Monday, May 27th 2019

AMD Announces Radeon RX 5700 Based on Navi: RDNA, 7nm, PCIe Gen4, GDDR6

AMD at its 2019 Computex keynote today unveiled the Radeon RX 5000 family of graphics cards that leverage its new Navi graphics architecture and 7 nm silicon fabrication process. Navi isn't just an incremental upgrade over Vega with a handful new technologies, but the biggest overhaul to AMD's GPU SIMD design since Graphics CoreNext, circa 2011. Called RDNA or Radeon DNA, the new compute unit by AMD is a clean-slate SIMD design with a 1.25X IPC uplift over Vega, an overhauled on-chip cache hierarchy, and a more streamlined graphics pipeline.

In addition, the architecture is designed to increase performance-per-Watt by 50 percent over Vega. The first part to leverage Navi is the Radeon RX 5700. AMD ran a side-by-side demo of the RX 5700 versus the GeForce RTX 2070 at Strange Brigade, where NVIDIA's $500 card was beaten. "Strange Brigade" is one game where AMD fares generally well as it is heavily optimized for asynchonous compute. Navi also ticks two big technology check-boxes, PCI-Express gen 4.0, and GDDR6 memory. AMD has planned a July availability for the RX 5700, and did not disclose pricing.
Add your own comment

202 Comments on AMD Announces Radeon RX 5700 Based on Navi: RDNA, 7nm, PCIe Gen4, GDDR6

#151
Casecutter
This is not surprising I think the "big Navi" was always getting the "Next-Gen" architecture it's just now they made RDNA as the "term" for the Gaming centric arrangement of the individual nucleotides (building block) that make "Next-Gen" architecture. Next-Gen is an all encompassing idea of such building blocks, and not one architecture that fits all. I see AMD/RTG splitting the Professional on a separate strand, that makes use of nucleotides that offer the best HPC, AI, or whatever the task requires. These nucleotides (bits and pieces) can also go down to the console market and APU's.

Remember AMD/RTG is setting itself up against a huge onslaught, and not just Nvidia or Enthusiast Gaming market which is just a pittance for profits. Compared that to what they might miss-out on if Intel makes strides into all these various emerging markets, while even later unlocks consoles and delivers their own true APU's. My hope is the Raja and all these other folks that were "brought-in" by Intel where compartmentalize, and didn't have full-understanding to the vision Lisa Sue (upper management) developed since mid 2016, but I think AMD has had such strategic overview compromised.

https://www.pcgamesn.com/intel/gpu-hires-amd-nvidia
Posted on Reply
#152
GoldenX
So, RDNA is compatible with the GCN ISA. I lost all hope of a proper OpenGL driver.
Posted on Reply
#153
John Naylor
Yawwwwn .... based upon recent years, I have learned that until I see test results here on TPU and elsewhere, it's not real. These announcements and single game testing never live up to the hype.
Posted on Reply
#154
Valantar
John Naylor said:

Yawwwwn .... based upon recent years, I have learned that until I see test results here on TPU and elsewhere, it's not real. These announcements and single game testing never live up to the hype.
We obviously need reviews, but there's little reason to suspect that AMD's engineering team can't come up with a more powerful and efficient architecture when aiming for a blank-slate design - it just takes time. This has been on the roadmaps for a few years already, so the arch is probably 4-5 years in the making, with efforts intensifying in the past couple. I just hope they've taken the necessary time to make it stick on the first try. Perhaps that was the reason for Navi being late? If so, I sure don't mind.
Posted on Reply
#155
bug
P4-630 said:

AMD Radeon RX 5000 is hybrid with elements of GCN - "pure" RDNA only in 2020 - Sweclockers
Amd/comments/but81o
https://www.sweclockers.com/nyhet/27618-amd-radeon-rx-5000-ar-hybrid-med-inslag-av-gcn-renodlad-rdna-forst-ar-2020

Navi die:

https://nl.hardware.info/nieuws/65723/computex-rx-5000-serie-van-amd-wordt-hybride-van-gcn-en-rdna-pure-rdna-komt-met-navi-20
If true, this takes Rebrandeon to the next level: different architectures not only within the same product line, but within the same chip line.
Posted on Reply
#156
Casecutter
Why is it we get one data point of say 90 (30 games x 3 resolutions) and we have all we need to see it as allegement of hyperbole.
Posted on Reply
#157
bug
Casecutter said:

Why is it we get one data point of say 90 (30 games x 3 resolutions) and we have all we need to see it as allegement of hyperbole.
Leaked info this close to a launch, usually depicts a best case scenario, that's why. It's not foolproof, but it's an educated guess ;)
Posted on Reply
#158
GoldenX
bug said:

If true, this takes Rebrandeon to the next level: different architectures not only within the same product line, but within the same chip line.
Pascal is the same thing with Maxwell then. Turing is Pascal with RTX on top of it then.
Posted on Reply
#159
bug
GoldenX said:

Pascal is the same thing with Maxwell then. Turing is Pascal with RTX on top of it then.
What?
Posted on Reply
#160
GoldenX
bug said:

What?
It's not a rebrand if you keep things from the old specs to keep compatibility. Also, G92 wants it's rebrand meme back.

So, no numbers on the 5700, it sounds like it will be another boring launch.
Posted on Reply
#161
bug
GoldenX said:

It's not a rebrand if you keep things from the old specs to keep compatibility. Also, G92 wants it's rebrand meme back.
Oh crap, selective memory strikes again.
Posted on Reply
#162
GoldenX
bug said:

Oh crap, selective memory strikes again.
All 3 companies are rebrand masters.
AMD has the disaster of old GCN cards getting into new series, RX500 vs RX400, etc.
Nvidia has the G92 fiasco (9 cards?), the 100, 300 and 800 series, and 90% of their mobile chips.
Intel has 4 series of CPUs with exactly the same IGP, they only added an U to the start of the name.

VIA is the only good guy here.
Posted on Reply
#163
bug
GoldenX said:

All 3 companies are rebrand masters.
AMD has the disaster of old GCN cards getting into new series, RX500 vs RX400, etc.
Nvidia has the G92 fiasco (9 cards?), the 100, 300 and 800 series, and 90% of their mobile chips.
Intel has 4 series of CPUs with exactly the same IGP, they only added an U to the start of the name.

VIA is the only good guy here.
This was about desktop GPUs, so that's what i was talking about.
AMD routinely does stuff like: https://en.wikipedia.org/wiki/AMD_Radeon_Rx_200_series#Chipset_table (i.e. everything from Terrascale to GCN 1.3, all under the 200 moniker).

The new ground they could be breaking is Navi chips actually being from different families if the above rumor is true: little Navi built with GCN blocks, big Navi without.
Posted on Reply
#164
GoldenX
bug said:

This was about desktop GPUs, so that's what i was talking about.
AMD routinely does stuff like: https://en.wikipedia.org/wiki/AMD_Radeon_Rx_200_series#Chipset_table (i.e. everything from Terrascale to GCN 1.3, all under the 200 moniker).

The new ground they could be breaking is Navi chips actually being from different families if the above rumor is true: little Navi built with GCN blocks, big Navi without.
The 100 and 300 series are desktop cards... "He that is without sin among you..."
Posted on Reply
#165
Valantar
bug said:

This was about desktop GPUs, so that's what i was talking about.
AMD routinely does stuff like: https://en.wikipedia.org/wiki/AMD_Radeon_Rx_200_series#Chipset_table (i.e. everything from Terrascale to GCN 1.3, all under the 200 moniker).

The new ground they could be breaking is Navi chips actually being from different families if the above rumor is true: little Navi built with GCN blocks, big Navi without.
Some GCN blocks. That distinction can matter quite a lot depending on what blocks they are. Calling this a rebrand, though? That's idiocy. Even if it carries over some parts of the design, it's a brand-new die design with brand new core components. If that's a rebrand, there has really never been a new chip designed, ever.
Posted on Reply
#166
bug
GoldenX said:

The 100 and 300 series are desktop cards... "He that is without sin among you..."
They may be rebrands (man, did you dig up series I never knew existed on the desktop), but they still don't mix different architectures under the same moniker.

Valantar said:

Some GCN blocks. That distinction can matter quite a lot depending on what blocks they are. Calling this a rebrand, though? That's idiocy. Even if it carries over some parts of the design, it's a brand-new die design with brand new core components. If that's a rebrand, there has really never been a new chip designed, ever.
Come on people. Is my English that bad? This is about putting the same label on unrelated products, not about rebranding.
Posted on Reply
#167
GoldenX
bug said:

They may be rebrands (man, did you dig up series I never knew existed on the desktop), but they still don't mix different architectures under the same moniker.
Low end 700 series are Fermi, low end 400 are Tesla, the 750Ti is Maxwell v1 in Kepler's lineup, G92 cards are in 4 different series (8000, 9000, 100, 200).
Nvidia has been having a good conduct lately, doesn't mean they are innocent.
Posted on Reply
#168
bug
GoldenX said:

Low end 700 series are Fermi, low end 400 are Tesla, the 750Ti is Maxwell v1 in Kepler's lineup, G92 cards are in 4 different series (8000, 9000, 100, 200).
Nvidia has been having a good conduct lately, doesn't mean they are innocent.
So, bottom line, you're ok if little Navi turns out a Frankenstein monster and big Navi is the completely new architecture. I'm not.
Posted on Reply
#169
GoldenX
bug said:

So, bottom line, you're ok if little Navi turns out a Frankenstein monster and big Navi is the completely new architecture. I'm not.
The 750ti was little Frankenstein monster Maxwell and no one complained.
What maters is if the product is good, internally it could be an Intel IGP for all I care.
Posted on Reply
#170
bug
GoldenX said:

The 750ti was little Frankenstein monster Maxwell and no one complained.
Do you even understand what that article was talking about? It said the little Navi could be built with Navi and GCN blocks and only the big Navi will be entirely new. The 750Ti was nothing like that.
GoldenX said:
What maters is if the product is good, internally it could be an Intel IGP for all I care.
In general, yes. But when you mix architectures like AMD does, you end up with missing features, depending on the model. Whether some codec isn't hardware accelerated on older parts or a new HDMI revision isn't supported, there's a lot of aspects where you can end up drawing the short straw.
Posted on Reply
#171
GoldenX
Getting AMD cards is being a beta tester for them, look at the Vega 56, beating the 64 when undervolted AND overclocked.
Posted on Reply
#172
medi01
Exactly what needs to happen for idiotic GCN news to stop?

GCN is an instruction set that is not getting dropped any time soon, definitely not sooner than nVidia drops its 11 years old CUDA.
As for "microarchitectures" AMDs' own Vega is quite different to Polaris.

GoldenX said:

So, RDNA is compatible with the GCN ISA. I lost all hope of a proper OpenGL driver.
And 2080 is compatible with 11 years old CUDA.
And, wait for it, Zen2 is compatible with 39 years old x86!!!

AMD just cannot innovate!

GoldenX said:

Getting AMD cards is being a beta tester for them, look at the Vega 56, beating the 64 when undervolted AND overclocked
Beating 2070 when overclocked + overvolted:



What trickery is this? How dare they sell us slower cards for cheap?
Posted on Reply
#173
bug
medi01 said:

Exactly what needs to happen for idiotic GCN news to stop?

GCN is an instruction set that is not getting dropped any time soon, definitely not sooner than nVidia drops its 11 years old CUDA.
As for "microarchitectures" AMDs' own Vega is quite different to Polaris.


And 2080 is compatible with 11 years old CUDA.
And, wait for it, Zen2 is compatible with 39 years old x86!!!

AMD just cannot innovate!


Beating 2070 when overclocked + overvolted:



What trickery is this? How dare they sell us slower cards for cheap?
Probably you informing yourself: https://en.wikipedia.org/wiki/Graphics_Core_Next
Graphics Core Next (GCN) is the codename for both a series of microarchitectures as well as for an instruction set
But I'm not holding my breath.
Posted on Reply
#174
Aquinus
Resident Wat-man
btarunr said:

Navi also ticks to big technology check-boxes
Typo. Should be two not to.
Posted on Reply
#175
GoldenX
medi01 said:
And 2080 is compatible with 11 years old CUDA.
And, wait for it, Zen2 is compatible with 39 years old x86!!!
And still AMD drivers are the worst at OpenGL since... ATI. They never solved it.

CUDA has a lot of driver optimization work, the ISA could be different between generations, Nvidia does the work to unify them in CUDA. You know, it's just another compute language.
Posted on Reply
Add your own comment