NVIDIA Titan might lose its performance crown soon.

[N19h7m4r3]

It's not too long before AMD officially announce their new R9 series of cards. Codenamed Volcanic Islands.

There's been quite a few leaks the past few days, and the first benchmarks have just been leaked. Showing the new flagship card outperforms the Titan.

http://videocardz.com/45753/amd-radeon-r9-290x-slightly-faster-gtx-titan

http://www.legitreviews.com/amd-sha...and-center-ready-to-launch-hawaii-gpus_124633

I'd very interested in the possibility adding a mid-highend (dual 6-pin power connections) R9 card to my MP, and does anyone here think these might be compatible in either OS X 10.8, or 10.9, considering that Apple is now moving towards AMD for the new Mac Pro.

AMD will also be live streaming the event straight from Hawaii
http://www.youtube.com/watch?v=bHfmM6QYWNM

 

[beaker7]

4 GB Ram? What year is this?

 

[deconstruct60]

4 GB Ram? What year is this?

2013 were cards aimed primarily aimed at gaming ( 2 DVI and one display port connector ) don't particularly need a whole lot more. The FirePro derivative of this likely will be 2x that amount if grumbling about computation RAM resources.

Titan is a rebrand release of the earlier K20 card. Similar issue

It's not too long before AMD officially announce their new R9 series of cards. Codenamed Volcanic Islands.

Announce and ship are likely two different things. The relative smaller die size suggests that AMD has time to market ( easier to get out of TSMC's fabs ) and costs as bigger objectives than ultimate performance crown.

So yet another flurry of fanboy chest beating from both sides as one temporarily leapfrogs the other.

If Apple is going to sync to FirePro derivations of this architecture it would likely be close to a 12 months from now until saw drivers. ( presuming doesn't ship for 1-2 months and probable sync to OS X 10.10 ).


P.S. It would be rather huge change of pace if derivatives of these got included in the Mac Pro 2013. But more likely those derivatives seems destin for the 2014 (perhaps early 2015) version.

 

[beaker7]

2013 were cards aimed primarily aimed at gaming ( 2 DVI and one display port connector ) don't particularly need a whole lot more.

Hi.

Don't tell me what I need.

Thnx.

 

[deconstruct60]

Hi.

Don't tell me what I need.

It isn't about what you want ( there is a difference between want and need/require ) and what the game's hard requirements. I'm not saying anything about individual wants. Come up with a list of game that require 2+GB of VRAM to work as part of their specificatios. If you can't compose a list of substantive length, that's what I'm talking about. Not your tangent to that.

 

[beaker7]

It isn't about what you want.

Actually, if a company is trying to sell me a product, that is exactly what it is about.

In any case, bottom of the barrel requirements of some hypothetical game are completely and utterly irrelevant to my needs.

 

[Erasmus]

2013 were cards aimed primarily aimed at gaming [which] don't particularly need a whole lot more.

Don't tell me what I need.

Beaker7, unless I am much mistaken, you are not a computer game.

Anyways, although I'm sure it won't happen, would love to see two of these in the new Mac Pro. It will be disappointing to see the new Mac Pro come out AFTER a new generation of components have been released, but still containing the old lot.

 

[N19h7m4r3]

Announce and ship are likely two different things. The relative smaller die size suggests that AMD has time to market ( easier to get out of TSMC's fabs ) and costs as bigger objectives than ultimate performance crown.

So yet another flurry of fanboy chest beating from both sides as one temporarily leapfrogs the other.

If Apple is going to sync to FirePro derivations of this architecture it would likely be close to a 12 months from now until saw drivers. ( presuming doesn't ship for 1-2 months and probable sync to OS X 10.10 ).


P.S. It would be rather huge change of pace if derivatives of these got included in the Mac Pro 2013. But more likely those derivatives seems destin for the 2014 (perhaps early 2015) version.

Oh I know announce and shipping are different, although an announcement is still important to get a clear list of specification until reviews are published.

What depends on this now is how long before the NDA gets lifted, and we get some proper information.

Also many suggest that it's technically a refresh of the current 7xxx series, if so the drivers might be very similar. That would open up the door for use in OS X earlier than we think.

 

[deconstruct60]

It will be disappointing to see the new Mac Pro come out AFTER a new generation of components have been released, but still containing the old lot.

While disappointing it probably shouldn't be unexpected. When has Apple's Mac Pro in complete sync with the latest, bleeding edge graphics? I think that is one reason for the shift to FirePro variants is that they come out at a pace closer to Apple's pace. If the expectations are how quickly "pro graphics" update cycles goes then that will lead to less expectation mismatch.

It will be disappointing if AMD also releases new FirePro at the same time as this more gamer oriented card. But if the new FirePro don't show up till mid-2014 then there isn't going to be a new generation of FirePro products for a while. Last year AMD put a significant delay gap before releasing the FirePro variants. However, last year AMD wasn't licensing FirePro out to Apple either. It is possible. It just doesn't fit the track record for either one.

[ It does point to yet another reason why Apple's custom FirePro implementation shouldn't be sky high priced though even if they are W7000/W9000 variants. ]

 

[deconstruct60]

Also many suggest that it's technically a refresh of the current 7xxx series, if so the drivers might be very similar. That would open up the door for use in OS X earlier than we think.

It isn't GCN v2, but it is a different resource allocation/implementation. Radical changes to the drivers probably aren't necessary, but not going to arrive at optimized results with old stuff (or minor tweaking to old stuff).

If the NDA lifts and AMD is releasing new FirePro with variations on these resource implementations this Fall, then sure it could be sooner. The problem is that there is no track record for Apple+AMD allocating the resources so that OS X drivers were a "release day" outcome. They could do that, but for a variety of reasons they haven't before.

 

[N19h7m4r3]

It isn't GCN v2, but it is a different resource allocation/implementation. Radical changes to the drivers probably aren't necessary, but not going to arrive at optimized results with old stuff (or minor tweaking to old stuff).

If the NDA lifts and AMD is releasing new FirePro with variations on these resource implementations this Fall, then sure it could be sooner. The problem is that there is no track record for Apple+AMD allocating the resources so that OS X drivers were a "release day" outcome. They could do that, but for a variety of reasons they haven't before.

Quite true, and looking at the release of the original 7xxx series it took them months to a year to finally have drivers optimised for performance.

Either way, it'll be interesting to see what AMD announce in detail.
Depending on it, and reviews later it'll decide if I buy a current 7xxx series for my Mac Pro or wait a bit.

Although I'm sure some folks here might pick one up early to try and work working.

 

[slughead]

It will be disappointing if AMD also releases new FirePro at the same time as this more gamer oriented card. But if the new FirePro don't show up till mid-2014 then there isn't going to be a new generation of FirePro products for a while. Last year AMD put a significant delay gap before releasing the FirePro variants. However, last year AMD wasn't licensing FirePro out to Apple either. It is possible. It just doesn't fit the track record for either one.

... But regardless of the release date of the new Fire Pro or this product, nMP users wont be able to upgrade while old MP users will.

 

[N19h7m4r3]

... But regardless of the release date of the new Fire Pro or this product, nMP users wont be able to upgrade while old MP users will.

That's also true, although for all we know Apple could have been working with AMD to possibly use these new cards in the nMP. MacVidCards already showed that his flashed 7xxx cards show up as FirePro's in 10.9 Mavericks.

At any rate we'll find out more from AMD in 7.5 hours!

Folks can watch the live stream here:

http://www.youtube.com/watch?feature=player_embedded&v=bHfmM6QYWNM

 

[beaker7]

... But regardless of the release date of the new Fire Pro or this product, nMP users wont be able to upgrade while old MP users will.

Sure you can. You'd just have to buy a new Mac Pro.

 

[deconstruct60]

... But regardless of the release date of the new Fire Pro or this product, nMP users wont be able to upgrade while old MP users will.

There is no guarantee of that. New cards require new drivers. To date, new cards targeted at Mac Pro PCI-e slots had drivers made for them by AMD (and add-in-card vendors ) and Apple. The 3rd party market at this point is limited to used Macs ( a stagnant and steadily shrinking pool of systems ). That is a market where few, if any, vendors are going to sink R&D effort into. 3rd party support of Mac Pro PCI-e slots by 'new' cards was luke warm when there was a growing pool of systems. Shrinking? Who is going to spend money on that?

Drivers are coupled to the OS. At some point, those old MP systems will be dropped by the OS. Even if try exactly match with the AMD FirePro card infrastructure (at likely substantially higher prices). The old systems are going to get left behind.

The Mac Pro 2013 GPU cards are going to be replaceable. It is not a question if can upgrade, but more so how many parts will be available and when. [ e.g., folks buying CPU/RAM daughterboards for current systems. There were never sold retail and yet folks a buying them. Same thing is likely going to happen for GPU daughterboards over time. ]

 

[N19h7m4r3]

Apparently the R9 280X and lower cards are just refreshed or rebranded 7xxx cards. While the Flagship R9 290X is the new Hawaii core.

here's some leaked pictures for you all.

Just 2.5 hours until we get some proper details though.

 

[deconstruct60]

Apparently the R9 280X and lower cards are just refreshed or rebranded 7xxx cards. While the Flagship R9 290X is the new Hawaii core.
.....
Just 2.5 hours until we get some proper details though.

More like 3.5 hours since they couldn't get the video stream working. ;-)


And really not that much more info in terms of clarifying what is/isn't new architecture wise (from graphics context). Audio processing additions (i.e., more CPU computational offloading ) , new API library, support for 4K monitors , and basic GCN approach

 

[N19h7m4r3]

More like 3.5 hours since they couldn't get the video stream working. ;-)


And really not that much more info in terms of clarifying what is/isn't new architecture wise (from graphics context). Audio processing additions (i.e., more CPU computational offloading ) , new API library, support for 4K monitors , and basic GCN approach

Yup, they didn't even show anything regarding the 290X, also all their benchmarks and demos were run using AMD CPU's. Which means the graphics performance numbers they did show would most likely be higher with a decent Intel system.

Going to have to wait for NDA to lift and Anad and other for full reviews.

At the moment though it seems I might just pick up "cheap" Mac 7950. It seems most of the cards use 6 and 8pin power connections so far.

I'm waiting another week before I make up mind though.

 

[deconstruct60]

Yup, they didn't even show anything regarding the 290X, also all their benchmarks and demos were run using AMD CPU's.

Well they do sell CPUs too. They were giving gobs of partners stage time, might as well let the folks in your same company get some too. I guess I missed it but the demos jumped straight into what was demo and the hardware was all "mystery". I wouldn't be surprised if some of the more intensive games were on 290X and just didn't mention it.

Additionally, as they pointed out a couple of times AMD won all the next generation game consoles. They are pushing the notion of "kill two birds with one stone" by optimizing games for consoles and legacy PC form factors by targeting AMD.

Going to have to wait for NDA to lift and Anad and other for full reviews.

That's the part that doesn't make alot of sense. Pragmatically invite 30K folks to come look at stuff and then hide the info about the stuff behind an NDA. if going to take pre-orders Oct 3 that is less than a week of NDA window. Really? The presentation created as many questions as it resolved. For example Anandtech chart here:

http://www.anandtech.com/show/7369/amd-announces-next-generation-r7-and-r9-video-cards

It really isn't "NA" ( not applicable) as much as the presentation wad devoid on facts on this. The transistors on the 290X are not a secret but the others are ? The pre order date of the others are also a secret? Have benchmark for others but top end one... secret? The card can order soon no price, the ones with no release date ... prices.

They didn't have to reveal all the details with multiple benchmarks and all features, but at least be consistent.

I'm waiting another week before I make up mind though.

Given the week later pre-order date, at least should be able to get a uniformly filled in chart in a week.

 

[N19h7m4r3]

I have to agree on all your points it was rather odd. The viewership did start dropping off significantly after they failed to mention the 290 cards.

Also to note that some of the demos were run on 7990's from what some spotted.

Folks at forums like XtremeSystems and Overclockers.co.uk that are also benchers and vendors noted that they have some cards already.

8-Pack one for the worlds best overclockers stated the Firestrike chart is wrong, citing that the CPU's most likely bottlenecked the cards. He's testing up to 4 of them at a time apparently.

The lack of real information is off putting though, some citing failure and other that it could mean stellar performance.

http://forums.overclockers.co.uk/showthread.php?p=25007620#post25007620

By next week I'll know if I'm going to order the Mac 7950 or an r9 card.


The Mantle API looks amazing though, especially since they claim it will be open and work with upper API's like DX, and OpenGL. It also allows upto 8 core CPU's to be "fully" used by software/games.

 

[deconstruct60]

I have to agree on all your points it was rather odd.

The objective of the presentation covered a wide variety of topics. In addition to the cards I think they are trying to spur interest in the upcoming developers conference. "Here is some info but if really want to know show up in San Jose in a month and a half. It isn't too late to get decent price on plane ticket." . There were a couple other set the stage for later events here. [ As a buzz building exercise not all that bad. ]

The gimmick of bringing folks to Hawaii to talk about the Hawaii code word GPU was rather wasted. But Hawaii is a nice place to go visit so I'm sure folks who got to go had a nice time. I think the expectation was that they would justify that a bit more.

The viewership did start dropping off significantly after they failed to mention the 290 cards.

And part of that was just length of presentation. Human attention span isn't that long if don't carefully craft pacing if there are competing interests for that attention.

The partner demos were largely too long. They should have demoed a thing and got off. If capped to 30 mins then AMD could have got through their points in 30 mins and whole thing wrapped up in an hour. Dragging things past an hour unless there is major change in topic/plot is going to cause problems.

Also to note that some of the demos were run on 7990's from what some spotted.

Not a good sign as to there being short gap between "pre-order" and shipping.

The lack of real information is off putting though, some citing failure and other that it could mean stellar performance.

If buzz was the objective, then numerous forums speculating is hardly a failure. At least not a failure for the short term.

I don't think that they were not trying to pitch this as some Titan killer. Some of the "failure" talk is of expection mismatch that were never really supported all that well. It is competitive not not a "killer". that is the danger of feeding the buzz generation machine. 3rd parties insert stuff and the "telephone game" spins it out of control.

The Mantle API looks amazing though, especially since they claim it will be open and work with upper API's like DX, and OpenGL. It also allows upto 8 core CPU's to be "fully" used by software/games.

8 cores in consoles makes that that core count mark entirely not surprising. Claiming open and it being open are two different things. It depends upon just how much wiggle room there is in the API's abstraction. If it is too tightly coupled to GCN features than it really isn't all that open. If is more aimed at common GPGPU capabilities then it is. Anadtech's article today speculates that it may be a variation on the XBox's new low level API. Again that skews it in one direction ( works on PS4? works for NVidia? works for OS X ? without a significant number of mismatches? )

Too low level and the API isn't portable. Either long term across generations of equipment or to multiple vendors.

There is also potential conflicts between stuff like eyefinity and crossfire and low level. There is a good reasons why APIs are generally has high as they are.

Will be interesting to see if the "True Sound" weaves back into the components that Apple is going to use. Also if this isn't a track to where the DSP engine is just layer on top of the GPGPU infrastructure. It would lend more support for why Apple's two GPU set up would be useful in a wider variety of contexts ( e.g., audio processing). If give audio software an API to a DSP they should not have lots of problems leveraging it.

 

[N19h7m4r3]

The objective of the presentation covered a wide variety of topics. In addition to the cards I think they are trying to spur interest in the upcoming developers conference. "Here is some info but if really want to know show up in San Jose in a month and a half. It isn't too late to get decent price on plane ticket." . There were a couple other set the stage for later events here. [ As a buzz building exercise not all that bad. ]

The gimmick of bringing folks to Hawaii to talk about the Hawaii code word GPU was rather wasted. But Hawaii is a nice place to go visit so I'm sure folks who got to go had a nice time. I think the expectation was that they would justify that a bit more.

Personally if I flew out all the way and had to sit through all of it, I would not be happy. Especially because of the real lack of GPU information.

And part of that was just length of presentation. Human attention span isn't that long if don't carefully craft pacing if there are competing interests for that attention.

The partner demos were largely too long. They should have demoed a thing and got off. If capped to 30 mins then AMD could have got through their points in 30 mins and whole thing wrapped up in an hour. Dragging things past an hour unless there is major change in topic/plot is going to cause problems.

This was very true, not to mention the one really monotonous chap, each time he came on they lost a few hundred viewers a minute.

The partner demo's, while nice I guess were far too long. Especially of the game Lichdom which I really did not care for at all.

Not a good sign as to there being short gap between "pre-order" and shipping.

I'm certainly waiting until then for solid information, if there's nothing I'll go with the card I have in mind.

8 cores in consoles makes that that core count mark entirely not surprising. Claiming open and it being open are two different things. It depends upon just how much wiggle room there is in the API's abstraction. If it is too tightly coupled to GCN features than it really isn't all that open. If is more aimed at common GPGPU capabilities then it is. Anadtech's article today speculates that it may be a variation on the XBox's new low level API. Again that skews it in one direction ( works on PS4? works for NVidia? works for OS X ? without a significant number of mismatches? )

I think most things that help take advantage of more cores efficiently is good, but like yourself, Anand , others and I, I do have the same concerns. John Carmack seems to think it would be great, especially for the SteamBox. Although he also said he won't be using it any time soon.
I wonder if that's any possible indication with it's compatibility/use with OpenGL.

I guess we'll just have to wait until next week when the pre-orders start and we possibly get some solid information.

 

[slughead]

There is no guarantee of that. New cards require new drivers. To date, new cards targeted at Mac Pro PCI-e slots had drivers made for them by AMD (and add-in-card vendors ) and Apple. The 3rd party market at this point is limited to used Macs ( a stagnant and steadily shrinking pool of systems ). That is a market where few, if any, vendors are going to sink R&D effort into. 3rd party support of Mac Pro PCI-e slots by 'new' cards was luke warm when there was a growing pool of systems. Shrinking? Who is going to spend money on that?

True, drivers may be a problem... but a bigger problem than relying on 3rd parties to create a totally new form-factor of each card or relying on Apple to release multiple video card options? Yeah I think not.

How's that 5870 working out for you? Keep in mind it is STILL the top video card available as a BTO option in the Mac Pro. This card was released in September 2009 - Four years and counting! Happy 4th birthday!

I took this nostalgic picture... Today.

You want to talk about small markets. The number of Old Mac Pros out there will outnumber the new Mac Pros for a very long time, and just look at the plethora of options Apple has laid out for us!

The Mac Pro 2013 GPU cards are going to be replaceable. It is not a question if can upgrade, but more so how many parts will be available and when. [ e.g., folks buying CPU/RAM daughterboards for current systems. There were never sold retail and yet folks a buying them. Same thing is likely going to happen for GPU daughterboards over time. ]

You don't know that, and even if it were true, the options will be extremely limited. For as few "mac editions" we see in the standard PCIe form-factor, the likelihood of even ONE nMP 3rd party card is close to zero. That means that unlike the current MP, you'll have to rely entirely on Apple for upgrades (RUH ROH). After Apple danes to actually upgrade the iTube with a decent video card (2015?) there might be a ridiculously-overpriced "replacement" part you can retrofit into the 2013 (14?) model--provided they don't alter the proprietary form-factor over the course of a year, making the newer models video cards incompatible with the original model. Ah, the horrifying realities of proprietary form-factors!

This is also assuming Apple doesn't abandon the iTube form-factor before they update the video cards. That's quite a lot of faith you're putting into this product lines popularity! How much R&D are they going to pump into cramming the latest cards into an iTube-sized heat-dissipation-nightmare before they decide to divert that budget into creating a new hockey-puck mouse for their Mac Mini?

 

[tomvos]

Claiming open and it being open are two different things.

Looks like your comment is correct.

"Nachdem wir unseren privaten Gesprächstermin mit AMDs verlorenem Sohn und Softwarechef Raja Koduri hatten, haben wir in Sachen offener Plattform noch einmal nachgehakt. AMD versteht Mantle nicht als offenen Standard à la OpenCL oder OpenGL. Man hofft aber, dass sich die API zu einer Art Industriestandard entwickeln wird und außer Dice und einigen bislang noch nicht näher genannten weiteren Entwicklern insbesondere die Unterstützung in den großen Lizenz-Engines etablieren wird. Denn die, so Koduri, hätten inzwischen einen weitaus wichtigeren Stellenwert als noch vor zehn Jahren. Daher sei auch eine Analogie zu Glide nicht ganz zutreffend. Falls aber ein anderer Hardwareanbieter AMD anspreche, um ebenfalls ein Backend für Mantle mit Spezialtreiber für seine eigene Hardware zu erstellen, werde AMD diese Ansuchen nicht von vornherein ablehnen, ergänzte Koduri."

Source: http://www.pcgameshardware.de/Grafikkarten-Grafikkarte-97980/Specials/AMD-Low-Level-API-Mantle-Battlefield-4-1090085/
​

A rough translation of the important tidbits:
- AMD does not consider Mantle to be an open standard like OpenCL or OpenGL.
- If another hardware vendor contacts AMD in order to create a backend for Mantle with a special device driver for his own hardware, then AMD does not generally decline such a solicitation.

For me this sounds as open as a G8-summit.

 

[pastrychef]

True, drivers may be a problem... but a bigger problem than relying on 3rd parties to create a totally new form-factor of each card or relying on Apple to release multiple video card options? Yeah I think not.

How's that 5870 working out for you? Keep in mind it is STILL the top video card available as a BTO option in the Mac Pro. This card was released in September 2009 - Four years and counting! Happy 4th birthday!

I took this nostalgic picture... Today.
Image

You want to talk about small markets. The number of Old Mac Pros out there will outnumber the new Mac Pros for a very long time, and just look at the plethora of options Apple has laid out for us!




You don't know that, and even if it were true, the options will be extremely limited. For as few "mac editions" we see in the standard PCIe form-factor, the likelihood of even ONE nMP 3rd party card is close to zero. That means that unlike the current MP, you'll have to rely entirely on Apple for upgrades (RUH ROH). After Apple danes to actually upgrade the iTube with a decent video card (2016?) there might be a ridiculously-overpriced "replacement" part you can retrofit into the 2014 (15?) model--provided they don't alter the proprietary form-factor over the course of a year, making the newer models video cards incompatible with the original model. Ah, the horrifying realities of proprietary form-factors!

This is also assuming Apple doesn't abandon the iTube form-factor before they update the video cards. That's quite a lot of faith you're putting into this product lines popularity! How much R&D are they going to pump into cramming the latest cards into an iTube-sized heat-dissipation-nightmare before they decide to divert that budget into creating a new hockey-puck mouse for their Mac Mini?

Are you going to try and hijack another thread and turn it in to a New Mac Pro hate thread? That horse has been beaten to death already.