• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[Sweclockers] AMD Tonga will have the power consumption of 270X but replace the 280X

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I don't know what Nvidia is doing with SLI, and the fact that the individual cards themselves are so close in power consumption otherwise baffles me even more. Whatever it is they're doing though, it's working.

Well, Tom's has rated their power consumption as equal even in multiGPU. Ehh, actually, Crossfire seems a lot worse (the 295x2 is running much cooler, so it's consuming less power than a typical Crossfire setup would).

http://www.tomshardware.com/reviews/radeon-r9-295x2-review-benchmark-performance,3799-16.html
 
Last edited:
I can't speak for every review, but the reason for [H]'s low SLI power usage is because they warm the cards up and the 780 ti's are throttling due to thermals.
1390541180GVyEsgrEO9_10_1.gif


This is the system power with a single card. It's only 112W less. We know the 780 ti draws more than 112W.
1383802230J422mbwkoS_7_1.gif


Is this really right or am I reading it wrong? 176W avg. through the PCIe and only 47W throught the PEG? That doesn't make sense.
02-Power-Draw-Reference-Gaming.png
 
Haha I knew it. Tonga is NOT based on Tahiti.
Tonga is based on Hawaii!

Thanks to Computerbase we know a lot more of the specifications of FirePro W8100.

Shader count: Still unknown but atleast 2048 shaders (32CUs) based on R9 M295X leak
Shader clock: 917 Mhz (Not known if it have additional turbo boost)
Bus Width: 512bit
Memory clock: 1250MHz
Memory Bandwidth: 320GB/s

4_m.jpg


This also means that R9 M295X have the memory bus chopped in half and have 12CUs disabled.
Source:http://www.computerbase.de/2014-06/amd-firepro-w8100-hawaii-nicht-tonga-gpu/
 
Last edited:
Sorry, I don't speak German and Google translate butchers German so as not to be reliable. What makes you think that Tonga has a memory bus that's chopped in half and 12CU's disabled?
AMD apparently instead of using an actual Tonga GPU in W8100. Used the Hawaii GL40 ASIC which is the equivalent of the R9 290 graphics with full DP capabilities.
 
Haha I knew it. Tonga is NOT based on Tahiti.
Tonga is based on Hawaii!

Thanks to Computerbase we know a lot more of the specifications of FirePro W8100.

Shader count: Still unknown but atleast 2048 shaders (32CUs) based on R9 M295X leak
Shader clock: 917 Mhz (Not known if it have additional turbo boost)
Bus Width: 512bit
Memory clock: 1250MHz
Memory Bandwidth: 320GB/s

4_m.jpg


This also means that R9 M295X have the memory bus chopped in half and have 12CUs disabled.
Source:http://www.computerbase.de/2014-06/amd-firepro-w8100-hawaii-nicht-tonga-gpu/

"Based on hawaii" do you mean it's Hawaii cut down? Because that would make very little sense to use Hawaii in a mobile setup instead of Tahiti.
 
Sorry, I don't speak German and Google translate butchers German so as not to be reliable. What makes you think that Tonga has a memory bus that's chopped in half and 12CU's disabled?

Because
A) We already know the specs of R9 M295X. It have 32CUs based on leak from ozone3d.

B) We know now that Tonga is based on Hawaii and we know that Hawaii have 44CUs. Which means M295X have 12CUs disabled.

C) Memory bandwidth is chopped in half because MXM cards (mobile cards) have a practical restriction of 256bit. Hawaii have 512bit.

Unless there is two Tonga chips. One for desktop and one for mobile.

But Hawaii on a mobile chip is possible. Hawaii is big die, 423mm2, but mobile chips like GTX 480M was 526mm2.
Who knows though, FirePro W8100 can be Hawaii while R9 M295X is an entirely different chip, Tonga that isnt based on Hawaii.

Either the CodeXL dump is wrong, or both are based on Hawaii.
http://www.pcgameshardware.de/screenshots/811x455/2014/05/beyond3d-tonga-codexl-2-pcgh.jpg

We will see 🙂
 
Last edited:
Tonga, Hawaii...Tahiti. They are all boring now.
The market really needs something new honestly and I'm pretty certain it's not in a close future. 🙁
 
Average what over what? How are they measuring this average? Likely they take measurements at different points in time and they average them, but how many?


Instantaneously is the only way power can accurately be measured anyway. It appears you don't understand much about how measurement power works.

Per Ryan Smith's mouth:

It's the highest average value over a 5 second window (this way we're measuring full power consumption, but not transient peaks). That said our Crysis 3 test is rather consistent, so it will hold for well over 5 seconds.

-Ryan Smith

There you go. Average over time. 780 TI's consume way less power than 290x's in crossfire. HardOCP and every other website agrees with Anandtech and comes to the same conclusions.
 
Its been confirmed that FirePro W8100 is Hawaii but not Tonga like CodeXL says.
I think it stood there all along and I apparantly didnt read everything. 😀

Auf Nachfrage bestätigte AMD uns gegenüber, dass bei der FirePro W8100 die Hawaii- und nicht Tonga-GPU zum Einsatz kommen wird.
On demand AMD has confirmed to us that a Hawaii and not Tonga GPU will be used at the FirePro W8100.
 
Last edited:
They'll need to deliver it at 4gb for $250-$300 and it needs to be slightly (5-10%) faster than 280x@1ghz.

I don't see why a more efficient 280x makes any sense at 2gb and I don't think power savings is enough to get folks interested here without a price reduction.

Can't come soon enough, but if it comes at a price premium to the 280x and if it has a 2gb/4gb price stupidity arrangement then i'm going to be disappointed.

Hopefully this will prompt nVidia to get more exciting in the $200-$400 space where they have been garbage for awhile now.
 
Last edited:
The mention of FirePro W8100 for Tonga might have been a typo. In which case it could be FirePro M8100 instead of FirePro W8100.

So, AMD will have the;
Radeon R9 M295X for Enthusiast Gaming Notebooks.
FirePro M8100 for Mobile Workstations.
Radeon R9 2x5 for Performance Gaming Desktops.
 
Last edited:
Tahiti debuted in 2011. Would be very, very good to see a VI part debuting this year at a so crucial price point(is 280x performance in 270x class, no?).
 
I fail to see the point in bringing new cards to the table if they are only to just out due the card that they are replacing for around the same price,
to me all they are doing is milking the general public of the money,
at the end of the day people want an improvement to the point of 30% to 40% on every new version of card not 5% or 10% thats crap.
 
I fail to see the point in bringing new cards to the table if they are only to just out due the card that they are replacing for around the same price,
to me all they are doing is milking the general public of the money,
at the end of the day people want an improvement to the point of 30% to 40% on every new version of card not 5% or 10% thats crap.

Exactly, they are milking the market and don't/can't have new architectures or GPUs for the moment because of TSMC.
 
For me, doing more with less is better, that's why i prefer new architectures showing up now at 28nm.
Would be so cool seeing "Maxwell effect" in a 280x chip consuming 270x power with no node shrink.
 
For me, doing more with less is better, that's why i prefer new architectures showing up now at 28nm.
Would be so cool seeing "Maxwell effect" in a 280x chip consuming 270x power with no node shrink.

Not much difference in consumption between a 280x and 270x.
Both AMD and nVidia have nothing good to release.

58715.png
 
Not much difference in consumption between a 280x and 270x.
Both AMD and nVidia have nothing good to release.

58715.png

50-75W is a pretty big difference in my book. You're forgetting that that graph is total power consumption, not just the GPU. If isolating the GPU power useage were possible, the difference percentage-wise would be pretty high. (~25-35%) If that's not impressive to you, what would be?
 
Last edited:
50W is a pretty big difference in my book. You're forgetting that that graph is total power consumption, not just the GPU. If isolating the GPU power useage were possible, the difference percentage-wise would be pretty high. (~25%) If that's not impressive to you, what would be? 50%?


You are right :colbert:.

But honestly, what we need right now is MORE POWAH.

I personally don't care about efficiency and there are already many efficient cards on the market to satisfy everyone.

50 Watts =
2860_835b648e28cd56859500dcea222694bccf44a4ad_original_x_323_1337885093.jpg

________
(This bulb must also be 0 watt on idle 😉 )
 
Back
Top