New Macbook Pro with vega graphics coming soon

[leman]

ITo me Vega 56 to Vega 20 looks like rx580 to rx560 and I’m eager to learn if this is in fact true.

Well, rx580 and rx560 are two different chips, so I guess that's what koyoot is saying...
[doublepost=1541018429][/doublepost]

Well, on a long shot I gave apple support a friendly call to see if they could make any exceptions on my September purchase, but it was a little too long after the 15-days. Sigh. I guess I'll go back to my original plan of saving up for an egpu.

As I said earlier in the thread, I contacted support and they told me that they will upgrade my model (bought in July) if I pay the configuration price. I will drop by Apple Store this week to make sure I understood them correctly.

At least is it a safe bet that if I grab a Vega 56 or 64 at some point it'll ultimately pack more "oomph" than this mobile Vega 20? (just trying to put a positive spin on the disappointment)

Of course the desktop GPU will be faster by factor of 2 at least.

 

[Thysanoptera]

Well, rx580 and rx560 are two different chips, so I guess that's what koyoot is saying...

Different chips with identical feature sets and performance/watt ratios, that's what I'm trying to figure out. So far the only difference mentioned by anyone was your "optimized geometry engine", which somehow doesnt convince me Vega 12 (that's the correct name, right?) is 50% more efficient than Vega 10.

 

[koyoot]

Different chips with identical feature sets and performance/watt ratios, that's what I'm trying to figure out. So far the only difference mentioned by anyone was your "optimized geometry engine", which somehow doesnt convince me Vega 12 (that's the correct name, right?) is 50% more efficient than Vega 10.

I have a big problem with that "optimized geometry engine"...

Big Vega is famous for its Geometry Engine, that is capable of discarding primitives, that are not visible in the field of view of 3dGraphics(in games), and increase the framerate this way, and mitigate the fact, that there are only 4 Shader/Geometry Engines in the design, which massively bottlenecks the front end of the GPU, considering that 4 Shader/Geometry Engines, have to feed 64 CU's(each SE has 16 CU's).

The problem is... it does not work, at all. NGG path is completely disabled for Vega, and will NOT be available in any way, shape or form for Vega. Yes, Vega 12 is better design than Polaris 11, because you get 4 Shader Engines, which means with each clock, you will get 4 primitives, instead of just two, so the GPU will not be bound on the front end. But, Im not sure it really has "Optimized Geometry Engine". Primitive Shaders, Primitive Culling is coming back with Navi. Without some very specific hardware features(Operand Reuse Cache, or whatever it will be called by AMD), it is impossible for AMD to do what they wanted to do through drivers.

Apple themselves on their site says: "Enhanced Compute Engine" which relates to simply Rapid Packed Math: FP16x2 from FP32 cores, compared to FP16x1 from FP32 cores.

I was personally expecting much worse from Vega 12 in terms of core clock/core count in 35W TDP envelope. 1.3 GHz, from 1280 GCN core chip is pretty much unheard of, till now.

P.S. What matters for GPU power consumption and core clocks is ALWAYS physical design. Vega Mobile is different than Vega 10, for obvious reasons. Voltage curve is different, and MOST LIKELY, the fact that we have 4 Shader Engines, with just 5 CU's in each SE is playing the biggest role in so low power consumption of Vega 12, with such high clock speed. Yes, HBM2@ 1.5 Gbps is consuming just 3W's of power. But all the rest is due to good physical design(good engineering) from AMD.

 

[Thysanoptera]

Big Vega is famous for its Geometry Engine, that is capable of discarding primitives, that are not visible in the field of view of 3dGraphics(in games), and increase the framerate this way, and mitigate the fact, that there are only 4 Shader/Geometry Engines in the design.

The problem is... it does not work, at all. NGG path is completely disabled for Vega, and will NOT be available in any way, shape or form for Vega. Yes, Vega 12 is better design than Polaris 11, because you get 4 Shader Engines, which means with each clock, you will get 4 primitives, instead of just two, so the GPU will not be bound on the front end. But, Im not sure it really has "Optimized Geometry Engine".

Thats the kind of information I was looking for, I didn't know about the NGG paths. I'm confused about the second part though, 4 Shader Engines were in it since Fiji, even before Polaris.

 

[koyoot]

Thats the kind of information I was looking for, I didn't know about the NGG paths. I'm confused about the second part though, 4 Shader Engines were in it since Fiji, even before Polaris.

AMD cannot go higher than 4 Shader Engines, because their Architecture does not have Operand Reuse Cache, which allows those GPUs, to not only beyond 4 Shader/Geometry Engines, but also increases the efficiency and feeds those GPUs properly with data(CUs "know" which instructions were executed, or not executed, and not executed instructions go back to the pipeline to be executed out of order). Saves Register File Size, saves memory bandwidth, saves power, the instruction data is smaller, and easier to execute for the GPU.

Navi appears to have all of those features, and at this point it appears it is confirmed that Navi will have NGG path, and Primitive Shaders fully enabled.

Its good that AMD went with "wider" front end on this GPU, and narrower back-end. 4 Shader/Geometry Engine/5 CU design is much better, than 2 Shader/Geometry Engine 10 CU one, because of the lack of Operand Reuse Cache. GPU will be properly fed with work. It may be first very balanced AMD design since mythical, today, Hawaii GPUs.

 

[cube]

People keep wondering what the performance will be but there's already a "vega" 20 and 24 out on the market...

They're in the Intel NUC 8s...
Base model has a RX Vega M GL with a vega chip, 20 CUs and RX Vega M GH with 24 CUs

Both got a 4 GB single stack of HBM2 as well.

Only difference is that the boost clock speed is a bit lower at around 1200 mhz.

Since they perform at between a 1050 and a 1050 TI, I fully expect these Vega 20 chips to perform at about the same level as the 1050ti and Vega 16 to be at about a 1050.

What is the 560X comparible in the nvidia relm?

Read the thread. Kaby Lake G has FAKE Vega.

 

[koyoot]

Vega 12 clocked at 1.3 GHz is around 64% faster than Radeon Pro 560X. In graphics tests it is between 57 and 72% faster.

Here is direct comparison between the two: https://www.3dmark.com/compare/3dm11/12875524/3dm11/12886777#

The GPU itself appears to be slightly slower than GTX 1060 Max-Q, and on par with Vega M GH.

Pretty nice computer you will have there: 85W, 6 core/12 Thread CPU, 16 GB of RAM, GTX 1060 Max-Q levels of performance. Not bad at all.

I really, really want this GPU on desktop, mainstream market, tho.

 

[littlepud]

I wonder if mobile Vega will be at risk of gpugate down the road, given it's a brand new physical design and package being put into a thin (and hot) chassis. The Radeon 4xx/5xx series have been proven to be reliable workhorses in the 2016-2018 models so far.

 

[ThisBougieLife]

I'm definitely interested in this, but not sure I'm going to risk getting a new MBP if this T2 kernel panic-gate continues.

 

[Thysanoptera]

Navi appears to have all of those features, and at this point it appears it is confirmed that Navi will have NGG path, and Primitive Shaders fully enabled.

So this is not coming to Vega 12? I thought you mentioned this as an example of why Vega 12 will be better.

Its good that AMD went with "wider" front end on this GPU, and narrower back-end. 4 Shader/Geometry Engine/5 CU design is much better, than 2 Shader/Geometry Engine 10 CU one

So this is the first time they reduced significantly the number of CUs within the same shader engine. Just wondering why didn't they do it earlier if that brings so many benefits.

Vega 12 clocked at 1.3 GHz is around 64% faster than Radeon Pro 560X. In graphics tests it is between 57 and 72% faster.

Here is direct comparison between the two: https://www.3dmark.com/compare/3dm11/12875524/3dm11/12886777#

I don't believe it. Not a 35W CPU. This is 1050Ti, desktop, 75W GPU. Why would they test it on old, low end Haswell? And why doesn't this show up in search results? Like - if this is true I'm calling my broker, selling all the stuff I have and buying AMD stock, because Nvidia just got owned. AMD jumped 4 generations in one mobile SKU without even whispering about it. No way.

 

[cube]

I don't believe it. Not a 35W CPU. This is 1050Ti, desktop, 75W GPU. Why would they test it on old, low end Haswell? And why doesn't this show up in search results? Like - if this is true I'm calling my broker, selling all the stuff I have and buying AMD stock, because Nvidia just got owned. AMD jumped 4 generations in one mobile SKU without even whispering about it. No way.

RX 560 has 16CUs. Vega 20 would be equivalent to about 26 Polaris CUs. And then it has HBM instead of GDDR5.

The RX 560"X" in the test is clearly on a mobile platform.

 

[leman]

I have a big problem with that "optimized geometry engine"...

I'm only quitting what AMD is saying in their marketing videos about mobile Vega, no idea what they mean with it. I just assume that it's some sort of incremental improvement over over Vega chip or they probably won't mention it. Or maybe I'm completely mistaken.

Here is direct comparison between the two: https://www.3dmark.com/compare/3dm11/12875524/3dm11/12886777#

I'd take that benchmark with a big grain fo salt since we know nothing about the test platform or even if the results are genuine. The only thing that makes me thing they could be genuine is how conveniently these results overlap with Apple's claims about Vega's performance

 

[koyoot]

So this is not coming to Vega 12? I thought you mentioned this as an example of why Vega 12 will be better.

I said that Vega does not have NGG path enabled, because it would be for AMD too expensive to develop drivers, for. Every specific game would require specifically compiled shader paths for this feature to work. Navi will make it completely hardware level, without driver intervention. What I have said, about Vega, what is making it better is front end wider, than back end, which is exact point of next paragraph.

So this is the first time they reduced significantly the number of CUs within the same shader engine. Just wondering why didn't they do it earlier if that brings so many benefits.

Probably because they were money constrained, they had to compete with Nvidia, and frankly - who cared about it? Even today, if you will have equally performin GPUs from Nvidia and AMD, people in 70% will pick Nvidia. Cursed market. AMD has zero incentive in developing GPUs, because people are simply not buying them, even if they have very good products.

I don't believe it. Not a 35W CPU. This is 1050Ti, desktop, 75W GPU. Why would they test it on old, low end Haswell? And why doesn't this show up in search results? Like - if this is true I'm calling my broker, selling all the stuff I have and buying AMD stock, because Nvidia just got owned. AMD jumped 4 generations in one mobile SKU without even whispering about it. No way.

You do not have to believe in it.

It was not test done by Apple, but by AMD.

P.S. You should've buy AMD when it was 0.89$ . Few weeks ago AMD was 30$ and it appears that thanks to Rome, next year it will be pretty nice bloodbath in stock market, with AMD going up rapidly .

I'm only quitting what AMD is saying in their marketing videos about mobile Vega, no idea what they mean with it. I just assume that it's some sort of incremental improvement over over Vega chip or they probably won't mention it. Or maybe I'm completely mistaken.



I'd take that benchmark with a big grain fo salt since we know nothing about the test platform or even if the results are genuine. The only thing that makes me thing they could be genuine is how conveniently these results overlap with Apple's claims about Vega's performance

They are genuine .

Its funny how people are having hard time believing that AMD could have delivered properly powerful hardware...

 

[leman]

They are genuine .

Do you work for AMD? You seem just awfully sure

 

[koyoot]

Do you work for AMD? You seem just awfully sure

No, I wish I would .

I am confident they are genuine .

 

[Queen6]

I fervently hope AMD does pull it off and gives Nvida a run for it's money - In the mean time this notebook with a 1070 is just killing it...

Q-6

 

[sickntired]

Well, rx580 and rx560 are two different chips, so I guess that's what koyoot is saying...
[doublepost=1541018429][/doublepost]

As I said earlier in the thread, I contacted support and they told me that they will upgrade my model (bought in July) if I pay the configuration price. I will drop by Apple Store this week to make sure I understood them correctly.



Of course the desktop GPU will be faster by factor of 2 at least.

I bought mine in August. If they will let you buy the one with the new Vega graphics for just the difference in price, that would be awesome. I hope you have the rep's name that you spoke to -- that may help you out later. It would actually even be better if you had that rep open a "ticket" so that there is a record of it. You would also probably have more luck in getting the laptop swapped through that rep rather than through the Apple Store because the configuration you want is most likely a BTO and you'd be at the whim of the Apple Store manager about honoring what that rep told you.

What did you say to the rep in the first place to have him/her even make that offer to you? If you are able to do it, that would be the first time I've seen this done when it's way pass the return period.

 

[stevemiller]

I bought mine in August. If they will let you buy the one with the new Vega graphics for just the difference in price, that would be awesome. I hope you have the rep's name that you spoke to -- that may help you out later. It would actually even be better if you had that rep open a "ticket" so that there is a record of it. You would also probably have more luck in getting the laptop swapped through that rep rather than through the Apple Store because the configuration you want is most likely a BTO and you'd be at the whim of the Apple Store manager about honoring what that rep told you.

What did you say to the rep in the first place to have him/her even make that offer to you? If you are able to do it, that would be the first time I've seen this done when it's way pass the return period.

Yeah I gave it a try with no luck and I got mine sept 21. So definitely it’s luck of the draw at best. Also the person I spoke to was simply operating off a script that I don’t think she understood. She suggested I try updating my operating system to get the better graphics card. Oh well.

 

[leman]

I bought mine in August. If they will let you buy the one with the new Vega graphics for just the difference in price, that would be awesome. I hope you have the rep's name that you spoke to -- that may help you out later. It would actually even be better if you had that rep open a "ticket" so that there is a record of it. You would also probably have more luck in getting the laptop swapped through that rep rather than through the Apple Store because the configuration you want is most likely a BTO and you'd be at the whim of the Apple Store manager about honoring what that rep told you.

Yeah, I have a ticket and a transcript.

What did you say to the rep in the first place to have him/her even make that offer to you? If you are able to do it, that would be the first time I've seen this done when it's way pass the return period.

I just explained the situation and they replied immediately (free translation): "thats possible at any time at one of the Apple Stores. A technician will examine your machine and if no damage is discovered you can simply upgrade as you wish".

I should have asked for more details, but I was a bit in a rush so I just let it at that. There is still a lot of room for misinterpretation. This is why I will go to an Apple Store next week and ask there directly.

 

[Sterkenburg]

She suggested I try updating my operating system to get the better graphics card. Oh well.

Lol, sorry man but that's hilarious XD

I spoke to a salesperson today as well, she suggested that I might be able to negotiate a solution in person at an Apple Store, but stressed that normally this would not be allowed, and that she couldn't guarantee anything.
Letting customers upgrade from a semi-new machine by paying the difference would be awesome from Apple, but I reckon that will be the exception, not the rule (i.e. if it's your lucky day they might make an exception for you).

 

[leman]

Yeah I gave it a try with no luck and I got mine sept 21. So definitely it’s luck of the draw at best. Also the person I spoke to was simply operating off a script that I don’t think she understood. She suggested I try updating my operating system to get the better graphics card. Oh well.

Well, that person was clearly incompetent. I'd try again

P.S. I think the wording is very important. I didn't ask if I can return my machine after 5 months. I asked whether I could upgrade the configuration of my existing machine to the new GPUs.

 

[littlepud]

Well, that person was clearly incompetent. I'd try again

P.S. I think the wording is very important. I didn't ask if I can return my machine after 5 months. I asked whether I could upgrade the configuration of my existing machine to the new GPUs.

Apple Stores are way more equipped to handle modular component replacements these days compared to past machine generations (MBP 2011 Radeongate I'm looking at you). It's conceivable they might send a new logic board to the store and swap it out for you as an upgrade, sending your exiting board back into the refurbished/service pool.

As much as I'd want a Vega, if it's only a CTO option I'll likely pass. I've found that stock configurations are the way to go with Apple -- the ability to exchange a DOA unit (or simply if you change your mind, say for color or for 13" vs 15") immediately in-store is very valuable.

 

[Ma2k5]

Do you work for AMD? You seem just awfully sure

Have you not seen his post history from years back? He has been hedging for AMD to supercede Nvidia for years (he thought mobile Nvidia was dead, for us only to see them making $$$$$ with a deal with Nintendo and still continue to dominate the laptop market), then he wonders why people find it hard to believe they will deliver anything to wow us . I don't know anyone who is more AMD-obssessed/fanboy than him on this forum.

Apparently AMD are good honest ethical company, all their marketing is 100% genuine. Any other company however, the devil, we must ignore their marketing etc.

Me, I couldn't care if it's an AMD, Nvidia or Intel, but people need to stop regurgitating stuff from websites as fact, which he does far too often. I have no issue with him being so obsessed with AMD that he reads every article about them everyday, but does he need to turn some of these threads into "AMDRumors"?

 

[leman]

So guys, very sorry to say it but it was false alarm. Just went to Apple store to clarify and it turns out the guy from the online support was a bit of an idiot. When he said I can "upgrade" the machine what was apparently meant is using the buy back option, which is obviously not helpful. Well, at least guys at the Apple Store were very apologetic.

I am very sorry for effectively spreading unhealthy rumours, I should have waited until I could clarify this 100%.

 

[Ma2k5]

So guys, very sorry to say it but it was false alarm. Just went to Apple store to clarify and it turns out the guy from the online support was a bit of an idiot. When he said I can "upgrade" the machine what was apparently meant is using the buy back option, which is obviously not helpful. Well, at least guys at the Apple Store were very apologetic.

I am very sorry for effectively spreading unhealthy rumours, I should have waited until I could clarify this 100%.

Not your fault, but it was the expected outcome. Didn't sound very "Apple" like, the whole pay the difference by handing in a used device.