.
That isn't even to mention that they would need an entirely new die to hit 75W GPUs and Arc die sizes are still insane. They fundamentally can't offer compelling performance at a $200 price point, their architecture just isn't mature enough.
- Joined
- Jun 13, 2005
- Messages
- 66,129
- Reaction score
- 37,455
Then I would surmise it explains why GTX 1650 sales persisted post-2020 during the COVID boom while GTX 1660 sales fell off a cliff. I only inferred that supposition because of your quibbling insistences:
Launch year for the 1650, 1660, 1660 Super, and 1660 Ti were all 2019. Your first figures don't show the base 1660, but the 1660 is outnumbered nearly 2-to-1 by the 1660 Ti, and nearly 3-to-1 by the 1660 Super on the Steam Survey today. So either everyone threw their 1660's in the garbage, nobody with a 1660 answers honestly what their variant is, or you're lumping all of the above together in your launch year sales comparison. Because if the 1660 outsold the 1650 4-to-1 in 2019, and yet has sold fewer total units through today, or a roughly equal amount, then the 1650 sold much better post-2020 through the COVID pandemic and in the years after than it did.
And go ahead and look at the kind of GPUs I'm talking about. They aren't selling? That's because NVIDIA isn't making enough of them. Look up every low-profile iteration of the RTX 3050 6GB you can find on PC Part Picker. Don't believe what PCPP tells you. Visit the links. Sold out, out of stock, back order, or best of all, "Available Soon". That's true for every AIC, and every store: Amazon, B&H, Newegg, whoever. If the appetite isn't there, then why aren't the cards gathering dust on shelves?
Even if Intel's expenditures demand aiming for a larger overall share of the market, I do not see the folly in targeting the niche. Because there is nothing on paper that makes these Battlemage GPUs more attractive than the A750 or A580. At least win something. This is little more than a hail mary praying that future driver efficiency pulls off some miracle. In the age of DLSS 2.0+ maturity, it reeks of desperation, and bad strategy.
Sales persisting after Year 1 is very normal. You're reading into a trend that doesn't exist. GTX 1660 sales fell because it was almost entirely replaced by 1660 Super, whereas 1650 Super sold in tandem with 1650 for prebuilds.
I forgot to include 1660, it did 160K through last year in prebuilds, it wasn't a big mover since Nvidia didn't want to cannibalize too many 2060 sales for OEMs. The comparison for assessing 75W TAM isn't 1650 vs 1660 it's 1650 vs 1660 and 2060. And that wasn't close, regardless of how great a value the 1650 was.
So in this case, the Steam figures sort of line up, but almost all of the Steam GPU rankings are well within MOE and useless.
It's not only what Nvidia wants, it's what US partners want to sell. No one wants to go sub 75W GPUs in their prebuilds in 2025, there's no margin worth chasing there. And as I mentioned, the lower VRAM variants for 3050 and 3060 were supply limited.
The basic market analysis mistake you're making here is that this mythical horde of gamers who want 75W GPUs didn't hold off on buying new GPUs. They just bought the next tier up, they haven't' been waiting for 5 years to upgrade to a product that doesn't exist anymore. And those in the 60 tier moved to 70, it's been fairly clear in sell-out the past two years.
It would have to be an entirely new die, that's a ridiculous R&D expenditure. The main appeal of Battlemage is they exist whereas Alchemist never hit high volume production -- this isn't to say that makes a difference, the Nvidia brand recognition is very real at this point.
And again, there's a basic math problem here. Intel is hitting 4060 performance levels at 150W or 180W with a die the size of Nvidia's 70 or 80 tier products. The roadmap was set out years ago, if that's the performance we're getting for Battlemage, the performance at 75W is likely to be pretty bad and worse than the mark they need to beat there (RX 7600ish) for SI's.
