Intel has started benchmarking their high end GPU's

PLEASE let these be good...and have actual stock.
That would be brilliant in the current climate, but I presume by the time they release, AMD and Nvidia might have their supply sorted
Without a doubt. Unless Intel manages some in house miracles the problem in general right now for everyone seems to be the lack of 7nm and beyond fabs. As much as TMSC can handle, everyone from phones to cars and space rockets want new chips..... and there just isn't enough production.

A few months ago I would have said the long term solution is chiplets to drive up yields and allow controllers to be fabbed on less then processes... but even AMD is having issues with Ryzen supply right now. It seems like TMSC is just slammed.

As much as I really expect Intels fab business to eventually fail hard and get spun off.... we really could use Intel fabs turning out GPUs.
 
This won't change anything, even if it could beat Nvidia and AMD GPUs if they don't offer XE cards as stand alone, retail cards.
If these are sold as add-in discrete cards from the like of Asus, MSI ect , it would change everything.
If.
 
As much as TMSC can handle, everyone from phones to cars and space rockets want new chips..... and there just isn't enough production.

I was looking up the automotive chips - either NXP Semiconductors of Infineon. I thought it was especially interesting that much of their TriCore products are still 32-bit chips based off of 90nm and 65nm process.
Now I realize it really depends what chip and for what function certainly is a big factor, but ARM chips from NXP could be 10nm or as old as the 28nm stuff. Do we actually have a specific list of what is currently in shortage?

I know Toyota has their 4-month stockpile - hmm, maybe I should look into what they're using.
 
This won't change anything, even if it could beat Nvidia and AMD GPUs if they don't offer XE cards as stand alone, retail cards.
If these are sold as add-in discrete cards from the like of Asus, MSI ect , it would change everything.
If.

shoot, they don’t even need to partner with other companies if they don’t want to. As long as they are available at retail to buy, if they’re all Intel-made and Intel-sold, it’s still competition.

Xe-HPG is, in theory, going to be available to buy just like any GeForce or Radeon card.

(I hope it happens, but it also won’t believe it’s actually happening until it actually does.)
 
If they can release anynything that is better than 'terrible' it will be a win for intel.
Well so far we at least have better than an AMD iGPU... which for casual 1080p is a good start.
 
I doubt the first gen is going to be mind blowing, but I'm happy to see Intel taking graphics more seriously.

Definitely room for a third player here, and if they have stock well it will sell out.
 
So this articles title is clickbaity as all hell but it shows AMD and NVidia starting to show up in the benchmark tests for the DX12U mesh shader tests that Intel has been.

TLDR;
3080
Traditional: 67 fps
Mesh: 584 fps

6800xt
Traditional: 28 fps
Mesh: 468 fps

neither card showed any visible difference between the two methods in side by side still frame comparisons.

These new render methods should start appearing in titles very soon, supported on consoles and the RTX2000 series.

https://www.pcgamer.com/nvidias-rtx...sPOF3BYYxoohJRgiFjPMusXVJeRpT2n2uV5tK_GJKnJi4
 
Last edited:
may be time for a new AMD owned fab?? that way they can not only maximize market share but maximize profit. heck they use to own global foundries back in the day.

GET ON IT AMD!!!
 
may be time for a new AMD owned fab?? that way they can not only maximize market share but maximize profit. heck they use to own global foundries back in the day.

GET ON IT AMD!!!
Fabs were one of the things that was causing problems for AMD back in the day, so don’t count on it.
 
Well TSMC is expanding, and things will be back to normal eventually, so I don't think AMD needs to do anything drastic.
 
may be time for a new AMD owned fab?? that way they can not only maximize market share but maximize profit. heck they use to own global foundries back in the day.

GET ON IT AMD!!!
I feel it would be extremely hard to compete with the TSMC of the world, specially for an entity that is really far from 100% focused on it.
 
Well TSMC is expanding, and things will be back to normal eventually, so I don't think AMD needs to do anything drastic.
Well TSMC is expanding their 5 and 3nm offerings they aren't building any more 7nm fabs. 5 and 7 are running at capacity it is their older larger nodes that are running at lower volumes, so nothing TSMC has planned in the immediate future will improve AMD's product availability. Once AMD has met Sony and Microsoft's delivery quotas sometime in July (at current estimates) then they will be able to slow production of the console parts which is currently making up 80% of their output. AMD simply released too many new products in too short a time on the most heavily constrained production process in the world and has no real means of meeting demand on it.
 
I feel it would be extremely hard to compete with the TSMC of the world, specially for an entity that is really far from 100% focused on it.
Even if they did manage to get TSMC to completely license them their production process the cost of a single fab plant costs TSMC upwards of 20 Billion and takes 3 to 5 years to build. AMD has signed an agreement for ~30 of TSMC's 7nm output and given that TSMC operates 4 fabs on that size would mean AMD would have to build 2 of them to beat that output, and AMD has neither the cash on hand nor the time to invest into that.
 
Well so far we at least have better than an AMD iGPU... which for casual 1080p is a good start.
Considering the AMD iGPUs are using tech almost a decade old at this point it's not a good sign that intel can only just barely beat it... I don't know that it's realistic to see RDNA 2 architecture very soon, but even just RDNA is a massive step up in efficiency and performance.
 
And where are the bench results for the Intel GPU??


Bottom of the chart?

What's that saying about wishing in one hand and shitting in the other?
 
Never gonna happen. Way too huge and risky of an investment.
how? they know what they're doing and how it's done..?! it's supply and demand. there is high demand, how risky could it be. it's not like the "chip market" is gonna see a slow down anytime soon.

Fabs were one of the things that was causing problems for AMD back in the day, so don’t count on it.

nah i think their aquisition of ATI along with their performance of Phenom falling behind core2duo and then shortly after the core i7 series along w/ Nvidia punishing them is what put them in the position to sell their fab.

man if they could start fabbing their own chips again their profits would soar and would give them more R&D money to keep it going and keep competing w/ intel. If it takes a few yrs to get going, SO WHAT?

You guys are literally living IN the 2nd chip drought of the decade but saying we don't need another fab? ok?
 
how? they know what they're doing and how it's done..?! it's supply and demand. there is high demand, how risky could it be. it's not like the "chip market" is gonna see a slow down anytime soon.



nah i think their aquisition of ATI along with their performance of Phenom falling behind core2duo and then shortly after the core i7 series along w/ Nvidia punishing them is what put them in the position to sell their fab.

man if they could start fabbing their own chips again their profits would soar and would give them more R&D money to keep it going and keep competing w/ intel. If it takes a few yrs to get going, SO WHAT?

You guys are literally living IN the 2nd chip drought of the decade but saying we don't need another fab? ok?
Profits would tank, a new Fab would cost a bare minimum of 20 Billion to build, that is if they managed to work out a licensing deal for free with TSMC. AMD does not have that degree of capital so they would have to take out some very heavy loans then spend some 3-5 years building the plant all while paying the interest on those loans and paying TSMC to build their chips in the meantime. Then once it was done it would likely then get a single year of operation before needing to be re-tooled for a newer node. Then that would leave them with a single fab they could dedicate to their chips, where TSMC currently gives them roughly 1.5 fabs worth of output so they would either need to build 2 fabs or still outsource their production. At this stage, there is no reasonable way for AMD to get back into fabbing their own chips with any degree of a reasonable ROI.
 
This is great. If it's good, Intel will make a killing since they have their own fabs. Amd. LOL
 
Is it shipping in volume yet? You can't pick one up at Best Buy.
2022 for consumer DG2, the 10nm is filled for the new Xeons and laptop line for July this year, they are really working to nail Q1 procurement season for the new lineup.
 
how? they know what they're doing and how it's done..?! it's supply and demand. there is high demand, how risky could it be. it's not like the "chip market" is gonna see a slow down anytime soon.
The risk is ending spending a fortune of a fabs that end up not competitive for the high margin cpu-gpu product they make and still end up using TSMC anyway for those seem significant, specially with how much value they can create in spending money and effort on what they are good at right now. With the list of TSMC project current in talk.... good luck for anyone to have something that keep up with those with the level of expertise that are getting and it is a role of dice that it would help in 2030 to have an impact for market share (that stock shortage because of those fabs bottleneck will be will still be an issue).
 
FTA it sounds like they still see iris as the GPU to beat which means this will likely be almost as big of a joke as the XE in a tech sense and a bigger joke if they actually try to market it as a gaming part. Teasers from Raja based on unreleased benchmarks only reinforce this suspicion.

nah i think their aquisition of ATI along with their performance of Phenom falling behind core2duo and then shortly after the core i7 series along w/ Nvidia punishing them is what put them in the position to sell their fab.
It was multiple missteps on new nodes with their fabs that mainly put them in a perilous situation, the ATI acquisition along with selling the fabs were big gambles but ultimately kept the company going. Just like Intel is dealing with now fab process missteps not only created issues there but made their CPUs even less competitive.

The only way I see them opening a fab or two of their own is they get a favorable licensing deal with TSMC to use their tech but I think they'd rather contract it out considering how much flexibility it gives them, not to mention they can see what they could be stuck with if they were still relying on GF.

New fabs will be coming but they take time to build so it's not something that's very market responsive. I also think we'll see further consolidation of the fab industry due to the risk involved and it wouldn't surprise me to see Intel get out of it if they can't turn things around soon with their 7/5nm processes.
 
Building a fab is like allowing the tail to wag the dog.

It is better for AMD to outsource from TSMC & Samsung rather than commit huge volumes to a fab partner (like what happened with bulldozer)
 
Then once it was done it would likely then get a single year of operation before needing to be re-tooled for a newer node.
the way it's looking, PS5/XBSX demand is gonna be high for a while being that nobody that wants one can get one w/ out paying at least double. not to mention everyone that wants ryzen 5xxx and RX6xxx. as far as i know AMD should be almost to the point of NO debt. that's what they've been using most of their Ryzen profits to do, if i remember correctly, is pay down their debt. In the long run, the way the world is headed, a new fab would pay most certainly pay itself off. but hey, that's just my opinion. what do i know?
 
the way it's looking, PS5/XBSX demand is gonna be high for a while being that nobody that wants one can get one w/ out paying at least double. not to mention everyone that wants ryzen 5xxx and RX6xxx. as far as i know AMD should be almost to the point of NO debt. that's what they've been using most of their Ryzen profits to do, if i remember correctly, is pay down their debt. In the long run, the way the world is headed, a new fab would pay most certainly pay itself off. but hey, that's just my opinion. what do i know?
It could but not with an ROI that would make investors happy, they couldn’t pay for it in 10 years or less. 2020 was their most profitable year ever and if they were to put 100% of their profits into paying for the fab it would take 13 years. But that again assumes that TSMC gives them the tech and know how on how to make the chips. If TSMC didn’t give them the tech and they had to research it themselves it very well could make Intel’s 10nm debacle look tame in comparison, AMD just doesn’t have the kind of revenue to support that. Given AMD’s roadmaps they intend to be on TSMC 3nm in 4 years so if they broke ground today for that fab plant they would need to have already locked in a process as good as or better than that.
We’re just going to have to accept that AMD is a boutique part dealer, small batches of high end products and tight availability.
 
Except Intel isn't fabricating it, TSMC is producing it because Intel still hasn't managed to ship anything below 14nm in volume.
I laughed than thought, “that can’t be right!” But looks like Intel has been shipping Ice Lake on 10nm since 2019, and Tiger lake in 2020. Intel plans to have their 4’th 10nm Fab operational and in full production by April. Granted that is for the laptop ranges, but they are holding their own against AMD in those departments while costing about the same and having drastically better availability.
 
Last edited:
how? they know what they're doing and how it's done..?! it's supply and demand. there is high demand, how risky could it be. it's not like the "chip market" is gonna see a slow down anytime soon.



nah i think their aquisition of ATI along with their performance of Phenom falling behind core2duo and then shortly after the core i7 series along w/ Nvidia punishing them is what put them in the position to sell their fab.

man if they could start fabbing their own chips again their profits would soar and would give them more R&D money to keep it going and keep competing w/ intel. If it takes a few yrs to get going, SO WHAT?

You guys are literally living IN the 2nd chip drought of the decade but saying we don't need another fab? ok?

You can think what you want, but that's not the reality of what happened to AMD. Ars did a great article about this back in 2013. This is part two of that article, but it details the situation around how the fabs were simply sucking way too much capital out of the company:

https://arstechnica.com/information...e-top-of-the-mountain-to-the-deepest-valleys/

I don't think you appreciate just how expensive fabs are to build and operate. Then, after spending a few billion dollars to get it up and running, you have to go head to head with Taiwan Semiconductor and their army of patents guarding the best process technology. They're ahead of everyone, including companies with deep pockets like Samsung and veteran fab operator Intel, who has struggled to break past 14nm, as well as GF, AMD's old fab operation who, at this point, would now have years of experience beyond what AMD has right now. AMD is going to get creamed if they try to enter that market at this point. They would need to invest billions of dollars just to be able to compete, and even then, there is no guarantee that they can.

I never said the world doesn't need another fab, I said the world doesn't need one from AMD when there are better alternatives, and AMD doesn't need a fab to bring the company to its knees once again. In spite if what you want to believe, no, this would not be a wildly profitable endeavor from AMD that will give it an R&D boost to crush Intel.
 
the way it's looking, PS5/XBSX demand is gonna be high for a while being that nobody that wants one can get one w/ out paying at least double. not to mention everyone that wants ryzen 5xxx and RX6xxx. as far as i know AMD should be almost to the point of NO debt. that's what they've been using most of their Ryzen profits to do, if i remember correctly, is pay down their debt. In the long run, the way the world is headed, a new fab would pay most certainly pay itself off. but hey, that's just my opinion. what do i know?

Not to sound rude so please don't take offense to this, but it's very clear that you don't understand the costs and complexity associated with chip fabrication. If fabrication were an easy business, I want you to ponder for a moment why even Intel had considered walking away from it late last year.

https://www.bloombergquint.com/business/intel-considers-what-was-once-heresy-not-manufacturing-chips

Again, what you're proposing isn't realistic from a business perspective. AMD is not the only chip manufacturer who doesn't fabricate their own chips. Nvidia, Qualcomm, and Broadcom, to name a few, outsource their fabrication as well. There's a really good reason for this. A new fab would not almost certainly pay for itself. They would need to take on a tremendous amount of debt, pay interest on that debt, hire specialists, pay construction firms, redirect R&D resources on the design side to fabrication, etc. There is a good chance it would actually bankrupt AMD; it almost did before. Again, if Intel considered moving away from being a primary manufacturer given their decades of experience and existing infrastructure, AMD would be next to hopeless at this point. Developing new process nodes is complicated work.
 
Back
Top