Baseless Next Generation Rumors with no Technical Merits [pre E3 2019] *spawn*
- Thread starter Arkham night 2
- Start date
-
- Tags
- non sequitur
- Status
Multi GPU or processing divided over more processors for more overall power. This is where I think customizations need to be for azure. MS needs to deliver the featureset, but once the generation needs more power they can’t go back and toss the old blades, they can leverage multi APU support to bring the power levels up appropriately.
DX12 supports this today so, perhaps that is a potential mitigation strategy for investment costs and ensuring a long life before needing to replace hardware.
Reason 1, According to Lisa Su Sony are partners with AM, along with Amazon and Google (not Microsoft).
Reason 2, ms have picked epyc as it's currently a better deal, I'm sure if intel could match or better the overall deal they would've gone a different route, it's not out of loyalty and as soon as intel are the better option they will switch.
Reason 3, Navi was designed for and with Sony in this partnership, both helping each other to get what was needed. 2/3rds of the Radeon team were put to work specifically for this. AMD are profiting from this work, or at least will do when they sell it in the desktop market place (and selling a version of it, most likely with what Sony added removed or disabled to Microsoft)
Do you need any more perfectly valid reasons why AMD may prefer Sony?
This made me think of the previous leak where it has smt disabled but hyper scheduler enabled in a command line info dump.
Depending on engine would be interesting if dev could select 8, 16, or 24 thread mode like it indicated. (minus OS allocation) I'd assume frequency would change depending on what was selected.
Example:
8t 3.3Ghz
16t 3Ghz
24t 2.8Ghz
See above.
They aren't going to turn down money, but they are the only show in town if you want to get both parts from the same vendor so they have everyone over a barrel and can really do what they want.
Intel and Nvidia already shot their bolt with the Xbox team at the first go.
She also took to the stage with Phil recently.
The fact of the matter is she can't say much until MS discloses what they are doing together, as MS is their client not the other way around.
It doesn't matter that everyone already know / believes that AMD is powering Scarlett.
She also never said anything regarding Sony until Sony talked about it.
It was all just rumors and speculation, not AMD saying we have a closer relationship with Sony than MS.
This doesn't mean they have a better rationship or more customized navi or anything than with MS.
If navi is more customized it's just because that's what Sony paid for.
Last edited:
Well, Lisa Su likes* Cerny. Does she like Spencer too ? I don't think so. She clearly prefers Sony.
*Twitter like
*Twitter like
But she follows Satya Nadella, Phil Spencer, XBox, Microsoft Surface, Microsoft Azure, Bill Gates, and Melinda Gates. So there you have it.
I know for a fact that she didn't sent Phil a birthday card, but cerny got one.
So would that just be migrating the future old hardware towards non-gaming? e.g. deep learning, Cloud-SDK, cloud back-up, cloud render farms, dedicated servers?
Multi-GPU for gaming is practically dead for a number of reasons, but on the HW side, they would need to build a high bandwidth interconnect between the GPUs.
Keep the old HW around for low-tier streaming quality?
------
What's the process currently for large scale data centers in terms of purchasing & replacing? Tangentially, what would Google's process be for Stadia?
-----
The Win10 Linux kernel being set for November 2020 is rather...... coincidental. I'm still not really sure what to think of it.
Maybe GabeN will finally get a competent Steambox solution? :V
Possibly. The multi-adaptor isn't dead! It just was seldom used. Shadow of the Tomb Raider gets excellent saturation and scaling on mGPU with DX12. It's also a prime candidate for CPU core scaling as well.
edit: as I understand it from forum comments, even the ray tracing can be done over multi-adaptor.
changed video for 2080TI ray traced SLI
The high bandwidth interconnect would be server side (i assume blade and possible hardware customizations), I think following along what Google said with Stadia; multi-processor work loads are available to developers if they require more.
I mean practically speaking, that’s effectively dead in terms of industry support.
So I guess I don’t really understand the cost benefit here if they have to build something that’s still very different compared to the console SKU.
Isn’t xCloud currently just a bunch of SAD Xbox One units holding hands? (Ahem)
That'd be so funny if Sony got a better deal from AMD and better tech simply because Lisu Su likes Mark Cerny. The guy;s not even a Sony employee.
It's the only advantage we have on them and the only way we can stall the robot takeover. Who do thay fair typos wtih grammer? Can the stand-under this??
They can't. I used to a be a product manager for a telco a while back in my career. My organization had it's standard business solutions in which we sell solutions to companies wanting to buy products and solutions from us. And then we have to work with other carriers, the departments there are called Partner Solutions. They are fully separated from the rest of the business. they have their own floors, they have their own servers, data etc. No one knows what types of deals or work is happening in partner solutions.
AMD semi-custom department is the same way. You're welcome to ask @Rys about it since he works for AMD RTG. He'll tell you the same, no one in RTG knows what's happening in semi-custom. And the semi-custom teams don't know what the other semi-custom teams are working on. At Apple, team members are not even allowed to ask other employees where they work and what projects they are working on. This applies to the whole campus. There are obvious needs for security, and I assure you the semi-custom department is difficult to learn information from. This is for obvious reasons otherwise AMD will be sued into financial ruin.
There is no such thing as loyalty. Businesses operate with profits. I've seen preference towards a vendor, like most companies prefer using Microsoft Office, but many companies will move to Google Docs because it's cheaper. Not because it's better. It's really that simple.
When it comes to large complex deals, the larger the deals are the more negotiation power there is across the board. So if MS is negotiating a large EYPC data centre contract, they are likely bundling in the Xbox 2 chip for the streaming in the data centres, which should cross over to the deals on the local console as well. This isn't some loyalty thing, it's just a lot of money and any product manger that has worked with a sales team knows that AMD will be more willing to reduce more margin here and there for their products for the net gain on the entire deal.
Last edited:
It's available on Stadia and possibly on Azure. Why notI mean practically speaking, that’s effectively dead in terms of industry support.
Besides, what happens if we can't shrink processors within respectable ROI. Shouldn't multi-adaptor be the next logical step? We see this with infinity fabric.
You’d keep the chip right. Everything else would be designed for a blade. 4 chips on 1 blade they have for the XBO concepts.
More correct to say 4 boards to a blade.
Custom 1S boards
- Status
