Vendor lockout of GPUs? *spawn*
- Thread starter Lurkmass
- Start date
I guess by that point Linux would be the less hostile platform in terms of compatibility than Windows and Microsoft loses their API monopoly advantage since they become redundant. Developers would rather just move on to Linux than to enable a new monopoly from Microsoft that doesn't work for them. Microsoft would be ruined since their business model relies on maximizing compatibility ...
DavidGraham
Veteran
Nope.
Intel simply can't lock out another vendor.
https://www.anandtech.com/show/4122/intel-settles-with-nvidia-more-money-fewer-problems-no-x86
If you looked at the next page it lists out in a table that Nvidia doesn't get a DMI/QPI Bus License or the x86 license. Intel virtually didn't have to give up anything aside from some chump change while Nvidia eventually forfeited producing nForce chipsets ...
If Nvidia truly did have a compelling case then they could've won far more guarantees by taking it to the court to ensure access to any interface technology that Intel uses but since they weren't confident enough they instead settled for cash while cross-licensing their graphics IPs to Intel ...
The key phrase being "several years" and it's been a decade at this point so any legally binding commitments imposed have already expired so it's only a matter of when they'll decide to block others again ...
They will run them on Intel and AMD x86 CPUs just like they do today. Because Intel isn’t going to abandon the massive x86 + windows PC ecosystem just because they have a discrete GPU now.
It really doesn’t. You’re simply overestimating Intel’s power over the PC ecosystem.
In your hypothetical scenario do both Intel and AMD completely drop support for industry standard peripheral interfaces like PCIe and Thunderbolt and Microsoft and enterprise clients will be fine with it? Because that’s what it would take to blacklist peripheral manufacturers. Even Apple uses Bluetooth and USB.
DavidGraham
Veteran
Because they wanted to, not because Intel forced them, NVIDIA was thinking about abandoning chipsets for quite sometime, they were transitioning to GPGPUs and mobile SoCs.
It's Intel that settled, not NVIDIA.
Until they get legally blocked from doing so again, and with a simple complaint to the FTC.
Intel certainly doesn’t seem to be interested in walled gardens. Maybe they’re under new management?
“In addition to Intel’s Thunderbolt silicon, next year Intel plans to make the Thunderbolt protocol specification available to the industry under a nonexclusive, royalty-free license. Releasing the Thunderbolt protocol specification in this manner is expected to greatly increase Thunderbolt adoption by encouraging third-party chip makers to build Thunderbolt-compatible chips. We expect industry chip development to accelerate a wide range of new devices and user experiences.”
https://newsroom.intel.com/editorials/envision-world-thunderbolt-3-everywhere/
“In addition to Intel’s Thunderbolt silicon, next year Intel plans to make the Thunderbolt protocol specification available to the industry under a nonexclusive, royalty-free license. Releasing the Thunderbolt protocol specification in this manner is expected to greatly increase Thunderbolt adoption by encouraging third-party chip makers to build Thunderbolt-compatible chips. We expect industry chip development to accelerate a wide range of new devices and user experiences.”
https://newsroom.intel.com/editorials/envision-world-thunderbolt-3-everywhere/
Yes, it's called USB4 now. However Intel hasn't made public any such plans for TB4 where they're currently at or future revisions
When did I imply that Intel will abandon the x86 architecture ? Even if they don't create a walled garden yet with this move they certainly get closer by doing it ...
Am I ?
PCIe is far more niche than either Bluetooth or USB in reach and Intel is going to introduce another proprietary interconnect interface for their upcoming GPU compute segment. Xe Link could be repurposed in the future to service graphics acceleration. Out of the gate Intel are already attempting to bundle their GPUs with their CPUs at high-end compute markets so what is stopping them from doing the same for the high-end graphics market ?
At the end of the day Intel still achieved their end goal which was to stop Nvidia from producing integrated graphics chipsets ...
As far as I know both parties have to be in agreement to reach settlement. Nvidia could've very well pressed Intel into court but they didn't so they decided to chicken out instead because they thought that the risk/reward benefit of continuing litigation wasn't worth it ...
What exactly would Nvidia have to be able to open a new case against Intel ? They don't have either an x86 license or access to all of Intel's future interfaces so no court would realistically entitle Nvidia to have rights to them ...
“Intel says that it does not charge royalty or licensing fees for the Thunderbolt specification. The company says will continue to be the case going forward with Thunderbolt 4, too.”
https://www.theverge.com/circuitbre...ndard-docks-cables-usb4-accessories-port-usbc
Oh, must have missed that. Let me rephrase it then: Intel hasn't shown any public estimates on when TB4 might become open (after they have TB5 ready probably like with TB3?)
What do you mean by “become open”? Are they charging royalties for TB4?
No, they're not charging royalties. Charging royalties would mean the tech was available for others. However that's not the case, Intel isn't allowing anyone else to make TB4 controllers at this time. If it supports TB4, it has Intel chip(s) on it .
Silent_Buddha
Legend
This is especially true now that AMD once again has a competitive CPU and in some workloads (applications that can take advantage of massive multhithreading) a far superior CPU to anything that Intel has. And unlike when Opteron and Athlon 64 were competitive with Intel, there isn't nearly as much industry or consumer resistance to buying AMD CPUs.
The only reason that Intel maintains as much of an advantage in CPU sales is that AMD is extremely limited compared to Intel with regard to wafer starts for their CPUs.
So, even in a hypothetical world where Intel wants to abandon the Windows PC market or at the very least alienate the Windows PC market by locking out hardware from other vendors being able to run on their Motherboards, that would backfire spectacularly on on them. All it would accomplish is to drive more of their consumers into wanting to switch to a different hardware vendor.
It's the same reason that AMD would never lock out NV GPUs on their platform even if they hypothetically controlled 75% of all PC CPU/MB sales.
Obviously where are some sectors where something like this could occur (supercomputers, for example), but those sectors will have little to no bearing on the Windows PC/server market.
It's amazing what actual competition does to make companies be more consumer friendly.
Regards,
SB
Indeed, MS knows how to flex its muscles at large IHVs. Back in the mid-90s, MS threatened Intel to drop support for future processors in Windows 95, if Intel insisted with their proprietary NSP framework (MMX, etc.), instead of using APIs from MS.
Ok, but what does have to do with being “open”? Intel is actively encouraging TB adoption by peripheral manufacturers. That was the whole reason for eliminating the royalty fee. Meaning they’re not interested in creating some new interface that only works with their hardware.
Motherboard manufacturers are free to add TB controllers to AMD systems. Those systems can then work with TB peripherals from any other manufacturer. No Intel lock-in required.
Last edited:
Yes, they are free to add Intel TB4 controller to AMD systems. AMD however doesn't have option to put TB4 controller in their CPUs or chipsets. Or MediaTek to make competing discrete TB4 controllers etc etc.
Why does it matter who makes the TB4 controller? The point is that anyone can buy the controller and build a TB4 capable motherboard and it doesn’t need to be an “Intel PC”. TB4 isn’t some proprietary interface that can only be used by Intel devices.
In case you missed it this conversation is about a hypothetical Intel strategy to create a proprietary device interface that can only be used by its CPUs and GPUs and therefore push competition out of the market. The point is that they’re currently doing the exact opposite.
What's the significant difference between USB4 and TB4 except for tighter mandatory specifications, which you can enforce only, if you own an ecosystem?
As far as I've seen from the respective wikipedia articles, there's mandatory PCIe-tunnel of 32 Gbps for TB4, which is optional at this speed for USB4 and TB4 uses some vt-d functionality to improve security against physical DMA attacks.
D
Deleted member 13524
Guest
TB4 doesn't use PCIe 4.0?
Decreasing the performance hit on eGPUs should have been TB4's main feature..
Decreasing the performance hit on eGPUs should have been TB4's main feature..
Similar threads
