AMD: Speculation, Rumors, and Discussion (Archive)

Status
Not open for further replies.
3dilettante said:
The improved single thread performance is a sub-item of the larger instruction buffer item. My interpretation is that the larger bullet point gives a feature, and then the indented items give the effect or a detail about it.
Click to expand...
Maybe, but an independent command stream was one of the requirements for the flexible scalar. That would entail increasing the instruction buffer, by one "per wave", instead of using the first lane for wave level operations. It's possible the ALUs are 4x(16+1)+1. That could give each ALU/executing wave a scalar to prefetch and run wave level ops in addition to a dedicated scalar for the CU. Might also explain part of the transistor disparity and why we aren't seeing more CUs. It seems odd they made all those slides and didn't mention how many processors were in there. Quantities on ACEs, HWS, TMUs, ROPS, etc. Everything but the actual processor count, which is the big number everyone likes to show.
 
fellix said:
And what's with the dedicated L/S units -- 16 per CU?
Click to expand...
Since the number of texture units do not change, it seems to be unchanged. Since GCN 1.0, it is capable of loading sixteen consecutive 32-bit (or four 128-bit) values from the L1 cache per clock, with the texture & data conversion unit sitting on the return data path.
 
Razor1 said:

near the end of the video they have furmark power consumption: its at 170 watts..... 3:48 mark

Its furmark but still.
Click to expand...
If we look at this https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/21.html, basically 480 furmark number is close to 270X and 760. With better (new) power management, I think on typical gaming situation 480 can be a bit lower than 270X. I can't find 1080 number since Nvidia throttle their card to never exceed their TDP(?), thus on game and on furmark, 1080 numbers are the same http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-10.html
 
Orion said:
Hmmm, rumors of 480 perf at 390x levels

http://www.guru3d.com/articles_pages/ashes_of_singularity_directx_12_benchmark_ii_review,7.html

That puts 390x about 30% higher than 980ti at crazy settings at 1440p.

The average gtx 1080 gap to 980 ti in dx12 is 30% higher performance, although some sites say it's about 20%.

If so this would suggest excellent performance in this app at these settings.
Click to expand...

I guess performance magically jumped from vanilla 290 to 390X in the past month. How else do you explain 1.83 CF scaling and 6% ahead of 1080 a month ago.
 
Jawed said:
So this guy appears to have RX 480:

http://www.overclock.net/t/1603915/coming-soon-unofficial-polaris-owners-thread/40#post_25297979

The GPU-Z screenie shows revision "C7". I don't remember ever having seen a revision that far. So, I had a rummage and found Fury X is revision C8:

http://forums.anandtech.com/showthread.php?t=2441548

I can't help thinking Polaris has had a troubled birth.

So, is that revision meaningful?
Click to expand...
SimBy said:
Comments?

"Wow, base metal layer had 3 spins and there were 7 minor spins after that at least... They have been working their behinds off to get this launched."

EDIT: Ah Jawed picked it up already.
Click to expand...
It's not GPU revision, it's used by AMD to differentiate SKUs which use same PCI device ID
 
BRiT said:
No reason why the users couldn't moderate things themselves and choose to not post in stale threads.

Anyways, here it is, here is the one single thread for all things AMD. Enjoy.
Click to expand...
All it took is one post to trigger you, grow a thicker skin. And while you're at it merge all the Nvidia threads as well, maybe merge ALL threads into one megathread to keep it interesting :LOL: This forum is really going downhill.
 
This single thread for all AMD is silly. All architectures all cards in the same thread? Impossible to find any info for the things you are specifically interested in.
Please split these threads back out into some sort of usable manner. Per architecture at least.
I also like the old "speculation thread, production thread" scheme.
 
Have a vote to determine final thread setup ... until then mimic the GTX 1080/1070 thread setup. I'd agree having multiple thread fractures/spinoffs does tend to dilute and sometimes shorten the lifespan of threads.
 
Last edited by a moderator:
Rurouni said:
If we look at this https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/21.html, basically 480 furmark number is close to 270X and 760. With better (new) power management, I think on typical gaming situation 480 can be a bit lower than 270X. I can't find 1080 number since Nvidia throttle their card to never exceed their TDP(?), thus on game and on furmark, 1080 numbers are the same http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-10.html
Click to expand...


Take 20% off the power usage, and its still very high lol.
 
Razor1 said:
Take 20% off the power usage, and its still very high lol.
Click to expand...
Why only take 20%? 270X on typical gaming load is 119W according to techpowerup while on furmark it is 173W. 285 on typical gaming load is 184w, but on furmark it is 250w. 980ti exhibited a greater than 20% difference between typical gaming load vs furmark. Basically what I mean is just wait for the review.
 
AMD Radeon RX 480 Video Reviews:
Thx Vcardz.
 
Last edited:
Status
Not open for further replies.

Similar threads

hoom
AMD: Zen 4, Speculation, Rumours and Discussion
8 9 10
Replies
199
Views
39K
MfA
M
Rys
Reopening Architecture and Products + rules and guidelines changes
3 4 5
Replies
90
Views
16K
Shifty Geezer
Shifty Geezer
D
AMD: Zen 2 (Ryzen/Threadripper 3000?, Epyc 8000?) Speculation, Rumours and Discussion
16 17 18
Replies
340
Views
153K
PSman1700
P
Shortbread
Next-Generation NVMe SSD and I/O Technology [PC, PS5, XBSX|S]
105 106 107
Replies
2K
Views
217K
Karamazov
K
L
A rambling history of AMD and nVidia's rise to the top
2
Replies
20
Views
6K
Leovinus
L
Back
Top