ASUS Crosshair X670E HERO review
Patriot Viper Venom DDR5 6200 CL40 review
AMD Ryzen 9 7900X processor review
AMD Ryzen 5 7600X processor review
AMD Ryzen 7 7700X review
AMD Ryzen 9 7950X review
DDR5 Ryzen 7 7700 DDR5 memory scaling review
Seasonic Arch Q503 chassis + Connect PSU review
GeForce RTX 4080/4090 announcements (preview)
Corsair HS55 Wireless Core Headset review
New AMD roadmap gives more insight in polaris 10 and 11
A new roadmap appeared online from AMD, in this roadmap you can spot some plans and positioning for the future graphics cards based on the Polaris 10 and 11 GPU design. The roadmap also makes it pretty clear that there will be no rebrands.
The new slide shows product positioning of Polaris and confirms that Polaris 11 will be a more mainstream model with the Polaris 10 to be a high-end SKU series. Polaris 10 is "Ellesmere" and Polaris 11 is "Baffin.
We think it is safe to say that Polaris 10 will be the R9 490X series with 490 and 480 models (high-end). Polaris 11 will be the GPU replacing the 370 and 360 products, likely R7 470 and R9 470. It is unlikely that AMD will announce all products at once, so we expect Polaris 10 appear first around the Computex timeframe with an official launch later. Polaris introduces HVEC (h.265) decode and encode hardware-acceleration and will also support the latest display output standards like DisplayPort 1.3 (how does 3840x2160 @ 120Hz and 1920x1080 @ 480Hz sound?) and of course, HDMI 2.0.
The roadmap also confirms that the Vega architecture is to be released in 2017. Vega will het HBM2 memory and the successor to the current (enthusiast) Fury models. Later on in 2018 the NAVI architecture will be released, positioned with 'next-gen' memory.
In a recent find AMD Polaris 11 was spotted, Device ID 67FF:C8 codenamed “Goose”. This would be the base GPU for a several entry-level products. Now, the CompuBench database reports back that this device has 16 CUs with a maximum clock frequency of 1000 MHz. Multiple your CUs (compute units) by the number of shader processors per cluster (assuming that AMD keeps 64 per cluster) and you'll notice that Polaris 11 in this configuration has 1024 Shader processors tied to a 128-bit bus and 2048 MB of memory.
Polaris 10, codenamed "Ellesmere," would then feature over 2304 stream processors (36 CUs); and Vega 10 featuring 4096 stream processors, with 64 CUs. Things could end up looking like this:
| AMD Polaris / Vega GPU Specs (rumored) | |||
|---|---|---|---|
| AMD Vega 10 | AMD Polaris 10 | AMD Polaris 11 (Dev_ID 67FF) | |
| GPU | Vega 10 / Greenland | Polaris 10 / Ellesmere | Polaris 11 / Baffin |
| Positioning | Enthusiast | High-end | Mainstream |
| Fabrication Process | 14nm FinFET | 14nm FinFET | 14nm FinFET |
| Compute Units | 64 | 36 | 16 |
| Stream Processors | 4096 | 2304 / 2560 | 1024 / 1280 |
| Computing Power | ~8.2 TFLOPs | ~3.7 TFLOPs | ~ 2.0 TFLOPs |
| Core clock | ~1000 MHz | ~800 MHz | ~1000 MHz |
| Effective Memory Clock | ~2000 MHz | ~6000 MHz | ~7000 MHz |
| Memory Bus | 4096-bit | 256-bit | 128-bit |
| Memory | 16GB HBM2 | 8GB GDDR5(x) | 2GB GDDR5 |
| Bandwidth | 1024 GB/s | 192 / 384 GB/s | 112 GB/s |
| Launch Date | 2017 | Q2 2016 | Q2 2016 |
Mind you that the specs shown in the CompuBench database might not be the full unlocked GPUs, so the CU numbers might even be higher. Which makes this news-item very speculative. We are inclined to say that Polaris 10 really would get 2560 shader processors.
Please click the thumbnail below to see the new roadmap slide.
New AMD roadmap shows Vega and Navi Architectures - 03/15/2016 08:28 AM
While the Polaris GPU architecture is to be released this year, two new architecture (or product series) code-names have surfaced in a roadmap shown in an AMD presentation, Vega and Navi....
Download the new AMD Catalyst 15.11.1 Beta Driver - 11/17/2015 01:42 PM
You can now download the new AMD Catalyst 15.11.1 Beta Driver, which contains quality and performance optimizations for the following titles: Star Wars: Battlefront Fallout 4 - Assassin's Creed Syn...
BIOSTAR Supports New AMD FM2+ Godavari APU - 05/11/2015 07:56 AM
With AMD's launch of the FM2+ "Godavari" class of APUs today, Biostar has been in preparation of a complete series of mainboards for the supporting chipsets such as the BIOSTAR Hi-Fi A88W...
MSI releases 8 new AMD Godavari ready FM2+ motherboards - 05/04/2015 09:16 AM
MSI launches 8 new AMD FM2+/FM2 socket based motherboards supporting the latest AMD Godavari APU. These new models are available in ATX, micro-ATX, and mini-ATX form factors and are backwards compatib...
New AMD GPUs and APUs after March - 01/22/2015 11:41 AM
That what she said .. ehm that would be the new CEO, Lisa Su. So AMD will release new GPUs and APUs in or after Q2 2015 "Going into the second quarter and the second half of the year with our ...
GeniusPr0
Senior Member
Posts: 1322
Joined: 2003-04-26
Senior Member
Posts: 1322
Joined: 2003-04-26
#5261490 Posted on: 04/22/2016 12:05 AM
you guys are behind
you guys are behind
Ieldra
Senior Member
Posts: 3490
Joined: 2007-01-27
Senior Member
Posts: 3490
Joined: 2007-01-27
#5261507 Posted on: 04/22/2016 12:55 AM
Nice rig
lol
Anyway. Stop saying "MHz don't matter", first of all it makes it seem like 'MHz' are a thing, secondly they do matter, they just don't completely describe the performance.
MHz is a unit of frequency, it tells you how many times in the cycle completes in one second.
Using the FP32 example; you have units that do floating point arithmetic, as of the last five years or so they implement the new FMA (fused multiply-add) instructions so you have 2 floating points ops per cycle.
You have 2816 floating point execution units on a 980Ti
So that's 2 every cycle on each of the 2816 units
1,500,000,000 cycles per second (1500mhz) --- (2 x 2816) x 1,500,000,000 = ~8,450,000,000,000 floating point operations per second.
You can also calculate other theoretical throughputs considering the frequency; for render output units for examples, or texture mapping units, the clock matters.
Obviously having 2x units at half the frequency is the same as half the units at double the frequency.
The Fury X has a FP32 performance of 8,600,000,000,000 flops
To match this performance a 980Ti needs to be clocked at (8,600,000,000,000)/(2x2816) = 1527 mhz
you guys are behind
Nice rig
lol
Anyway. Stop saying "MHz don't matter", first of all it makes it seem like 'MHz' are a thing, secondly they do matter, they just don't completely describe the performance.
MHz is a unit of frequency, it tells you how many times in the cycle completes in one second.
Using the FP32 example; you have units that do floating point arithmetic, as of the last five years or so they implement the new FMA (fused multiply-add) instructions so you have 2 floating points ops per cycle.
You have 2816 floating point execution units on a 980Ti
So that's 2 every cycle on each of the 2816 units
1,500,000,000 cycles per second (1500mhz) --- (2 x 2816) x 1,500,000,000 = ~8,450,000,000,000 floating point operations per second.
You can also calculate other theoretical throughputs considering the frequency; for render output units for examples, or texture mapping units, the clock matters.
Obviously having 2x units at half the frequency is the same as half the units at double the frequency.
The Fury X has a FP32 performance of 8,600,000,000,000 flops
To match this performance a 980Ti needs to be clocked at (8,600,000,000,000)/(2x2816) = 1527 mhz
GeniusPr0
Senior Member
Posts: 1322
Joined: 2003-04-26
Senior Member
Posts: 1322
Joined: 2003-04-26
#5261509 Posted on: 04/22/2016 12:58 AM
Fiji will sell along with Polaris. Just saying...
The performance per watt game is annoying since it depends on the application and settings in question to determine the power consumption. It's possible to remain very power efficient at a certain clock and explode with a small bit of tinkering. This is Maxwell II in a nutshell. I suspect Polaris is similar, and that's why AMD changed from up to 2.5x to up to 2.0x perf-per-watt. They had to increase clocks. Those with 290s and up probably have no reason to upgrade unless they want Pascal.
Nevertheless AMD confirmed its position with Polaris 10/11 for release in June. Hopefully it's not all that will be offered. These are essentially all notebook grade GPUs.
Fiji will sell along with Polaris. Just saying...
The performance per watt game is annoying since it depends on the application and settings in question to determine the power consumption. It's possible to remain very power efficient at a certain clock and explode with a small bit of tinkering. This is Maxwell II in a nutshell. I suspect Polaris is similar, and that's why AMD changed from up to 2.5x to up to 2.0x perf-per-watt. They had to increase clocks. Those with 290s and up probably have no reason to upgrade unless they want Pascal.
Nevertheless AMD confirmed its position with Polaris 10/11 for release in June. Hopefully it's not all that will be offered. These are essentially all notebook grade GPUs.
AMD demonstrated its “Polaris” 10 and 11 next-generation GPUs, with Polaris 11
targeting the notebook market and “Polaris” 10 aimed at the mainstream desktop and
high-end gaming notebook segment. “Polaris” architecture-based GPUs are expected to
deliver a 2x performance per watt improvement over current generation products and
are designed for intensive workloads including 4K video playback and virtual reality
(VR).
***61623; AMD continued to expand its leadership position in VR, unveiling new technologies and
collaborations across a variety of sectors, including gaming, education, and media.
o AMD introduced the Radeon™ Pro Duo GPU, part of the world's most powerful
platform for VR designed for creating and consuming VR content. AMD’s
Radeon™ Pro Duo GPU with its LiquidVR™ SDK is a platform aimed at most all
aspects of VR content creation: from entertainment to education, journalism,
medicine, and cinema.
o 20th Century Fox, New Regency, Ubisoft Motion Pictures, and VR development
studio Practical Magic chose the AMD Radeon™ Pro Duo GPU featuring the
AMD LiquidVR™ SDK to bring the upcoming ASSASSIN’S CREED movie VR
experience to life
targeting the notebook market and “Polaris” 10 aimed at the mainstream desktop and
high-end gaming notebook segment. “Polaris” architecture-based GPUs are expected to
deliver a 2x performance per watt improvement over current generation products and
are designed for intensive workloads including 4K video playback and virtual reality
(VR).
***61623; AMD continued to expand its leadership position in VR, unveiling new technologies and
collaborations across a variety of sectors, including gaming, education, and media.
o AMD introduced the Radeon™ Pro Duo GPU, part of the world's most powerful
platform for VR designed for creating and consuming VR content. AMD’s
Radeon™ Pro Duo GPU with its LiquidVR™ SDK is a platform aimed at most all
aspects of VR content creation: from entertainment to education, journalism,
medicine, and cinema.
o 20th Century Fox, New Regency, Ubisoft Motion Pictures, and VR development
studio Practical Magic chose the AMD Radeon™ Pro Duo GPU featuring the
AMD LiquidVR™ SDK to bring the upcoming ASSASSIN’S CREED movie VR
experience to life
Fox2232
Senior Member
Posts: 11809
Joined: 2012-07-20
Senior Member
Posts: 11809
Joined: 2012-07-20
#5261531 Posted on: 04/22/2016 01:54 AM
and that's why performance / (transistor*clock) is... a nonsense metric
look, performance is tangible metric, and so is transistor#
clock.. no one gives a **** about clock, because its significance is already accounted for in performance
what performance / (transistor*clock) metric does is... it punishes high-clocking arch/design. makes no sense.
Nonsense metric to many, not to me. It gives insight into architecture. Outlook for future.
When I had i7-720qm it had good performance per clock as 2nd and any generation after. But it was clocked too low to be great as it was 45nm chip.
Then 2nd generation came on 32nm and achievable clock speeds changed a lot.
Understanding technological (dis)advantages for each technology is important. If you happen to know that in 2 months some disadvantage will be gone...
and that's why performance / (transistor*clock) is... a nonsense metric
look, performance is tangible metric, and so is transistor#
clock.. no one gives a **** about clock, because its significance is already accounted for in performance
what performance / (transistor*clock) metric does is... it punishes high-clocking arch/design. makes no sense.
Nonsense metric to many, not to me. It gives insight into architecture. Outlook for future.
When I had i7-720qm it had good performance per clock as 2nd and any generation after. But it was clocked too low to be great as it was 45nm chip.
Then 2nd generation came on 32nm and achievable clock speeds changed a lot.
Understanding technological (dis)advantages for each technology is important. If you happen to know that in 2 months some disadvantage will be gone...
Click here to post a comment for this news story on the message forum.
Senior Member
Posts: 8325
Joined: 2008-07-31
Considering the fact that the Mhz don't matter and only the architecture, who cares?
I mean, your question is as valid as saying "back in the day":
"If a Pentium 4 can reach 3.73Ghz, why can't an Athlon 64?"
As though the fact that the Pentium 4 3.73Ghz stock actually somehow made it better then the Athlon 64
The point i'm trying to make is, there's no "Mhz" that's created equal, the "mhz" ultimately means nothing other then the speed at which the architecture is running, it only matters what the performance is, not what the Mhz is
^ Above reply to you too, since you seem to think that 800Mhz somehow means bad performance...not saying it WILL be good performance, but unless you know exactly how the new architecture(or updated) performs, then you can't make that call.
Right? I don't understand how people keep thinking this is how it works...