Overclock.net banner
Status
Not open for further replies.

[WCCF] GTX 980 Details Revealed 64 ROPs, HDMI 2.0, 2048 CUDA Cores and 5.2 Billion Transistors

25K views 303 replies 123 participants last post by  Alatar 
#1 ·
Source

Quote:
NVIDIA GeForce GTX 980 Graphics Card and GM204 GPU Detailed - 64 ROPs, HDMI 2.0, 2048 CUDA Cores and 5.2 Billion Transistors Operating At 165W
Quote:
So the final specifications of the GeForce GTX 980 include 2048 CUDA Cores, 128 TMUs, 64 ROPs. The core clock is maintained at 1126 MHz core and 1216 MHz boost while the memory is clocked in at 7 GHz effective clock which results in 224 GB/s bandwidth which might be enough thanks to the lower bandwidth dependency and increased efficiency. The TDP of the card is set at 165W while the power is fed through dual 6-Pin power connectors.
 
See less See more
1
  • Rep+
Reactions: keikei
#2 ·
Dat DP 1.3 tho...
frown.gif
 
#6 ·
Quote:
Originally Posted by Mand12 View Post

DP 1.3 was released yesterday. How can you reasonably expect it to be implemented?
That's the joke...
 
#11 ·
Quote:
Originally Posted by Swolern View Post

Man that 980 is clocked high, wondering about OC potential.....

And that memory bus.
mad.gif
As far as OC potential, keep in mind how low the TDP is, and how efficient Maxwell is. Push up to your actual power ceiling and you'll get a rather large performance increase.

And again, memory bus isn't anywhere close to as important as people think it is. There are lots of things that affect memory bandwidth and how it's used.
 
#12 ·
Quote:
Originally Posted by Mand12 View Post

Sorry. I'm expecting people to complain about that seriously.
Haha, no doubt some will. I personally expect it to be at least a year until DP 1.3 gets onto hardware. Same as HDMI 2.0 was ratified September of last year and just now made it onto these cards a year later.
 
#13 ·
..and it was a wise decision buying the 780 at 450€ last year, proven serveral times.
The GPU market isn't attractive at all.
 
#14 ·
Quote:
Originally Posted by omari79 View Post

a TDP of just 175W? o_O
My left eyebrow also went up with that statement.
 
#15 ·
Quote:
Originally Posted by Mand12 View Post

As far as OC potential, keep in mind how low the TDP is, and how efficient Maxwell is. Push up to your actual power ceiling and you'll get a rather large performance increase.

And again, memory bus isn't anywhere close to as important as people think it is. There are lots of things that affect memory bandwidth and how it's used.
If your pushing 7680x1440 it is very important.
 
#16 ·
Quote:
Originally Posted by Swolern View Post

If your pushing 7680x1440 it is very important.
Bus width is not the same thing as bandwidth, and even bandwidth is not the end of the story. How that bandwidth is utilized matters tremendously.

You can't just boil it down to 256-bit and claim it sucks. For reference:

GTX 280: 512 bit
GTX 480: 384 bit
GTX 580: 384 bit
GTX 680: 256 bit
GTX 780: 384 bit

Your logic would indicate that the 780 is inferior to the 280 for running 7680x1440.
 
#17 ·
Bus width is the new GHz myth. Probably even worse.

1) Bus width alone is pretty much irrelevant, bandwidth is the metric you should be looking at if you want to focus on memory in general.
2) Different architectures use bandwidth differently. A look at the recent Tonga release slides should tell you this.

In the end it's about performance. Smaller bus is preferable whenever possible due to power, cost, etc. Both Tonga and maxwell chips have improvements that aid with lower memory bandwidth. And this card will also give us a look at how NV cards with 64 ROPs will perform.
Quote:
Other interesting technology is Third Generation Delta Color Compression. This technology will help increase memory efficiency, which is definitely required with 256b interface.
http://videocardz.com/52362/only-at-vc-nvidia-geforce-gtx-980-final-specifications

Now imagine a GM200 card with that, a 384bit bus, 96 ROPs and around 3K cuda cores with sub 250W TDP and overclocking headroom for days. The 980 is all fine and good but the technology behind it (and the future cards based on the tech) is what really should get people excited.
 
#19 ·
Quote:
Originally Posted by th3illusiveman View Post

So is it 64 ROPS or 32? $599 or $499?
rolleyes.gif
64.

And price is unknown, lots of guesses, no hard confirmations.

Also some well known extreme OCers are saying 2300MHz on LN2.
 
#20 ·
Quote:
Originally Posted by Alatar View Post

Now imagine a GM200 card with that, a 384bit bus, 96 ROPs and around 3K cuda cores with sub 250W TDP and overclocking headroom for days. The 980 is all fine and good but the technology behind it (and the future cards based on the tech) is what really should get people excited.
Exactly, I am waiting for 980Ti... etc before selling my 780.
 
#23 ·
Quote:
Originally Posted by Alatar View Post

Bus width is the new GHz myth. Probably even worse.

1) Bus width alone is pretty much irrelevant, bandwidth is the metric you should be looking at if you want to focus on memory in general.
2) Different architectures use bandwidth differently. A look at the recent Tonga release slides should tell you this.

In the end it's about performance. Smaller bus is preferable whenever possible due to power, cost, etc. Both Tonga and maxwell chips have improvements that aid with lower memory bandwidth. And this card will also give us a look at how NV cards with 64 ROPs will perform.
http://videocardz.com/52362/only-at-vc-nvidia-geforce-gtx-980-final-specifications

Now imagine a GM200 card with that, a 384bit bus, 96 ROPs and around 3K cuda cores with sub 250W TDP and overclocking headroom for days. The 980 is all fine and good but the technology behind it (and the future cards based on the tech) is what really should get people excited.
^^This. Dropping them knowledge bombs like always.
thumb.gif


I was wondering what GM200 would have in terms of ROP count if GM204 has 64. 96 sounds about right.
 
Status
Not open for further replies.
You have insufficient privileges to reply here.
Top