The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
I said FinFet cards in general, not RX480 in particular.


The RX480 is nowadays well above a 970 in performance in all but the most extreme gameworks titles. Just check any review that is less than 1 month old. Yes the 8GB will inevitably make a difference in most titles, some day.

Apart from that there's the fact that it behaves a lot better on new APIs, it's a 1.5 years newer card, it has Freesync / Adaptive Sync, it has HDCP 2.2 through HDMI 2.0, HEVC hardware decoder/encoder, and new drivers are still bringing significant performance upgrades and new features (last of which was yesterday).

Define what you mean by "driver support" and how much are you willing to bet?
Serious question, because I'll take you up on that bet.
RX480 to me is like AMD's 2016 16nm GTX 970. Or an updated Hawaii. Of course it has new features. It's still in the same class though essentially. That doesn't mean I think 970 is a better value!

AMD drops driver support for their cards before NV does, historically without exception. Radeon 4000 doesn't have official support for >= Win 8.1. Fermi support is better than AMD DX11. Have you not noticed this? Intel might be worse .. I have a feeling that people here might not use hardware long enough to care/notice though.
 
Last edited:
"support" can be "dropped" in different ways. Considering how much kepler performance has fallen behind since the introduction of maxwell (and I don't see why maxwell would be any less special-case optimization dependent), I would stay clear of maxwell as soon as pascal was introduced... (not to mention the incredible bad drivers (shader compiler at least) for my maxwell 1 laptop - that path seemed deprecated by maxwell 2 before it ever got stable)
 
Considering how much kepler performance has fallen behind since the introduction of maxwell
Kepler aged fine, I owned a card just 3 months ago, Kepler performed worse mainly due to it's memory restrictions which affected it's ability to perform well in an age where games are memory hungry, Many games were compute heavy as well, and Kepler had worse compute performance than Maxwell and GCN.
 
"support" can be "dropped" in different ways. Considering how much kepler performance has fallen behind since the introduction of maxwell (and I don't see why maxwell would be any less special-case optimization dependent), I would stay clear of maxwell as soon as pascal was introduced... (not to mention the incredible bad drivers (shader compiler at least) for my maxwell 1 laptop - that path seemed deprecated by maxwell 2 before it ever got stable)


That had nothing to do with driver support.....two different things.
 
I don't see why maxwell would be any less special-case optimization dependent), I would stay clear of maxwell as soon as pascal was introduced...
Pascal is more or less a shrink of Maxwell with some minor optimization. The SM instruction set is identical (as in: you can run Maxwell handcrafted assembler CUDA shaders on Pascal without any problems.)
Kepler has a completely different shader core architecture, and it doesn't have the tiling functionality of Maxwell (and Pascal) either.

Did you really manage to not see any of that?
 
Surprised this news has not had more traction but this is the closest thread I think to where it should be:
AMD files patent infringement against LG, MediaTek, Visio, and some others.
https://www.techpowerup.com/230260/amd-files-patent-infringement-complaint-against-lg-vizio-others
According to the complaint, these patents generally relate to architectures for graphics processing unit (GPU) circuitry. The '506 patent relates to "a graphics processing architecture that enables a large amount of graphics data to be rendered to a frame buffer". The '133 patent relates to specialized "texture" processing circuitry that is employed by GPUs. Lastly, the '454 patent relates to a "unified shader" hardware architecture for GPUs. The complaint specifically refers to various televisions and smartphones, specifically, towards the graphics processing systems within those televisions and smartphones - as infringing products.

They are looking for a limited exclusion followed by a more permanent cease-desist.
Here is a copy of the actual filing: http://www.itcblog.com/images/amdcomplaint.pdf
Ironic that Nvidia makes a lot of news-waves with their patent infringements but this does not even register with most even if (big if for it to be 100%) it may have a more solid case.
Cheers
 
I believe Samsung licensed, or was at least looking at according to some stories last year, that IP so it looks like standard patent enforcement to avoid losing them. If they seek large monetary gains then it gets more interesting.
 
I believe Samsung licensed, or was at least looking at according to some stories last year, that IP so it looks like standard patent enforcement to avoid losing them. If they seek large monetary gains then it gets more interesting.
Seems AMD is looking to stop those companies' products in the US, so they must want a license agreement from them, several of these are massive companies (LG and MediaTek) with many products.

A week earlier AMD along with a few other companies reported to being investigated for patent infringement: http://www.ipwatchdog.com/2017/01/23/qualcomm-sony-lg-targeted-section-337-complaint/id=77323/

All getting messy with all the convergence of tech we see these days.
Cheers
 
Last edited:
OK. Got information back on this. Everything I have mentioned here is definitively correct.

Intel is licensing AMD GPU technology. No money has changed hands yet, so there is not financial impact till late in the year, hence nothing in the current earnings report.

The first product AMD is working on for Intel is a Kaby Lake processor variant that will positioned in the entry-to-mid level performance segment. It is in fact an MCM (multi-chip-module), and will not be on-die with the KB CPU. It is scheduled to come to market before the end of the year. I would expect more collaboration between AMD and Intel in the future on the graphics side.

And you can take all that to the bank.
https://hardforum.com/threads/from-...in-futility-h.1900681/page-72#post-1042797295
 
This is from last Q's conference call

Devinder Kumar

Yes, I think, we've talked about IP monetization in particular partnerships. So, AMD with the treasure full of IP that we have, we have 10,000 plus patterns, half of them are US-based. And we saw over the last two or three years that we can go ahead and partner with folks, where we don't want to directly enter market. We will be very careful in terms of who we partner with, where it doesn't come back and directly compete with us in areas that we want to go put products in, and very happy to do on a fair basis a deal with the partner to go ahead and monetize that IP, get cash, benefit the P&L and balance sheet in particular."

So what ever they are doing with Intel, it won't be something that is directly competing in current markets.
 

This would have made a lot of sense back in 2012-2016 when AMD was nowhere near having a competitive CPU architecture in the market, but now that they have Raven Bridge RyZen releasing in the same time frame as this Kaby Lake + Radeon MCM, it looks like they're just nullifying their USP.


Perhaps AMD is very aggressively trying to take over GPU marketshare in the PC market.
 

This is from last Q's conference call

Devinder Kumar

Yes, I think, we've talked about IP monetization in particular partnerships. So, AMD with the treasure full of IP that we have, we have 10,000 plus patterns, half of them are US-based. And we saw over the last two or three years that we can go ahead and partner with folks, where we don't want to directly enter market. We will be very careful in terms of who we partner with, where it doesn't come back and directly compete with us in areas that we want to go put products in, and very happy to do on a fair basis a deal with the partner to go ahead and monetize that IP, get cash, benefit the P&L and balance sheet in particular."

So what ever they are doing with Intel, it won't be something that is directly competing in current markets.

The only real option is Apple - they have the money to throw in to get others making whatever they need, it fits the schedule (they tend to update MacBooks late in the year), it won't step on AMD's feet (AMD isn't providing CPUs or APUs to Apple), and Apple actually would have use for it, put Kaby + Polaris11 (or 12?) on MCM and you get to shrink MacBooks motherboard a LOT, thus enabling them to make the laptop smaller or have bigger battery
 
The only real option is Apple - they have the money to throw in to get others making whatever they need, it fits the schedule (they tend to update MacBooks late in the year), it won't step on AMD's feet (AMD isn't providing CPUs or APUs to Apple), and Apple actually would have use for it, put Kaby + Polaris11 (or 12?) on MCM and you get to shrink MacBooks motherboard a LOT, thus enabling them to make the laptop smaller or have bigger battery


Yeah going by AMD's statements and what ya just stated, that looks like the only viable option AMD would be will to concede to Intel, cause otherwise, right now low end to mid range, is the only place AMD is making making money in the graphics market, so giving that up would just be suicide in the short term.
 
Surely, but at thesame time the Intel contract licences IP is too really important. ( and today have just been announced Lake Crest ) .. its hard for Intel to pass over the patent from AMD ( and Nvidia ) when we are speaking about GPU's ( whatever is the purpose of them ).

We are not speaking about product theres, but patent IP and technology.
 
Last edited:
The only real option is Apple - they have the money to throw in to get others making whatever they need, it fits the schedule (they tend to update MacBooks late in the year), it won't step on AMD's feet (AMD isn't providing CPUs or APUs to Apple), and Apple actually would have use for it, put Kaby + Polaris11 (or 12?) on MCM and you get to shrink MacBooks motherboard a LOT, thus enabling them to make the laptop smaller or have bigger battery

But there's Raven Ridge to do the exact same thing.

Not to mention:

https://twitter.com/BitsAndChipsEng/status/805515262124584960
 
Raven Ridge might not be ready in time; plus its CPU might be slower than whatever Apple is currently using.
 
The only real option is Apple - they have the money to throw in to get others making whatever they need, it fits the schedule (they tend to update MacBooks late in the year), it won't step on AMD's feet (AMD isn't providing CPUs or APUs to Apple), and Apple actually would have use for it, put Kaby + Polaris11 (or 12?) on MCM and you get to shrink MacBooks motherboard a LOT, thus enabling them to make the laptop smaller or have bigger battery
That still begs the question, why not consider a Zen APU? That would seem far simpler than combining an Intel CPU and AMD GPU for a MCM. Especially at the bottom of the stack. For wider distribution the fabric could be withheld. CPU+GPU connected by PCIE instead of mesh. No different than a discrete GPU, but with the compact form factor. That should leave AMD with a superior APU product while Intel still has a quality product and small form factor. MCM really only makes sense with HBM, or a more robust (FPGAs?) processor configuration. APU with HBM and then Optane for system memory. That would definitely be power efficient for a MacBook.

On the flip side, AMD could get a nice cut and competition isn't a concern.
 
Status
Not open for further replies.
Back
Top