Associate
You probably don't need them, no (for temps anyway)
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
**Nvidia GeForce GTX 970 / 980 Review Thread**
- Thread starter Boomstick777
- Start date
You probably don't need them, no (for temps anyway)
Associate
- Joined
- 20 Oct 2010
- Posts
- 1,194
I was expecting a bit more from the 980 as a new current flagship product, although from the specifications i was kind of expecting exactly what we saw today. AMD lets see what you have to bring to the market!
Associate
can we now begin speculating how this performance will scale to the Titan replacement?
Associate
- Joined
- 24 Oct 2013
- Posts
- 7
Don't worry there will be plenty along who will.
Here is a few to get on with.
AMD are not exactly going to fold and in recent times have shown an improvement.
You can not compare shaders on different architectures, NVidia and AMD designs are totally different.
The 980 is the successor for the GTX 770.
When AMD launched the 290X the last thing they were expecting was for NVidia to produce the 780ti, it caught them totally wrong footed.
The GM204 could be scaled up nearer to 100% on 20nm for full fat Maxwell, the GK110 chips already use nearly 7 billion transistors @28nm.
There is no full Tonga (285X) on the way and apparently there never was so LtMatt tells us.
The 980 is not 75% faster than a 285.
There are a few to start with.
Man I feel like I have to teach u to.
AMD will not fold but this puts them in more danger than you think. AMD currently has 940 billion in the cash and cash equivalents. They have 2 billion in debt. The problem with this is AMD needs 1 billion in cash to survive. This is to keep the creditors at bay and to pay for day to day operations. Since AMD is below this threshold, any further losses are dangerous. The counter to this is AMD got that console deal, that will save them. The problem with this is their debt interest is 49 million a quarter which wipes out most of this. So as a result, they lost 36 million this quarter and 20 million the last quarter even with the console deal going on. So what is a price cut between 150 dollars to 50 dollars going to do their margins and profits? I think you can fill in the rest.
Of course you can't compare shaders and such. I just used that to show that they are really in the same midrange chip class. The most important indicators are the power consumption(application use) and the die size(cost to manufacture chip) for chip class and how much the architecture can be scaled up. What die size and power consumption tell us is how much u can scale up an architecture. Small size = how many components you can put on the chip, power consumption ties to both number of gpu components and clocks.
If we look at the power consumption of the first maxwell chip, gtx 750 ti, it's the only thing near the performance/ energy efficiency of the gtx 980/970. That's because the more efficient an architecture is the more it can be scaled up without having to compromises clocks.
The gtx 770 and gtx 680 are the same chip.
Both of the chips are gk104 chips. The gtx 770 just has a better stepping which allow it to reach slightly higher clocks. But they are the same. You can even flash a gtx 770 into a gtx 680.Everyone knew the gtx 780 ti was coming out. The first Gk110 products were neutered products and it was only a matter of time and a new stepping(which improves yields) for the full gk110 chip to come out. Nvidia did the same thing with the gtx 480 and gtx 580. It didn't surprise AMD as it didn't surprise most people.
GM200 is going to be produced at 28nm and so will Fiji. Hence why AMD is using a hybrid water cooler to cool the chip(it wouldn't need this at 20nm unless they went super big). Producing a big monolithic chip like fiji or big maxwell(if they follow same protocol as the past) are a big no no on a new process like 20nm because producing the chip without getting to know the intricacies of the process makes producing big chips super expensive with a high amount of the products having defects. Nvidia learned this and its why a midrange keplar came out first. GM200 could be scaled up to 300%(but with lower clocks) if it was made on 20nm. The rumor going on right now is all the card makers are going to blow off 20nm and go to 16nm FinFet.
There is no 285x product on the way and its because it serves no purpose for the moment. It's no secret the chip the r9 285 is built on isn't a full chip. AMD has admitted this to most review websites. This is my theory, but what if Tonga has a yield problem and most chips come out with a defect so 99% percent of the chips can be turned into a r9 285 but only 1 of them can be turned into a fullfat Tonga? What would releasing fullfat Tonga at the same ti,e as r9 285 do? It would only serve to depress the price of r9 285. I have a feeling its going to be released with full fiji as part of the 300 series when yields are better.
About the 75% faster?
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/26.html
I like techpowerup for the number of games they test and variety.
Since performance percentages of 2560*1600 and 1920*1080 are the same and they are the most popular resolution, let use those..
100/57 = 1.754 or 75.4% faster.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,358
Well it is looking like a roughly 35% increase per core with better overclocking headroom, so Titan2 you are looking at roughly 50% overall increase in performance?
Caporegime
Soldato
Uggh, they've gone with EK instead of Swiftech again
What??? Tell me that's a joke? Swiftech are crap!
EVGA dropped them after the Titan Black launch
Look at it this way, Swiftech blocks have never been anything great to look at, but with light up EVGA logos they were very cool, it added to the card, it was something extra, a reason to buy the HC instead of just fitting a block yourself.
With the EK blocks that's gone because EK blocks look good anyway and the plexi coated in black design with an uncovered EVGA logo just looks rubbish by comparison to the standard block (and EK's plexi 980 blocks are pure sex), the is no reason to buy the HC for me now as I can buy the regular card plus the EK block and fit it (without effecting my EVGA warranty) save a load and get an infinitely better looking card, hell the OCUK tech labs 980's will basically be the same card just better looking.
Associate
Wonder how long until price-settles... are we going to be seeing really cheap 780s and 780Tis? AMD price-cut imminent?
Man of Honour
As I said earlier I am done with this the thread is about the 970/80 reviews.
You should not assume as much as you do
My first job was for an accountancy firm doing accounts.
Most of your facts in your last post are still wrong and I am not going to take time to correct them as this is not the place.
Soldato
Are there any reviews floating about which compare the various custom cooled 970s?
980GTX EK - Hydro Copper
Any word on release/clocks/price?
I can't even find where that image is from
Soldato
Who cares what it looks like? EVGA using EK is the best news for ages. The Swiftech blocks on the Hydro coppers don't even actively cool the VRMs. I've gone from EK to Hydro Titans and the difference in temperature is pretty staggering.
Just no!
Soldato
You're talking nonsense, off-topic nonsense at that. Take it to another thread please.
Associate
- Joined
- 20 Oct 2010
- Posts
- 1,194
I was looking for it myself and couldn't find any. I would like to see how custom cooled 970 ones perform compared to ref one.
Associate
Would a single 970 be significantly better than my single 7970?
Associate
- Joined
- 24 Oct 2013
- Posts
- 7
I agree with this. Since the GPC are arrange in clusters of 4 SMM and since the gtx 980 is made of 4 GPC, you have 16 SMM or 128x16 = 2048 shaders. SO something scaled up should contain additional blocks up of 1 GPC or 4 SMM more. Unless they change the configuration again like they did with the gtx 750 ti(which had 5 SMM in one GPC).
So scaling up and considering the current die size(398mm2), they should be able to fit 6 GPC or 24 SMM or 3072 shaders, 96 ROPS, 384 bit bus if they tighten the transistor density a bit. They will need to do this because these large chips they focus on double precision performance, which adds a tad more to the size. So I am guessing a slightly bigger maybe 620mm2 chip, with 30-40% more performance than this chip. But of course 1200 dollar price tag or something ridiculous. It's going to be hard to get good yields from a chip like this and we are likely to see something with just 20-25% percent more performance from the cut down version meant to be sold to most people.
Soldato
Associate
loving that post.