Hey, does anyone else remember when Globalfoundries announced that they were skipping 10nm and going straight to 7nm? Anyone remember how that worked out for them?
The GPU division has tried and failed to become a player in the discrete consumer GPU market.3.5 years into new management. Let's see what we got.
-Completely flubbed on demand expectations. Somehow expected the artificial increase to continue
-Cancelled Falcon Shores hybrid until who knows when? Now a GPU only variant is maybe 2025, which AMD/Nvidia has already.
-Intel 3 was supposed to offer density gains at the high performance line too. The actual Intel 3 is only 10% gain, if you chose the "HD" variant with lower performance
-18A was supposed to be 10% over 20A, and 20A was supposed to be 15% over Intel 3. Now we find out 18A is only 15% over Intel 3, with a disappointing 30% density gain. So really rather than 5N4Y, which with extra performance and density was somewhat justified, we have: 7, 4, 4+, 20A, 20A+, or 3N4Y, assuming we count 7, which makes no sense anyways but let's give them that.
-Qualcomm left, Softbank left, and now Broadcomm is on a thin line.
-The whole IDM 2.0 was about separating the two enough, not eventual goal of complete separation of the business. Big fail there. Since Pat also failed demand projections this one is also on him.
They need better IP for that.The GPU division has tried and failed to become a player in the discrete consumer GPU market.
There were two 6+8 tiles, one on N3B and 20A. What's interesting is that AFAIK there was a small number of ARL samples shipped to OEMs/AIBs using the 20A 6+8 tile, so it was a working node, but perhaps cost/yield was much better on N3B.
What do you think happens to DIY PC and x86 if Intel goes bankrupt? If 70% of the x86 market goes under, that’s basically the death of x86. The transition to ARM would be rapid at that point. I don’t really care about the x86 ISA in particular but rather the DIY PC market that’s attached to it. If x86 winds down, I’d guess the DIY PC market as it is today changes (probably for the worse).I don't share your pessimism. I think it could be significantly better. Wintel has dominated too long already, and needs to go IMO. If Nvidia shows up with an all team green gaming PC, it'll likely be as Valve did it; with their own OS.
Everyone repeats the same tropes. 1. We need company XYZ for competition. 2. Don't root for other corpos, none of them are your friends. My responses are -
1. It's late 2024, other competition is already here. May the best ones win.
2. The one in question has proven to be anticompetitive and anti-consumer to the level of being an active enemy. The other corpos don't need to be my friends. They just need to conduct biz without resorting to what amounts to bribery and intimidation, provide good end user service and support, and they are already a superior alternative. Hell, I'd settle for just being the "lesser evil".
What do you think happens to DIY PC and x86 if Intel goes bankrupt? If 70% of the x86 market goes under, that’s basically the death of x86. The transition to ARM would be rapid at that point. I don’t really care about the x86 ISA in particular but rather the DIY PC market that’s attached to it. If x86 winds down, I’d guess the DIY PC market as it is today changes (probably for the worse).
There’s other intangibles as well, who invents the next USB, PCIe, ATX, etc? The way AMD operates currently is optimized for being the 2nd largest x86 license holder. They stick back and produce less volume at higher margins, allowing Intel to take the brunt of the major investments and volume of sales for x86. If AMD continued to make x86, their role and how the company operates would have to change.
As far as lesser evils and what not, you’re kidding yourself if you think there is something morally superior about AMD or that Intel is uniquely evil. Intel operated how a near monopoly typically behaves, this isn’t unique to Intel. I don’t think AMD keeps the pricing on something like the 7800X3D at <$400 when they have monopoly power, that’d be a disservice to their shareholders if they did.
When I say 70% of x86 market, I mean 70% of x86 CPUs shipped across all segments - which is still the case (although probably not for much longer). That’s nothing to do with a “crack pipe”, that’s the current market share.Easy there, with the crack pipe. Intel is 10% of DIY:
Similarly on Amazon, where Intel has only 1 CPU of top 15 CPUs - at #14
Amazon Best Sellers: Best Computer CPU Processors
Discover the best Computer CPU Processors in Best Sellers. Find the top 100 most popular items in Amazon Computers & Accessories Best Sellers.www.amazon.com
With extra $7 to $10 in quarterly revenue, that's a lot of money to fund research into new technologies - of which Intel has done less and less as of late.
All while pocketing a large part of that $7 to $10 billion quarterly revenue mainly due to market distorting monopolistic practices.
The way Intel operated as an evil monopoly is not unique. Microsoft, Google have been behaving in equally evil ways, and NVidia is catching up.
But there is another extreme, TSMC, which is an example of a benevolent monopolist.
On a scale of -10 of most evil, where Intel is to +10 where TSMC is, my guess AMD would be in a positive territory.
When I say 70% of x86 market, I mean 70% of x86 CPUs shipped across all segments - which is still the case (although probably not for much longer). That’s nothing to do with a “crack pipe”, that’s the current market share.
DIY PC would not be the same because x86 would no longer be the dominant ISA. As of today there is no market for creating an ARM based DIY PC. Perhaps that changes but who knows what that looks like.You started with "What do you think happens to DIY PC".
Regarding the well over 70%, close to 80% market share in client PC while the underlying quality of the product doesn't justify even 50%, a better question is: what happens to the market for CPUs when the corrupting influence of Intel is removed from it.
I, for one, am curious to find out.
That shouldn't be too much of an issue, actually.There is an issue of placing the order with TSMC and get allocation. If decision to can 20A just took place, Intel may not have an allocation for sufficient volume of 6+8 dies from TSMC.
DIY PC would not be the same because x86 would no longer be the dominant ISA. As of today there is no market for creating an ARM based DIY PC. Perhaps that changes but who knows what that looks like.
As I said in an earlier post, Intel and AMD are the ying-yang of x86, they’re intertwined and co dependent. Neither Intel or AMD alone would have the market power to keep x86 as the dominant ISA. I get it, you and most of this forum are still angry because of something Intel did 20 years ago abusing their market position. That ugly history doesn’t change the fact of what I’m saying though.
If Intel does go down and if this forum still exists, I hope I’m still here to bump this post after the fallout takes place.
Edit: Also, what happens to the gaming dGPU market? That entire segment relies on the DIY x86 duopoly. I’m not saying this market disappears but it definitely changes in some way or another.
That shouldn't be too much of an issue, actually.
Going by rumors, the 20A tile was only going to be used in some desktop models, and not at all in mobile, indicating volume of this tile wasn't going to be that high in the first place.
And in desktop, they can mitigate volume constraints by controlling DIY demand for the 6+8 TSMC models through price (keeping price and margin a tad higher than they otherwise might have) and fill any remaining volume gap by continuing to make 65W 6+8 Raptors. I mean, in the price region of 6+8 models, their competition from AMD is mostly 6C and 8C models without V-Cache.
That being said, I wonder if demand for the 6+8 models is really going to be so high that their booked TSMC capacity would be insufficient. I have my doubts, even if ARL turns out relatively good.
There is an issue of placing the order with TSMC and get allocation. If decision to can 20A just took place, Intel may not have an allocation for sufficient volume of 6+8 dies from TSMC.
It's not a mental block, I was replying to a comment that said this: "Wintel has dominated too long already, and needs to go IMO."It seems like you have some mental block preventing you from conceiving a reality in which Intel is a not a dominant player. As if the possibility of Intel losing hegemony over client PC = end of the world, end of DIY, end of x86.
As if TSMC could not just allocate more wafers from Intel CPUs to AMD CPUs, and everything else stays the same.
It seems that you can conceive of a possibility of TSMC shifting more wafers to Arm, buy you can't conceive the possibility of TSMC shifting more wafers to AMD.
Karma
Desktop 6+8 would amount to relatively meagre volumes.Maybe Intel committed to buy x number of wafers from TSMC, but losing market share to AMD means that they need to cut back on their internal sourcing to be able to hit the order commitment to TSMC. In that case, killing 20A which was dual sourced made perfect sense since they already had the designs done for TSMC and they could save money to boot.
Intel's stock price is cheap. Anyone can try to snap up a few shares if they think it'll give them leverage. Buy enough shares and you can force a spinoff of the fabs.WOW. I would be shocked if this come true. Qualcomm x86 on the way?
They'd buy the teams, they wouldn't get an x86 license any day.WOW. I would be shocked if this come true. Qualcomm x86 on the way?
Exclusive-Qualcomm has explored acquiring pieces of Intel chip design business, sources say
SAN FRANCISCO/NEW YORK (Reuters) -Qualcomm has explored the possibility of acquiring portions of Intel's design business to boost the company's product portfolio, according to two sources familiar with the matter. The mobile chipmaker has examined acquiring different pieces of Intel, which is...finance.yahoo.com
I think AMD would approve of an x86 license transfer to Qualcomm. Competing against one x86 CPU design company is easier than competing against many ARM CPU design companies.They'd buy the teams, they wouldn't get an x86 license any day.
They just bought the Nuvia team, why would they buy yet another CPU design team just to make ARM CPUs (and start over from scratch)? Without an x86 license and their existing designs the team is significantly less valuable. If Intel is serious about selling them off to raise cash they should include the license.They'd buy the teams, they wouldn't get an x86 license any day.
Yep- stuff like that, I'd guess.I thought it would be the team which co-designs laptop boards with OEMs?
If Intel is serious about selling them off to raise cash they should include the license.