Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

AMD Confirms RX 480 At $199 USD, Other APU & Polaris Announcements

It appears that HBM2 APUs are coming, but not this year. None of roadmaps are pointing this out, but work on HBM2 tech on APUs is already undergoing. AMD is going to release them next year for two reasons: 14 nm+ process, which will allow for higher frequencies and lower thermal designs, and second reason is... HBM2 prices. AMD aims at 50-55$ manufacturing cost for the chips with HBM2, TSV's and Interposer, to make them sell at 199-259$ price points.

Ryzen Die size is around 190 mm2. Ryzen APU will be slightly bigger. Each die, with packaging and validation costs AMD 40$. Interesting times ahead.

One on engineering teams is aiming at 3.5 GHz all core Turbo, and 1.2 GHz core clocks on the GPU, for 65W TDP.
 
16173958820l.jpg

This is comparing 35W, Mobile Eng Sample with A12 9800, with 65W TDP. A12 9800 has 8CU GPU with 1108 MHz core clock, and DDR4 memory compatibility.

The 35W APU have had 3.0/3.3 GHz CPU core clock, and 800 MHz GPU core clock. 65W TDP GPU target is 1.1-1.2 GHz. So basically 65W TDP version will be around 30% faster than mobile version in terms of graphics.

If anyone is interested. A12 9800 with 2400 MHz DDR4, is able to average 50 FPS in Overwatch in 1080p/Medium preset. RX 550 is getting 68 FPS in High preset in the same resolution.
 
AidenShaw said:
As in "secured a contract for future deliveries" - the iMac Xeon isn't available for seven months.
Click to expand...

I am not quite familiar about CPU models. Do you know if Apple means Skylake EX? Or Cannonlake-EP? Or something else? Didn't know which "on the way" Xeon has 8 cores and Turbo 4.5GHz. Do you mind tell me a bit more (e.g. the CPU model) so that I can goggle a bit more info about it? Million thanks in advance.
 
h9826790 said:
I am not quite familiar about CPU models. Do you know if Apple means Skylake EX? Or Cannonlake-EP? Or something else? Didn't know which "on the way" Xeon has 8 cores and Turbo 4.5GHz. Do you mind tell me a bit more (e.g. the CPU model) so that I can goggle a bit more info about it? Million thanks in advance.
Click to expand...
Xeon's that use Skylake architecture.
 
h9826790 said:
I am not quite familiar about CPU models. Do you know if Apple means Skylake EX? Or Cannonlake-EP?
Click to expand...

There are no -EX or -EP going forward. There is -SP (scalable: 2+ packages ) and -W (workstation: 1 package)

http://www.anandtech.com/show/11332...-skylakesp-in-bronze-silver-gold-and-platinum

The iMac Pro has a -W. The Mac Pro will highly likely have a -W also.

Technically Intel hasn't introduced the -W models (e.g., whether sticking with Xeon E5 1600 v5 as a name), but they are coupled to the -X series ( similar to the way -E and -EP were ).

http://www.anandtech.com/show/11464...ging-18core-hcc-silicon-to-consumers-for-1999

The iMac Pro (and Mac Pro ) does not have the -X series. Apple mentioned Xeon explicitly in intro and the memory is listed as ECC RAM which -X series doesn't handle. Intel is just not pulling the Xeon E5 1600 v5 introduction schedule forward to compete with Ryzen. They are probably waiting for all details of Threadripper to drop and then counter with the rest of the Xeon E5 1600 info. That is going to take time. It is also going to take time for the initial bubble for AMD's GPU to settle down so that Apple can get a steady supply. That plus some perhaps 1-2 other components' availabiliity is the slide into December for the iMac Pro.

( Just because AMD releases Vega models Apple chose perhaps in late Summer-early Fall doesn't Apple is going to be available to get them at reasonable cost and binned to their thermal constraints at that time. )
 
Last edited:
deconstruct60 said:
There are no -EX or -EP going forward. There is -SP (scalable: 2+ packages ) and -W (workstation: 1 package)

http://www.anandtech.com/show/11332...-skylakesp-in-bronze-silver-gold-and-platinum

The iMac Pro has a -W. The Mac Pro will highly likely have a -W also.

Technically Intel hasn't introduced the -W models (e.g., whether sticking with Xeon E5 1600 v5 as a name), but they are coupled to the -X series ( similar to the way -E and -EP were ).

http://www.anandtech.com/show/11464...ging-18core-hcc-silicon-to-consumers-for-1999

The iMac Pro (and Mac Pro ) does not have the -X series. Apple mentioned Xeon explicitly in intro and the memory is listed as ECC RAM which -X series doesn't handle. Intel is just not pulling the Xeon E5 1600 v5 introduction schedule forward to compete with Ryzen. They are probably waiting for all details of Threadripper to drop and then counter with the rest of the Xeon E5 1600 info. That is going to take time. It is also going to take time for the initial bubble for AMD's GPU to settle down so that Apple can get a steady supply. That plus some perhaps 1-2 other components' availabiliity is the slide into December for the iMac Pro.

( Just because AMD releases Vega models Apple chose perhaps in late Summer-early Fall doesn't Apple is going to be available to get them at reasonable cost and binned to their thermal constraints at that time. )
Click to expand...

Thanks for the detailed reply. I have a clearer picture now. The iMac Pro should have the Xeon version (ECC RAM) of 7820X.

A bit like the i7-990X and W3690.
 
http://www.techarp.com/articles/amd-vega-memory-architecture/

Vega memory subsystem analysis, and Q&A video.

  • AMD Vega was specifically architected to handle big datasets, with a heterogenous memory architecture, a wide and flat address space, and a High Bandwidth Cache Controller (see 1:34).
  • Large amounts of DRAM can be used to handle big datasets, but this is not the best solution because DRAM is costly and consumes lots of power (see 2:54).
  • AMD chose to design a heterogenous memory architecture to support various memory technologies like HBM2 and even non-volatile memory (e.g. Radeon Solid State Graphics) (see 4:40 and 8:13).
  • At any given moment, the amount of data processed by the GPU is limited, so it doesn’t make sense to store a large dataset in DRAM. It would be better to cache the data required by the GPU on very fast memory (e.g. HBM2), and intelligently move them according to the GPU’s requirements (see 5:40).
  • The AMD Vega’s heterogenous memory architecture allows for easy integration of future memory technologies like storage-class memory (flash memory that can be accessed in bytes, instead of blocks) (see 8:13).
  • The AMD Vega has a 64-bit flat address space for its shaders (see 12:08, 12:36 and 18:21), but like NVIDIA, AMD is (very likely) limiting the addressable memory to 49-bits, giving it 512 TB of addressable memory.
  • AMD Vega has full access to the CPU’s 48-bit address space, with additional bits beyond that used to handle its own internal memory, storage and registers (see 12:16). This ties back to the High Bandwidth Cache Controller and heterogenous memory architecture, which allows the use of different memory and storage types.
  • Game developers currently try to manage data and memory usage, often extremely conservatively to support graphics cards with limited amounts of graphics memory (see 16:29).
  • With the introduction of AMD Vega, AMD wants game developers to leave data and memory management to the GPU. Its High Bandwidth Cache Controller and heterogenous memory system will automatically handle it for them (see 17:19).
  • The memory architectural advantages of AMD Vega will initially have little impact on gaming performance (due to the current conservative approach of game developers). This will change when developers hand over data and memory management to the GPU. (see 24:42).
  • The improved memory architecture in AMD Vega will mainly benefit AI applications (e.g. deep machine learning) with their large datasets (see 24:52).
 
koyoot said:
https://www.scan.co.uk/shop/compute...d-radeon-vega-frontier-edition-graphics-cards

999$ for air cooled version. 1499$ for water cooled version.
Click to expand...
  • “Vega” GPU Architecture.
  • 64 Next-Gen Compute Units (nCUs) (4096 Stream Processors)
  • 16GB High Bandwidth Cache (HBC) Memory.
  • 483 GB/s Memory Bandwidth.
  • 90 Gpixels/s Fillrate
  • 13.1 TFLOPS Peak FP32 Compute Performance
  • 26.2 TFLOPS Peak FP16 Compute Performance
  • Power Connectors 2 x 8-pin PCIe
  • 3x DisplayPort™ 1.4 HBR3/HDR Ready, & 1x HDMI™ 4K60 Display Outputs
  • DirectX® 12.1, OpenGL® 4.5, OpenCL™ 2.0, & Vulkan® 1.0 API Support
  • Dual-slot, full length (10.5”) Form Factor.
  • Air Cooling Solution. Liquid-cooled versions also available.

No mention about TDP yet. 2 x 8-pin PCIe doesn't sound very promising...

I suppose iMac Pro's Vega 56 and Vega 64 numbers refer to the amount of CU's?
 
jeanlain said:
300W TDP is a lot. Why would the liquid-cooled card consume more if it has the same TFLOPS?
Click to expand...

No, all numbers (except the one that specifically stated for liquid cool), are for the air cool version (check the small print). The liquid cool version will be more powerful (higher TFLOPS).
 
h9826790 said:
No, all numbers (except the one that specifically stated for liquid cool), are for the air cool version (check the small print). The liquid cool version will be more powerful (higher TFLOPS).
Click to expand...
No it doesn't. It has the same specs as air cooled version.

It appears also that all links have been taken down right now.

300W means what the board can handle. Just by looking at the characteristics of the GloFo/Samsung process vs 28 nm, the GPU should have the same power consumption, as Fury X(246W Under load). My prediction, before any leaks was that GPU's actual power consumption will be around 260W. We will see If I will be correct on this.

And just to give you an example. Radeon Pro Vega 64 has to fit in 175W TDP for iMac, with 1.35 GHz core clock.

I wish the liquid cooled will be also released for gaming market and not cost more than 1000$ :(. Dream is max 699$.

P.S. Im wondering if anyone will spot the elephant in the room in those specs and benchmarks. Cinebench is relying heavily on CPU performance, and causes overhead for the GPUs. So actually the GPUs will be faster than that.
 
koyoot said:
No it doesn't. It has the same specs as air cooled version.

It appears also that all links have been taken down right now.

300W means what the board can handle. Just by looking at the characteristics of the GloFo/Samsung process vs 28 nm, the GPU should have the same power consumption, as Fury X(246W Under load). My prediction, before any leaks was that GPU's actual power consumption will be around 260W. We will see If I will be correct on this.

And just to give you an example. Radeon Pro Vega 64 has to fit in 175W TDP for iMac, with 1.35 GHz core clock.

I wish the liquid cooled will be also released for gaming market and not cost more than 1000$ :(. Dream is max 699$.

P.S. Im wondering if anyone will spot the elephant in the room in those specs and benchmarks. Cinebench is relying heavily on CPU performance, and causes overhead for the GPUs. So actually the GPUs will be faster than that.
Click to expand...

Where this 175W coming from?

TBH, I very doubt if the iMac Pro can handle the heat. Unless Apple assume their buyer only use the iMac Pro for Youtube, Facebook, etc. That's cooling system is ridiculously under perform to cool a high core count Xeon and a AMD Vega 64.

In fact, they just make the same mistake as the 6,1. Down volt / clock the GPU to fit in the thermal envelope. This GPU should be good up to 13.1 TFLOPS, but only can deliver up to 11 TFLOPS on the iMac Pro. A guarantee 20% penalty even before the GPU launch just because "form over function". Really bad idea.

Anyway, back to that 300 vs 375W. If exact same spec, is it possible AMD assume the air cooled version will be thermal throttling? Therefore, when under stress, the GPU will warm up quickly and hit the max temperature, forced to stay at 300W. But the liquid cooled version can continuously draw 375W due to liquid cooling. That may also means the air cooled version can only deliver 13.1TFLOPS for a short period of time. But the liquid cooled version may able to deliver that without time limit?
 
  • Like
Reactions: Synchro3
h9826790 said:
Anyway, back to that 300 vs 375W. If exact same spec, is it possible AMD assume the air cooled version will be thermal throttling? Therefore, when under stress, the GPU will warm up quickly and hit the max temperature, forced to stay at 300W. But the liquid cooled version can continuously draw 375W due to liquid cooling. That may also means the air cooled version can only deliver 13.1TFLOPS for a short period of time. But the liquid cooled version may able to deliver that without time limit?
Click to expand...
No. It will not throttle. More likely Water Cooled version has higher Cooling capacity for higher possibility of OC.

The thing is those GPUs are supposed to be stable as hell, therefore, Raja have said that gaming cards will be faster than those, and more gaming focused. 2 Gbps HBM2 for 512 GB/s, 1650-1700 MHz GPU, and you end up very close to 300W.

All I am saying: those numbers indicate how much cooling power is in the shroud, not the actual power draw of the GPUs. Process alone should make the Founder Edition on par with Fury X in terms of power consumption(250W).
 
koyoot said:
No. It will not throttle. More likely Water Cooled version has higher Cooling capacity for higher possibility of OC.

The thing is those GPUs are supposed to be stable as hell, therefore, Raja have said that gaming cards will be faster than those, and more gaming focused. 2 Gbps HBM2 for 512 GB/s, 1650-1700 MHz GPU, and you end up very close to 300W.

All I am saying: those numbers indicate how much cooling power is in the shroud, not the actual power draw of the GPUs. Process alone should make the Founder Edition on par with Fury X in terms of power consumption(250W).
Click to expand...

So, you mean that TDP is purely talking about the cooler's capacity, nothing to do with the graphic card's real power draw?
 
h9826790 said:
So, you mean that TDP is purely talking about the cooler's capacity, nothing to do with the graphic card's real power draw?
Click to expand...
IMO, in this case - yes. Otherwise there would be no sense for two GPUs one with air cooler, and lower TBP, and second with water cooling, with higher TBP to have identical specs.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back