AMD Talks Next Generation Coherent Interconnect Fabric Connecting Polaris GPUs, Zen CPUs and HPC APU

[andybird123]

100GB/s over 4 lanes is a whole lot better than 64GB/s over 16, in some cases perhaps even better than 200GB/s over 16.

PCIe lanes are a limiting factor, if you can push more bandwidth over fewer lanes it matters because you have more lanes available for more of the same and / other PCIe lane IO.

but the articles don't actually say that it operates over pcie lanes, it uses GMI lanes on MCM packages, they haven't actually detailed how it would work externally (NV link uses specific additional connectors, not pcie ones)

its all irrelevant as its not coming to desktop, this is HPC stuff

 

[humbug]

Well thank god for PCIe 4.0 and hopefully 5.0 following that. If NVidia and now it would seem AMD had their way, with NVlink and now this coherent fabric, first starting off in the server and HPC market, just how long until they decide it is needed on the desktop side of things, 2 or 3 years and then we could have well this motherboard has AMD's super fast system on it whereas this one has NVidia's.

It is bad enough having certain games perform much better or worse on each sides hardware, now with the whole variable refresh rate thing you have to chose a side before you buy a monitor. How long before it extends to motherboards and then what would be next..?

Before anyone starts the whole yes but this one uses open standards yada yada yada.... We all know (or we should if we where honest with our selves) that that is meaningless when you only have two players in the market and neither will use what the other side develops regardless.

This is the result of the only two vendors operating in the same space with different agendas.

Open Standards is not the way remember, propitiatory lock-ins is better.

Different motherboards specific for each vendor, game features specific for each vendor, different screens for each vendor. Jen-Hsun Huang reassures you its all good and wholesome. Buy Nvidia you wont regret being locked out of everything else..

 

[bru]

This is the result of the only two vendors operating in the same space with different agendas.

Open Standards is not the way remember, propitiatory lock-ins is better.

Different motherboards specific for each vendor, game features specific for each vendor, different screens for each vendor. Jen-Hsun Huang reassures you its all good and wholesome. Buy Nvidia you wont regret being locked out of everything else..

Sarcasm aside, the point being is that when there are only two players , people can scream open standard or proprietary tech all they like, neither side is going to use the others willingly. So they are both as good as locked down regardless.

You buy a Gsync monitor you are tied to NVIDIA for VRR, you buy a freesync monitor you are tied to AMD for VRR. Yes I know that Intel have said they are interested, but how long is that going to take 3 years 4 maybe more.

 

[TheRealDeal]

Sarcasm aside, the point being is that when there are only two players , people can scream open standard or proprietary tech all they like, neither side is going to use the others willingly. So they are both as good as locked down regardless.

You buy a Gsync monitor you are tied to NVIDIA for VRR, you buy a freesync monitor you are tied to AMD for VRR. Yes I know that Intel have said they are interested, but how long is that going to take 3 years 4 maybe more.

Only because Nvidia so far have chosen not to support the open standard adaptive sync which has nothing to do with AMD. I doubt AMD would be aloud to support G-Sync.

 

[r7slayer]

Sarcasm aside, the point being is that when there are only two players , people can scream open standard or proprietary tech all they like, neither side is going to use the others willingly. So they are both as good as locked down regardless.

You buy a Gsync monitor you are tied to NVIDIA for VRR, you buy a freesync monitor you are tied to AMD for VRR. Yes I know that Intel have said they are interested, but how long is that going to take 3 years 4 maybe more.

Your not "tied" to AMD with Async monitors. It's a open standard. I wonder if they will still call them "Freesync" monitors when another company finally does start using the standard such as Intel.
However nVidia do not allow anyone else to support Gsync.

 

[CAT-THE-FIFTH]

Plus the other aspect is the cost of the FreeSync compatible monitors tends not to be massively higher than similar monitors lacking support AFAIK. I am yet to see many sub £250 G-Sync monitors ATM,and there seems to be a few sub £250 Freesync ones now.

Edit!!

Not sure how it got to Freesync/G-sync already on the first page.

I think we might need to get back onto topic!!

 

[pete910]

Not sure how it got to Freesync/G-sync already on the first page.

I think we might need to get back onto topic!!

Your on overclockers remember

 

[drunkenmaster]

Sarcasm aside, the point being is that when there are only two players , people can scream open standard or proprietary tech all they like, neither side is going to use the others willingly. So they are both as good as locked down regardless.

You buy a Gsync monitor you are tied to NVIDIA for VRR, you buy a freesync monitor you are tied to AMD for VRR. Yes I know that Intel have said they are interested, but how long is that going to take 3 years 4 maybe more.

The difference is Intel, and ARM guys as well in any devices they build, can use adaptive sync as well, it isn't locked in. Also precisely because it is an open standard it's part of the official spec so people aren't being charged extra for it.

So regardless of it being a two player system(it isn't btw) open standards still have massive benefits. Same spec screen on AMD side being 200-300 cheaper, often more. I got a 27" 1400p 144hz screen for £270 brand new, can you see a g-sync screen with those specs anywhere near that price? Open standards in a small market still makes a huge huge difference.

If Nvidia did Mantle but for themselves and never intended to share it both devs wouldn't have trusted it, it wouldn't have been shared with other parties and the rest of the industry wouldn't have caught up overnight as a result. HSA is going to have huge huge real world benefits soon primarily because AMD created a foundations separate from AMD and joint with multiple other industry players.

Just because there are two main players in desktop graphics doesn't in any way mean it's almost the same thing. If it were my screen would have cost £300 more, HSA would be dead, DX11 would be the only way forward at the moment and, well, many other worse things would be happening. Oh, all gpu cards would cost £50-100 more because Nvidia would have had to come up with their own memory standard which would mean two memory types being produced in smaller quantities and a huge extra R&D expense which would have to be paid back.

 

[bru]

Before anyone starts the whole yes but this one uses open standards yada yada yada.... We all know (or we should if we where honest with our selves) that that is meaningless when you only have two players in the market and neither will use what the other side develops regardless.

Well what a surprise, nobody took any notice of that at all.

 

[drunkenmaster]

You said something completely and utterly wrong. You stated open standards don't matter because there are only two players. Even with adaptive sync, before Intel officially support it(but will) the difference is a often £200 increase in price with the non open standard vs no increase in price with the open standard. That isn't a small difference, that is a monumentally huge difference. When you claim something is meaningless with a huge amount of evidence it absolutely isn't, expect to get called on it. £270 I paid for my Acer, what was the cheapest comparable Nvidia screen at the time I bought it, about £600....... meaningless.... okay.

If you didn't want it to be noticed, you shouldn't have said it.

 

[bru]

Price is not the issue, if you want to use VRR you either use NVidia or AMD, you cannot use VVR with the opposing tech at the moment. But you guys continue to say that its cheaper or it open, that is not the point I was making.

 

[drunkenmaster]

Price is not the issue, if you want to use VRR you either use NVidia or AMD, you cannot use VVR with the opposing tech at the moment. But you guys continue to say that its cheaper or it open, that is not the point I was making.

No, you're backtracking, you used VRR as an example of open standards being meaningless. It's painfully obvious that is what you meant and what you said. This thread is about a completely different tech and the point being discussed was that it would be an open standard, that is the common link here. YOu then used VRR to say open standards are meaningless because they make no difference.

Now you're trying to change the argument because what you claimed is patently false. Likewise if Nvidia add NVLink to a motherboard, AMD, Intel can't CHOOSE to use it without paying whenever they want.

Tomorrow AMD couldn't make a g-sync driver and use G-sync. Tomorrow Nvidia and Intel can release an adaptive sync driver and use any adaptive sync screen. You're trying to use the fact that they haven't YET to say there is no difference when again in reality that is still a monumental, fundamental difference in the implementations.

The screens AREN'T Freesync screens, they are adaptive sync screens with Freesync certification(which is free) labelled on them. There is nothing preventing Intel from providing a adaptive sync driver, giving it a name and having the screens certified for free(if they choose to do it freely) as well.

So your point you're trying to say you're making is invalid, and clearly the real point you were actually making is also invalid.

 

[bru]

So your point you're trying to say you're making is invalid, and clearly the real point you were actually making is also invalid.

Not at all the point I made was that at the moment you want VRR you have to choose one or the other, simple as that.
NVidia might decide to take up adaptive sync tomorrow, or offer Gsync to AMD and they accept, but until that happens your have to choose one or the other if you want VRR

 

[andybird123]

. There is nothing preventing Intel from providing a adaptive sync driver, giving it a name and having the screens certified for free(if they choose to do it freely) as well.

.

I would think the fact they've said their hardware doesn't support it would prevent them at this point.

 

[meldarthx]

Also NVlink won't ever see the light of day on x86 system; its for power pc if that.

We will most likely see AMD's fabric on consumer systems. Second - Intel has already said they will support adaptive-sync; just a matter of when. I have a feeling when that happens Nvidia will join most likely.

Adaptive-sync is open for Intel; Nvida; Arm etc. G-sync is not open at all.....for anyone; that's the difference.

 

[Mauller]

Nvlink could come to consumer boards, but not in the way that Coherent Fabric would.

It would only be used as communication between GPU's whle PCIe talks to the CPU.

But Coherent Fabric can do the whole shebang in an AMD only system.

And it is liable to happen considering in the past you had either SLI or Xfire motherboards. And i think Intel would lose the ability to host both at once, unless it picked up Coherent Fabric and Nvlink support. Which i doubt.

 

[meldarthx]

Nvlink could come to consumer boards, but not in the way that Coherent Fabric would.

It would only be used as communication between GPU's whle PCIe talks to the CPU.

But Coherent Fabric can do the whole shebang in an AMD only system.

And it is liable to happen considering in the past you had either SLI or Xfire motherboards. And i think Intel would lose the ability to host both at once, unless it picked up Coherent Fabric and Nvlink support. Which i doubt.

nope NVlink will never see the light of day in x86. One Nvidia doesn't make chipsets for x86 anymore; most MB manufactors aren't exactly happy paying sli tax. 3 read up on NVlink - its for power pc; its purely for HTC systems.

 

[meldarthx]

Nvlink could come to consumer boards, but not in the way that Coherent Fabric would.

It would only be used as communication between GPU's whle PCIe talks to the CPU.

But Coherent Fabric can do the whole shebang in an AMD only system.

And it is liable to happen considering in the past you had either SLI or Xfire motherboards. And i think Intel would lose the ability to host both at once, unless it picked up Coherent Fabric and Nvlink support. Which i doubt.

you forget Intel wants to cut out dgpus all together from their mbs.....

 

[melmac]

I would think the fact they've said their hardware doesn't support it would prevent them at this point.

Where did they say that? I remember reading that they would support adaptive sync in the future but would not be implementing it in their current line up. But I don't remember reading anything about their hardware not supporting it.

If their iGPUs support embedded display port 1.3 then they support adaptive sync.

 

[flopper]

So they're saying that there bridgeless crossfire (XDMA?) is not capable of doing the job it currently manages with the 28nm chipped cards.

There's a lot that really bothers me about the current Fiji cards, Another example is they state they have a couple of techs working full time on memory management for Fiji, That's obviously only a temporary measure so once they introduce the next gen cards will it continue? Will it be a relevant requirement for next gen HBM cards or will HBM1 owners become neglected.
I really like my Fury Tri-x as Sapphire did a bang up job but has it got a lifespan ticking away a lot faster than usual?

Memory management compression is important due to it allows faster exchange. any save with swapping textures faster allows better latency.
Latency is important especially with VR and also with crossfire/sli as until today dual or more cards are a bad idea. anyone buying more than a single card have a worse gaming experience than me with a single card.

VR however now force the manufacturer to create latency low in an unheard level and the Nvdia 980ti for example cant do that as their hardware isnt good enough for it. AMD furyX is. So AMD thinks ahead for the future vs someone else who forces you to upgrade more often and people like that for some reason?

AMD lasts longer due to their tech is better and more advanced.
Even Intel with all the billions of research failed to make a GPU with larrabee and still isnt any good with their GPU solutions.

VR and 240hz screens with HDR is coming and then you want, compression for memory and cards with low latency and the Polaris tech and the upcoming AMD interconnect is a way to deal with the bandwidth needed and the latency management. The next few years will have a increased demand for bandwidth and latency low that the designers and engineers are working on.

Polaris the brighter future