New Upcoming ATI/AMD GPU's Thread: Leaks, Hopes & Aftermarket GPU's

Yeah, thats kind of how I see it too.

 

Spoiler: new Leaks

The image shows an early test board, having a standard rectangular PCB. We can also see some labels attached to the PCB/board, such as ‘Typical Samsung 16Gb‘, which appears to indicate the Memory Type used on this board. The leaker says that this GPU is having 3+3+2 modules (8 in total), which means we are looking at a 256-bit wide Memory interface bus for this early engineering sample. There would be a total of 8 GDDR6 dies at the front of the PCB in a 3+3+2 layout.

Another label reads, ‘Full ???? Typical XT A0 ASIC’. This means the sample is a Gaming GPU, because the XT variant/moniker is usually reserved for the Radeon RX series, and not Radeon Pro which target the HPC/AI market. The board doesn’t appear to be a final retail version, since we can clearly see some dip switches and various voltage measurement points as well. At the back of the PCB there are PCB markings for the GDDR6 memory. The card is allegedly using 16 GB GDDR6 memory along a 256-bit bus interface.

From the photo we can also see a huge CPU-type cooler being attached to this graphics processor, few LED load indicators and at least two power connectors, including one 8-pin, to run diagnostics in a test environment. The LEDs are part of the GPU Tach which has been featured on several high-end Radeon RX graphics cards before, and acts like a load meter to measure the amount of load put on the GPU.

AMD will officially unveil the Radeon RX 6000 series on October 28.

THX to Metal Messiah

 

Well I sold my 5700 XT and got myself 980 to tide things over. I can do this for two months.

 

Test that 980 see if its any faster than rx570 these days.

 

What's good in upcoming generation - non reference Radeon versions will actually be quiet.
As always, AIB's will put the same coolers that they design for Nvidia, but this time those coolers will be massive

 

ATI no longer exist.
The company is AMD...and you seem to be right on what's being targeted.
AMD will lose badly to Nvidia with this strategy of holding onto information until the very last minute.
Nvidia...with very few exceptions always seem to meet expectations on performance.
This time around, they've done so on pricing as well.
They're also the mainstream company for all things GPU related...with faster support and more complete software solutions.
My guess is AMD do not yet have a GPU to compete with RTX3080 at this point.
That may be what is driving the silence.
Hopefully I am wrong.
If they had, they'd provide information to keep perspective buyers from getting the RTX3080 at launch.
Nvidia's AIB seem to be already spooled up and ready to go with product, so every one of those sales is one less for AMD.
It doesn't matter if they come out with a cheaper card, if the money is already spent.

 

Rambus Advances HBM2E Performance to 4.0 Gbps for AI/ML Training Applications

Rambus Inc. (NASDAQ: RMBS), a premier silicon IP and chip provider making data faster and safer, today announced it has achieved a record 4 Gbps performance with the Rambus HBM2E memory interface solution consisting of a fully-integrated PHY and controller.
Paired with the industry's fastest HBM2E DRAM from SK hynix operating at 3.6 Gbps, the solution can deliver 460 GB/s of bandwidth from a single HBM2E device.
This performance meets the terabyte-scale bandwidth needs of accelerators targeting the most demanding AI/ML training and high-performance computing (HPC) applications.

"With this achievement by Rambus, designers of AI and HPC systems can now implement systems using the world's fastest HBM2E DRAM running at 3.6 Gbps from SK hynix," said Uksong Kang, vice president of product planning at SK hynix.
"In July, we announced full-scale mass-production of HBM2E for state-of-the-art computing applications demanding the highest bandwidth available."



The fully-integrated, production-ready Rambus HBM2E memory subsystem runs at 4 Gbps without PHY voltage overdrive.
Rambus teamed with SK hynix and Alchip to implement the HBM2E 2.5D system to validate in silicon the Rambus HBM2E PHY and Memory Controller IP using TSMC's N7 process and CoWoS (Chip-on-Wafer-on-Substrate) advanced packaging technologies.
Co-designing with the engineering team from Rambus, Alchip led the interposer and package substrate design.

"This advancement of Rambus and its partners, using TSMC's advanced process and packaging technologies, is another important achievement of our ongoing collaboration with Rambus,"
said Suk Lee, senior director of the Design Infrastructure Management Division at TSMC.
"We look forward to a continued partnership with Rambus to enable the highest performance in AI/ML and HPC applications."

"Alchip brought a demonstrated track record of success in 7nm and 2.5D package design to this initiative," said Johnny Shen, CEO of Alchip Technologies.
"We're extremely proud of our contributions to Rambus' breakthrough achievement."

Rambus has 30 years of high-speed memory design applied to the most demanding computing applications. Its renowned signal integrity expertise was key to achieving an HBM2E memory interface capable of 4 Gbps operation.
This raises a new benchmark for meeting the insatiable bandwidth requirements of AI/ML training.

"With silicon operation up to 4 Gbps, designers can future-proof their HBM2E implementations and can be confident of ample margin for 3.6 Gbps designs," said Matthew Jones, senior director and general manager of IP cores at Rambus.
"As part of every customer engagement, Rambus provides reference designs for the 2.5D package and interposer to ensure first-time right implementations for mission-critical AI/ML designs."

Benefits of the Rambus HBM2E Memory Interface (PHY and Controller):

• Achieves the industry's highest speed of 4 Gbps per pin, delivering a system bandwidth of 460 GB from a single 3.6 Gbps HBM2E DRAM 3D device.
• Fully-integrated and verified HBM2E PHY and Controller reduces ASIC design complexity and speeds time to market
• Includes 2.5D package and interposer reference design as part of IP license
• Provides access to Rambus system and SI/PI experts helping ASIC designers to ensure maximum signal and power integrity for devices and systems
• Features LabStation development environment that enables quick system bring-up, characterization and debug
• Supports high-performance applications including state-of-the-art AI/ML training and high-performance computing (HPC) systems

For more information on the Rambus Interface IP, including our PHYs and Controllers, please visit rambus.com/interface-ip.

 

I do not know, why you even care.
You go with assumption that AMD has nothing competitive with 3080.
And 3070 will be available just before AMD's official launch. Where we may have some extra info anyway.

Nothing to be gained from doom and gloom mentality.

 

I hope, I really really hope... That they've streamlined and optimized their drivers this time around.
I like me some AMD, they've always been the performance per buck company for me.

I look forwards to the release.

 

I have zero doubts that this round AMD is going to compete with Nvidia's high-end. Lisa herself said HALO cards, didn't she.

 

I will tell what i see (Crystall ball = On)

There will be 2 SKUs

1. 80 CUs HBM2 BigNavi maby an XTX Model with LC 650-700€ (~RTX3080)
2. 72 CUs HBM2 smaller Navi (Similar to Vega 64/56 & Fury) 500-550€

3. 64 CUs GDDR6 18GB for the massess up to 449€ (5700XT + 20-30%)
4. 56 CUs cut down version up to 300-349€

There You have it -> this is most positive outcome i see
All SKUs have RT + DX12U

 

I hope you're right. That 80cu Navi would be sweet at that price. I'm not quite that optimistic though.

 

Yes my cristall ball has the same information

also XTX with 16G HBM2E 2 stacks

 

I don't know what to think anymore.... I need to focus on waiting!!

 

Thats because leakers 'maby' think that CDNA Big MI 100 TFLOPs GPU will be avaible also for gaming

 

Not hyped (for balance).

 

@PrEzi : I would not worry. RX 5700 XT replacement will have 64CUs (1.6x over Navi10), gaming clock above 2.05GHz (1.06x above Navi10 minimum).
Improved caches (CU utilization+energy efficiency), data compression (memory bandwidth+energy efficiency), improved command processor. Unknown if Dual or Quad CU, nor if CU is still 2x wave 32. (But surely improved use of shared resources over Navi10.)

And 64CUs RDNA != 64CUs GCN. Would AMD make similar improvements like with RDNA1, One would not have to be worried even if GPU had 120 CUs.

@OnnA : Steve does something he should not. Puts out very negative information and leaves it there for very long time before correcting on it. But with amount of empty talk, most of viewers are not going to get to the end of video as he repeats his negativity few times around and it is boring. (Like running fake news in TV, and then 2 weeks later dementing it in one sentence while 90% of people who head this fake news will not hear they they corrected it.)
Steve looks angry, that he has no idea what's going on... losing relevancy in process. But other internet PC-gaming/HW news producers do not know either. And that's even while they have quite some RDNA2 information already. We do not know final configurations, but that's hardly important when someone knows expected performance per transistor investment, per clock, per watt, ...

 

To be taken with "Grain of Salt" as with all leaked benchmarks.
AMD’s Next-Gen Radeon RX 6000 Series ‘RDNA 2’ Graphics Cards Alleged Benchmarks Leak Out

https://wccftech.com/amd-next-gen-radeon-rx-6000-series-rdna-2-graphics-cards-benchmarks-leak/

 

Problem with AotS benchmark is that it is rather unreliable for GPUs as it changes a lot with CPU(and its clock). Has poor resolution scaling. And does not really tell you actual clock used.
It only tells you that there is AMD's GPU that performs better than RX 5700 XT under rather vague conditions. And that it scores around most of 2080 Ti-s, but worse than OC editions of 2080 Ti with OC.
It is similar to geekbench, where I always ignore top scores and look at most common scores to see what is out of the box performance.

Yet, with this AMD's GPU we do not even know if it is underclocked for some specific test, or it is some maximum OC with increased voltage.
Nor we can know if it is mainstream, high-end or top chip.