Tuesday, May 26th 2015
NVIDIA GeForce GTX 980 Ti Clock Speeds Revealed
NVIDIA's upcoming GeForce GTX 980 Ti graphics card is shaping up to be the "almost Titan-X for two-thirds its price" product the company wants, out in the market. A leaked GPU-Z screenshot of the card by Korean tech-publication HardwareBattle (the same site that broke the card's core config,) reveals its reference clock speeds. All the values displayed by GPU-Z 0.8.2 in the screenshot are pulled from the system, and not an internal lookup table (all the LUT-based values are grayed out, because version 0.8.2 lacks those values for the GTX 980 Ti). The card offers clock speeds that are similar to those of the GTX Titan-X. The core is clocked at 1000 MHz, with a maximum GPU Boost frequency of 1076 MHz (1089 MHz on the GTX Titan-X), while the memory ticks at 7012 MHz (GDDR5-effective).
From our older article, it's known that the GTX 980 Ti will feature a lower CUDA core count, at 2,816 cores, compared to 3,072 on the GTX Titan-X. The TMU count is proportionately lower, at 176. The ROP count is a bigger mystery than Nessie. The card features 6 GB of GDDR5 memory, across a 384-bit wide memory interface. While the reference board design is something that's beginning to look dated, NVIDIA will allow its AIC (add-in card) partners to come up with custom-design boards factory-overclocked to Kingdom come, from day-one. The GeForce GTX 980 Ti is expected to be launched on the sidelines of Computex 2015, in the first week of June.
Sources:
HardwareBattle, VideoCardz
From our older article, it's known that the GTX 980 Ti will feature a lower CUDA core count, at 2,816 cores, compared to 3,072 on the GTX Titan-X. The TMU count is proportionately lower, at 176. The ROP count is a bigger mystery than Nessie. The card features 6 GB of GDDR5 memory, across a 384-bit wide memory interface. While the reference board design is something that's beginning to look dated, NVIDIA will allow its AIC (add-in card) partners to come up with custom-design boards factory-overclocked to Kingdom come, from day-one. The GeForce GTX 980 Ti is expected to be launched on the sidelines of Computex 2015, in the first week of June.
75 Comments on NVIDIA GeForce GTX 980 Ti Clock Speeds Revealed
Tesla 65nm
Fermi 40nm
Kepler 28nm
Maxwell 28nm
We didn't get a die shrink with Maxwell and so there wasn't as big a leap in performance due to the chip being more efficient so that we could have more performance per watt from that alone as we are used to. Maxwell is actually a damned good architecture when you take this into account for the performance that it does give on the same process as Kepler.
If I recall correctly the first RMA was very shortly after purchase but that was for many 8800GS cards (possibly not just eVGA labeled). The replacement was new in retail packaging and had a slightly different cooler / fan design (which suggest that they had a problem with the initial design).
As for drivers though, I've had more problems with drivers on ATI cards then nVidia cards. I had to RMA an AIW 1900X card once but that was about it.
Edit:
Like I said before though, I have no grudges either way and I will buy ATI and nVidia cards to this day. If anything I wish there was a third or fourth manufacturer bringing competing GPU's to the market as well.
How does everyone know firsthand how good EVGA's customer support is? They must have A LOT of cards break.
But yeah I will also agree that in my experience AMD cards are just hardier overclockers overall.
560 Ti (Me): Broke twice in 8 months
560 Ti (Brother): Broken SLI flicker
460 SE (Friend 1): Broke in 8 months
650 Ti (Friend 1): Broke within 6 months
560 Ti (Friend 2): Broke twice in 4 months
560 Ti (Friend 3): Still working
We all switched and you can see why. We never abused these cards AT ALL. Meanwhile we have had 5750, 6870, 6850's, 7970's, many 7950's, and 6450 that have never had any problems no matter how much we overclock them.
P.S. Any of my friends with non-EVGA nvidia cards had zero issues as far as I know.
260 SSC
580 SC
780ti Classified Hydrocopper (current)
780ti Classified (custom block - also hydrocopper).
Both 780ti's massively overclocked above 1350Mhz on custom BIOS at 1.35v (PSU overloaded and shut down). First card actually rocked up to 1386Mhz at 1.37v before sli.
A lot of folk rubbish brands but it's really an availability heuristic. I've never had a card fail my most recent AMD were 5850's and 7970's. They didn't fail either. I'm sure you could browse selectively but this is the first thing i came across:
www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/#NVIDIAFailureRates
No mention of brands but the generic is that AMD of late are actually worse.
As we never are provided the actual types of failures it's odd to speculate it is from the AMD chips themselves as I'm sure both AMD and Asus would have dyno/check/quality to sort prior to surface mounting to the PCB. Puget Systems only in-house testing found more immediate shortcomings, 5 min Furmark probably being the worst, and one stress test that of late which is recognized by Nvidia card and limits power.
Actually Asus was known for issue with its DirectCU implementation for the first Hawaii production, and this perhaps being reported, plainly cooling or Asus PCB quality.
Puget Systems comes across as "advertising" it is more AMD reference card condition although one should see it as an Asus shortcoming, almost to the point it smelled as some sort of "Tier 0" stunt from the time period.
Anyway, 980Ti looks set to deliver good performance, as long as the 6GB cap is minded.
550ti sli (me) bought a week after release, still work to this day.
550ti sli 2GB (friend 1) both work, were just upgraded
geforce 770 sli (friend 2) running on 1200 mhz OC for a year, plenty of time in folding@home, both still work great
geforce 660 (friend 2) still running, used as folding@home server
geforce 580 (friend 3) was still working as of a year ago, when it was sold.
geforce 780 (friend 3) still working
only the 550ti gpus were reference models, the rest were custom PCBs. reference high end models seem to have issues, but those 550tis were just tanks.
I myself went with pny's 3 fan 770s, wonderful cards, but would have gone with evga if these didnt exist.
videocardz.com/55659/leaked-driver-confirms-hawaii-rebrands-in-radeon-300-series
This ties in with Fiji having a new brand name, maybe true, maybe not, either way OMG release something new already AMD.
I'll go by reviews that come out and then choose what I want to get.
Same story, Guru3D twist:
www.guru3d.com/news-story/amd-radeon-r9-390r9-380r9-370-and-r9-360-series-rebrands.html