New AMD Radeon Pro 5600M Mobile GPU Brings Desktop-Class Graphics Performance and Enhanced Power Efficiency to 16-inch MacBook Pro

[AleksandarK]

AMD today announced availability of the new AMD Radeon Pro 5600M mobile GPU for the 16-inch MacBook Pro. Designed to deliver desktop-class graphics performance in an efficient mobile form factor, this new GPU powers computationally heavy workloads, enabling pro users to maximize productivity while on-the-go.

The AMD Radeon Pro 5600M GPU is built upon industry-leading 7 nm process technology and advanced AMD RDNA architecture to power a diverse range of pro applications, including video editing, color grading, application development, game creation and more. With 40 compute units and 8 GB of ultra-fast, low-power High Bandwidth Memory (HBM2), the AMD Radeon Pro 5600M GPU delivers superfast performance and excellent power efficiency in a single GPU package.



"More than ever, pro users require the freedom to create wherever their work takes them," said Scott Herkelman, corporate vice president and general manager, Graphics Business Unit at AMD. "With AMD RDNA architecture, the AMD Radeon Pro 5600M GPU offers the optimal combination of compute horsepower and power efficiency, providing an outstanding mobile graphics solution to power a wide range of demanding workloads on-the-go."

Key capabilities and features of the AMD Radeon Pro 5600M GPU include:

• Exceptional compute performance - Equipped with 40 compute units, the AMD Radeon Pro 5600M GPU delivers up to 5.3 TFLOPS of single-precision (FP32) floating point performance.
• HBM2 Memory - 8 GB of HBM2 with 394 GB/s bandwidth provides ultra-fast transfer speeds to power data-intensive pro applications.
• AMD RDNA architecture - AMD RDNA architecture delivers exceptional performance matched with remarkable power efficiency optimized for mobile computing platforms.

The new AMD Radeon Pro 5600M GPU adds a new super high-performance option to the existing Radeon Pro 5300M and 5500M GPU options for the 16-inch MacBook Pro. For more information about the MacBook Pro visit https://www.apple.com/macbook-pro-16/.

View at TechPowerUp Main Site

 

[Daven]

Too expensive but at least its the full, fat Navi at 2560 SPs (don't let the 5600 in the model fool you).

Edit: This also means that AMD was able to get the full Navi below 60 Watts as the MBP complete system tops out around 100 W.

 

[Assimilator]

Wow, that's effectively a 5700 XT, but with HBM2. Clocks will be super low to limit the power consumption (the reason HBM is used), so it won't be particularly fast... but it will be exceptionally expensive.

 

[Daven]

Wow, that's effectively a 5700 XT, but with HBM2. Clocks will be super low to limit the power consumption (the reason HBM is used), so it won't be particularly fast... but it will be exceptionally expensive.

I missed the HBM2 spec. I think that's also because of the limited PCB space in the MBP.

 

[xman2007]

I take it this is an apple only sku? I'd love to see this in an ryzen laptop as it would probably be a better option than the vega igp or Nvidia 1650/60 options

 

[Assimilator]

I missed the HBM2 spec. I think that's also because of the limited PCB space in the MBP.

Yup.

I take it this is an apple only sku? I'd love to see this in an ryzen laptop as it would probably be a better option than the vega igp or Nvidia 1650/60 options

This chip will be clocked so low and downclock so often that it probably won't even match performance of a 5500. Further, because (a) it's custom (b) it uses HBM it will cost far more cash than any sensible PC owner is willing to pay.

 

[SL2]

Why do they call it 5600 tho? Neither AMD nor Nvidia misses a chance to give a slower mobile chip the same number as the faster desktop counterpart.

 

[Assimilator]

Why do they call it 5600 tho? Neither AMD nor Nvidia misses a chance to give a slower mobile chip the same number as the faster desktop counterpart.

Mobile GPU naming schemes have always been a s**tshow, even compared to what happens on desktop with rebrands.

At least we're finally at a place where desktop GPUs are power-efficient enough to be put into laptops as-is, with nothing more than a downclock.

 

[SL2]

Mobile GPU naming schemes have always been a s**tshow, even compared to what happens on desktop with rebrands.

Yeah, kind of what I was implying, but you rarely see this kind of understated model numbers IMO (even if this is slower than a DT 5600XT).
Usually it's like same SP count = same model number. Or worse.

 

[Kaotik]

This is Navi 12, chip designed exclusively for Apple like Vega 12 (sold as Radeon Pro Vega 16 & 20 on Macs) before it.
On most basic, simplified level it's Navi 10 with GDDR6 memory controllers switched to HBM2.

 

[PerfectWave]

this is a RADEON PRO not just a radeon. GOT IT?

 

[john_]

Wow, that's effectively a 5700 XT, but with HBM2. Clocks will be super low to limit the power consumption (the reason HBM is used), so it won't be particularly fast... but it will be exceptionally expensive.

I was reading estimations of about 1050 MHz speed for the GPU.

 

[Kaotik]

I was reading estimations of about 1050 MHz speed for the GPU.

No need to estimate, it's max clocks are 1035 MHz, as stated on AMD's specs page

 

[THANATOS]

5600M has 32% more GFlops and 2x more bandwidth, so It won't be slower than 5500M.
AMD Radeon Pro 5600M - 2560SP; Turbo 1035Mhz; 5300 GFlops; 50W TDP
AMD Radeon Pro 5500M - 1536SP; Turbo 1300Mhz; 4000 GFlops; 50W TDP
It has more CU, but also lower clocks and more efficient HBM2, so 50W TDP could be enough without too much throttling.

 

[catulitechup]

very interesting

sadly amd dont have more products with HBM2 in desktop (RX 5500 with HBM2 could be have very interesting TDP)

 

[Darmok N Jalad]

I wouldn’t be too fast to assume frequency will be really low. Recall that Vega in Renoir is clocked at up to 1.75GHz on a 15W package that includes an 8C/16T CPU. Not the same architecture, but I don’t see Navi going backward here anyway. There appears to be a really nice sweetspot along the power/frequency curves for AMD’s architectures on 7nm, and I suspect it’s somewhere around 1.7GHz for the GPU, and 4.2GHz for the CPU.

Also, the next generation consoles probably aren’t going to be massive power consumers, as that makes it really hard to drive down complexity and cost over time, which is usually a major goal for margins and adoption rate. The PS5 is packing a GPU that hits 2.23GHz. If PS4 Pro or XoX are any indication, I doubt Sony is letting that custom chip exceed 200W. The PS4 Pro could barely pull 150W from the wall socket, so the custom chip was even less than that.

Edit,
Well, the other article suggests it's around 1.0GHz. A little disappointing, but maybe it met Apple's ask.

 

[Kaotik]

I wouldn’t be too fast to assume frequency will be really low. Recall that Vega in Renoir is clocked at up to 1.75GHz on a 15W package that includes an 8C/16T CPU. Not the same architecture, but I don’t see Navi going backward here anyway. There appears to be a really nice sweetspot along the power/frequency curves for AMD’s architectures on 7nm, and I suspect it’s somewhere around 1.7GHz for the GPU, and 4.2GHz for the CPU.

Also, the next generation consoles probably aren’t going to be massive power consumers, as that makes it really hard to drive down complexity and cost over time, which is usually a major goal for margins and adoption rate. The PS5 is packing a GPU that hits 2.23GHz. If PS4 Pro or XoX are any indication, I doubt Sony is letting that custom chip exceed 200W. The PS4 Pro could barely pull 150W from the wall socket, so the custom chip was even less than that.

Edit,
Well, the other article suggests it's around 1.0GHz. A little disappointing, but maybe it met Apple's ask.

It's not suggested, it's been public since AMD announced the damn thing, 1035 MHz, and there's nothing disappointing about it. It's not that the chip couldn't clock higher, it's the fact that it's TDP is 50 watts and the thing is same size as Navi 10 (RX5700XT)
Next gen consoles are using architecture which is 50%+ more power efficient, this is RDNA1, not RDNA2 like consoles.

 

[delshay]

I missed the HBM2 spec. I think that's also because of the limited PCB space in the MBP.

We need HBM2 for Nano cards, you can't do a Nano card without HBM2.

 

[Caring1]

I was under the assumption HBM ran hotter than "normal" vRam, is that true?

 

[Minus Infinity]

Gee maybe all the semiconductor experts can actually wait until we see the benchmarks before proclaiming it's uselessness.

 

[1d10t]

It's a shame that only Apple exclusive.

I was under the assumption HBM ran hotter than "normal" vRam, is that true?

I assume you never had R9 Nano before.
Run "hotter" because they are on the same die, so any GPU Core activity would likely heating up HBM stacks as well since both under the same base plate.

 

[Caring1]

I assume you never had R9 Nano before.
Run "hotter" because they are on the same die, so any GPU Core activity would likely heating up HBM stacks as well since both under the same base plate.

I'm not an early adopter, I prefer to wait and read reviews and get feedback from owners.
I did want a Nano with HBM but reading about heat issues and lack of overclocking, as well as power consumption deterred me.
I really hope these new cards have improved in those departments.

 

[1d10t]

I'm not an early adopter, I prefer to wait and read reviews and get feedback from owners.
I did want a Nano with HBM but reading about heat issues and lack of overclocking, as well as power consumption deterred me.
I really hope these new cards have improved in those departments.

Same here, got my second hand R9 Nano in late 2017 for about $300 and paired with ASRock B350 ITX and Ryzen 7 1700
As for temperature, I never expected R9 Nano perform quite well with just single axial fan in constrained space ( ITX chassis ), 80°C in furmark is remarkable.
TDP is just 150W, right next to GTX 970 but with performance of GTX 980Ti. If you find thats too aggressive you can tune it down with OverdriveNTool.
This card is the third AMD card that excites me most, below ATI 9700 Pro and unlockable 5850 / 6950 of course

 

[BoboOOZ]

I was under the assumption HBM ran hotter than "normal" vRam, is that true?

Total power consumption and thus total TDP are significantly lower than GDDR. But it requires adequate cooling since all the dies are stacked.
But since everything is on the same die, it's much easier to cool with a single copper base plate.