-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Nvidia Titan
- Thread starter Boulard83
- Start date
OP
- Joined
- Jan 19, 2009
- Thread Starter
- #62
Yeah but the GK110, if all this is just 75% real, is a huge jump from the GTX680 GK104. In fact, Nvidia never released the GK100 ... Remind me the GF100 failure with the uber power usage and heat. nvidia skipped on the GK100 release and waited to make sure the 28nm GPU was mature enough to hit the shelf.
- Joined
- May 10, 2009
How huge really depends on clocks. The 480 to 580 jump was ~23% (half from higher clocks, half from more cores), that's one of the higher same-process same-architecture jumps out there. It maintained the same price point, too.
If the clocks stay the same from GK104 to GK110 we're looking at 75% more performance per clock (if the K20X is a full GK110 die, anyway) absolute maximum, more likely less due to overhead. I seriously doubt the clocks will stay the same, given the Tesla cards TDP and clock speeds. The K20X has 2688 shaders at 732MHz and 235w TDP. Given the reaction of the small process sizes to higher voltage and clocks, I don't see the clock speed getting all that much higher even if the TDP is would up to 250-260w, especially given that the Tesla cards a low leakage bins.
If it stays at that clock speed, we're looking at 74.4% more shaders running 37% slower. (Or if you prefer looking at it the other way around, at 72% the speed)
Faster? Yes. OMGOMGOMGOMG faster? Not really. Probably about 25%, if it's run at the 732MHz base with similar boosting to the 680.
They did exactly the same thing with the 480 and 580, the 480 was a cut down die, the full die went into GPGPU boards for a while until they had a large enough pile to put them in consumer cards.
If the clocks stay the same from GK104 to GK110 we're looking at 75% more performance per clock (if the K20X is a full GK110 die, anyway) absolute maximum, more likely less due to overhead. I seriously doubt the clocks will stay the same, given the Tesla cards TDP and clock speeds. The K20X has 2688 shaders at 732MHz and 235w TDP. Given the reaction of the small process sizes to higher voltage and clocks, I don't see the clock speed getting all that much higher even if the TDP is would up to 250-260w, especially given that the Tesla cards a low leakage bins.
If it stays at that clock speed, we're looking at 74.4% more shaders running 37% slower. (Or if you prefer looking at it the other way around, at 72% the speed)
Faster? Yes. OMGOMGOMGOMG faster? Not really. Probably about 25%, if it's run at the 732MHz base with similar boosting to the 680.
They did exactly the same thing with the 480 and 580, the 480 was a cut down die, the full die went into GPGPU boards for a while until they had a large enough pile to put them in consumer cards.
- Joined
- Feb 17, 2010
- Location
- Manassas, VA
The HD8xxx series wont be that much more powerfull than the 7xxx. HD8000 series are a refresh, not a new genI didnt read anywhere anything about a big performance gain from the HD8000 series
real nextgen AMD and Nvidia cards are for the end of 2013.
HD8000 OEM is a refresh. HD8000 retail is a new gen.
- Joined
- Oct 14, 2007
- Joined
- Oct 29, 2012
Will this do anything else except improve tri-monitor performance?
I almost feel like high clock speeds are here to stay. They'll probably find a way to get these 780s to ship in the 800-900mhz range.
Then again, they want to continue making sure tesla is appealing to its customer base. Meh.
OP
- Joined
- Jan 19, 2009
- Thread Starter
- #68
OP
- Joined
- Jan 19, 2009
- Thread Starter
- #70
Dude, you run a tri monitor setup with a single 670... you need another or a 680 or 7970...
I currently play BF3 most of the time. Running at ULTRA no MSAA @ 50-60fps. not bad for a single card
I'm waiting for a better performing single card, wich the 680 isnt really ... you're still capped with the 256bit memory bus. I dont really like AMD CCC and Eyefinity. I'm not a 100% Nvidia Fanboy but when it comes to surround/eyefinity ... i prefer Nvidia drivers.
Edit : Its been more than 2 years that i run surround/eyefinity setup. I think that i know what it need and what it is to run this kinda setup.
Last edited:
- Joined
- Feb 26, 2003
Right.. but this is all we have... let us drool in peace...and stop drinking the Haterade!
Just thinking about history, how many cards come out showing an awesome 3dmark score to only perform like pure crap in games...
Not many(any)? I cant think of one since 2004 when I started benchmarking and gaming on PC's
. Of course that doesnt mean there werent any, however, it feels like you are alluding to this be a common thing, and I disagree with that sentiment. Its never a direct comparison but its plenty 'good enough' to judge, especially since 3DMark Vantage and forward. Its not like it scores high in 3d11 and then cant play modern games at full tilt at 1080p.
Last edited:
- Joined
- May 10, 2009
Such as?
- Joined
- Oct 29, 2012
I'm also curious. I've seen cards not perform AS WELL as they SHOULD in regards to their score, but never "pure crap" as quoted.
Especially if the scores end up really coming in at this "holy flippin' hell" level
Similar threads
