The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
Grall said:
Newer games are likely coded in a way that suits scalar GPUs. You would therefore see a proportionally lower utilization on the more rigid, wonkier VLIW-era products, and thus lower performance as well.
Click to expand...

what surprises me is the variation from one game to the other, considering they are all targeting the same hardware

Max Max which is based on the same engine from the Just Cause game
http://pclab.pl/art65638-10.html
behind the 260X not by much, while on Just Cause 3 it's under half.
similar results for the newest Assassins Creed I posted earlier

Metal Gear 5,
http://pclab.pl/art65645-10.html
clearly ahead of the 260x

while Wticher 3 on low settings (looking at Miasto City)
May 2015
260X 30-34.2 (min-avg)
6970 27-32.1
http://pclab.pl/art63116-7.html

October 2015
260x 29-35.1
6970 18-21.6
http://pclab.pl/art66374-5.html

seems like the same test scene and settings, looking at the 260X also the same performance, but the 6970 dropped performance badly with the latest updates and drivers!?
 
Yeah, something seems wrong with those 6970 numbers in Witcher 3.
Regardless, the 6970 should still be good for most games at 1080p and medium/high settings.
 
Davros said:
Possibly more bad news for amd
It looks like AMD will be hitting a brick wall with the sales of its Radeon R9 Fury X, as Asetek has sent both AMD and GIGABYTE cease and desist letters over the watercooling on their video cards. GIGABYTE's GTX 980 Waterforce cards are also set to see sales stopped.
http://www.tweaktown.com/news/48808/asetek-demands-suspension-sales-amds-radeon-r9-fury/index.html
Click to expand...
AMD actually responded to that already, that the jury didn't actually specify Fury X's cooler as infringing one (as it doesn't use any "official cooler" by cooler master), but that they'll defer to cooler master on the details
http://www.gamersnexus.net/industry/2220-amd-response-to-asetek-cnd-on-fury-x
“We are aware that Asetek has sued Cooler Master. While we defer to Cooler Master regarding the details of the litigation, we understand that the jury in that case did not find that the Cooler Master heat sink currently used with the Radeon Fury X infringed any of Asetek’s patents.”
Click to expand...
 
I doubt this particular SKU is in any way deterministic for the bottom line of the graphics division anyway. AMD could easily handover the design to the board partners to deal with the issue the best way they could find.
 
Either TechRadar got very confused with their latest mini-interview with Lisa Su, or there are Zen APUs coming out this year after all:
Rolling graphics and compute capabilities into a single solution, the APUs will feature AMD's Zen processors that will be manufactured using the 14nm FinFET process, rather than the 28nm process AMD has used since 2011. Laptops featuring the upcoming APUs are scheduled to debut later in the year.
Click to expand...

http://www.techradar.com/news/mobil...wer-play-to-win-over-the-laptop-elite-1313138
 
Su possibly misspoke, or there was a quote mangling somewhere. AMD's 7th gen APUs are not FinFET or Zen.

In other news, AMD's Seattle chip has been launched. Notably, it is being linked to a server fabric--the former Calxeda's fabric.
 
Or it's reference to the rumored HPC APU, which is supposedly Zen-based, and some rumors suggest it COULD emerge late this year. Mainstream APUs are another matter, of course
 
Well Bristol Ridge was supposed to be available in consumer products Q2 2016 or early Q3. I think if that chip was coming within that timeframe, we should have seen Bristol Ridge laptops at CES.. but we didn't.

Maybe Bristol Ridge was canned and Zen APUs were pushed forward? That's the unlikeliest possibility coming from AMD, though.

Kaotik said:
Or it's reference to the rumored HPC APU
Click to expand...
The sentences clearly point to laptop chips though...
 
Are there any articles on AMD's Seattle launch that mention the integrated Freedom Fabric hardware that was silently dropped from the feature set and has now been replaced by another fabric?
There were some rather evasive statements that the IP was available in the future, but with AMD going even further and using Silver Lining's Calxeda interconnect is this not the complete repudiation of the SeaMicro purchase it seems to be?
(edit: on top of the 99% repudiation that has already transpired)
 
ToTTenTranz said:
Well Bristol Ridge was supposed to be available in consumer products Q2 2016 or early Q3. I think if that chip was coming within that timeframe, we should have seen Bristol Ridge laptops at CES.. but we didn't.

Maybe Bristol Ridge was canned and Zen APUs were pushed forward? That's the unlikeliest possibility coming from AMD, though.
Click to expand...
Bristol Ridge is supposed to come to desktop first I think, laptop later (and considering laptops already have Carrizo, which is same as Bristol Ridge except for disabled DDR4-support, there's no similar rush for it as there is for desktop which is still stuck at Kaveri/Godavari)
 
Why didn't AMD use Jaguar/Puma cores in this Seattle chip?

And does Seattle actually have a segment to be successful in? Where does something like this get used in volume?
 
swaaye said:
Why didn't AMD use Jaguar/Puma cores in this Seattle chip?
Click to expand...

The point of it was to make an ARM-based server chip to get AMD into the market of ARM-based servers. Project Skybridge would have had an ARM and x86 cores that would fit into the same surrounding chip infrastructure, but that was cancelled.

Maybe Jaguar/Puma at 28nm wouldn't really differentiate itself enough from the Opterons. It would likely be slaughtered by the x86 competition, and the consoles have shown that Jaguar's module priorities lie elsewhere than scaling beyond 4 cores. ARM's infrastructure seems to have at least included that.
 
Yeah I remember that x86+ARM initiative now. I guess K12 is still in the works as well.

I wasn't aware of the 8-core Jaguar consoles having efficiency issues. Cortex A57 doesn't sound especially thrilling at this point either though.
 
Dont forget that this ARM chip was announced 3 years ago... Outside dont burn money for nothing, i dont really see why they realize it now.. The real deal will be Zen x86 and ARM one.
 
swaaye said:
Yeah I remember that x86+ARM initiative now. I guess K12 is still in the works as well.

I wasn't aware of the 8-core Jaguar consoles having efficiency issues. Cortex A57 doesn't sound especially thrilling at this point either though.
Click to expand...

The 2-module consoles have significant latency penalties associated with crossing between the L2 caches on-chip. Seattle shows an L3 and a different arrangement of the caches on-die. It's different, and the L3 at least puts forward the possibility that software wouldn't be actively discouraged from sharing data on-chip like it is for the consoles.
Another item that I forgot to mention is that at some point AMD had already lost its ability to design in Jaguar's space. Zen/K12 sucked up resources, and the Jaguar team was mostly gone. Perhaps a 28nm Jaguar-based Seattle had stopped being an option.

As was mentioned, this chip had a very prolonged development period, so it was locked into A57 years ago. Seattle also lost the onboard fabric inherited from the SeaMicro purchase, although unless this is a new die it's probably still in the chip. That fabric was likely more important to the product distinguishing itself than the A57, which was not a leading core even years ago in the server space.
 
AMD's 4Q2015 earnings (losses) are in:

AMD has turned in its financial results for the fourth quarter of 2015. The company posted an operating loss of $49 million and a net loss of $102 million on $958 million in revenue.

http://techreport.com/news/29621/amd-narrows-its-losses-in-the-fourth-quarter-of-2015

For now, though, AMD projects a 14% sequential decrease in revenue, plus or minus 3%, in the 13-week first quarter of this year.
Click to expand...

The forecast for 2016Q1 means a greater loss than this quarter.
 
Status
Not open for further replies.

Similar threads

F
Dedicated gaming GPUs bottom of the priority list?
2
Replies
34
Views
3K
Seanspeed
S
F
RDNA5 Speculation
Replies
6
Views
2K
Kaotik
Kaotik
T
Re-litigating Radeon market performance *spawn
2 3
Replies
40
Views
3K
Dictator
D
D
Intel, AMD, Microsoft & Others Team Up For Develop The “UALink”, Direct Competitor To NVIDIA NVLink
Replies
3
Views
910
pcchen
P
xpea
Datacenter GPU market supply and demand (2024)
2
Replies
28
Views
3K
Erinyes
E
Back
Top