- Jun 8, 2003
- 14,387
- 480
- 126
Elfear
Diamond Member
- May 30, 2004
- 7,097
- 644
- 126
Actually many reviewers called a spade a spade when the GTX 680 moved up a tier.
Techreport
“In terms of basic, per-clock rates, the GK104 stacks up reasonably well against today's best graphics chips. However, if the name "GK104" isn't enough of a clue for you, have a look at some of the vitals. This chip's memory interface is only 256 bits wide, all told, and its die size is smaller than the middle-class GF114 chip that powers the GeForce GTX 560 series. The GK104 is also substantially smaller, and comprised of fewer transistors, than the Tahiti GPU behind AMD's Radeon HD 7900 series cards. Although the product based on it is called the GeForce GTX 680, the GK104 is not a top-of-the-line, reticle-busting monster. For the Kepler generation, Nvidia has chosen to bring a smaller chip to market first.”
"Also, the GeForce GTX 680 is a massive generational improvement, extracting roughly twice the performance of the GeForce GTX 560 Ti from a similar class of GPU. Still, we're a little disappointed Nvidia isn't passing along more of those gains to consumers in the form of higher performance per dollar, as has happened in the past. Half a grand is a lot to ask for a mid-sized chip on a card with a 256-bit memory interface."
Tech Radar
"But the GTX 680 is not a natural successor to historical heavy weights like the Nvidia GeForce GTX 580, 480 and 280. The first hint that the GTX 680 might be a bit different is its GK104 internal codename. That codename also invokes parallels with something similar that happened at Nvidia's main rival AMD. More on that in moment.
If you knew nothing else about the GTX 680 other than the GK104 codename, you'd assume it was the successor to the GF104 (and its respun cousin the GF114), the graphics chip that begat the Nvidia GeForce GTX 560.
That was the second tier, rather than range-topping, graphics chip in Nvidia's outgoing GeForce 600 series of GPUs.”
But in the past, Nvidia has typically taken the opportunity to keep the overall chip size up and stuff in more transistors. Thus when the GTX 480 replaced the GTX 280, the transistor count more than doubled from 1.4 billion to 3 billion.
But the new GeForce GTX 680 only ups the ante to 3.5 billion. What's more, the new GTX 680 only sports a 256-bit memory bus. That's somewhat inevitable given the physical proportions of the GK104 chip and the limitations that places on the number of available contact points. But it means Nvidia's flagship GPU now has a narrower memory bus than its top chip of four years ago.
Now, you could say none of this matters. The only critical metric is performance. The GTX 680 is the fastest single you can buy. The end. But that's to ignore both what might have been and, arguably, what ought to be.
So, here's the bottom line. The GTX 680 is not a high end GPU. But it's being sold for the price of a high end GPU and that very likely makes it hugely profitable. Somewhere deep inside Nvidia's labs, there's another, much larger and more powerful GeForce 600 Series GPU."
Extremetech
“Those of you familiar with Nvidia’s historic naming schemes will recognize the GK104 moniker as one that Team Green typically would reserve for a mid-range GPU.”
Hexus
"Nvidia pulled off an enviable trick with its high-end GeForce GTX 600-series cards. This trick encapsulated two parts: a change in architecture from Fermi (GTX 580) to Kepler (GTX 680) along with an accompanying shrinking of the transistors that make up the GPU, down from 40nm to 32nm.
And what do you know, Nvidia managed to increase performance by 30-odd per cent, drop power by around 50W, and herald a new GeForce performance champion at the same $499 price point. But the real benefit to Nvidia lay in substantially reducing production costs, because, by historical standards, GTX 680 uses a mid-range-sized die.
In tenuous car parlance this was like swapping out an expensive-to-produce 3.0-litre engine, replacing it with a cheaper, more powerful 1.6-litre, and pocketing the difference. The Kepler architecture's impressive performance-per-watt metric has now been distilled much further down the product stack and, crucially, been propagated into the notebook market, where perf-per-watt is king."
Guru3D
“The GeForce GTX 680 being reviewed today is based on the new Kepler GPU architecture. Interestingly enough it is based on the 28nm GK104 GPU which typically would have indicated a mid-range product.”
PcPer
“The chip that powers the GTX 680 is the GK104, and it is oddly enough the more "midrange/enthusiast" offering.”
- Jun 8, 2003
- 14,387
- 480
- 126
threads closed. you guys ruined it with your die size crap AGAIN,
I said the list was based on price & performance.
The question asked was are we paying more for mid range performance, not die size.
I asked the die size, power consumption, wafer cost argument stays out of the thread.,
Now even the mods ruin threads..
thanks
good day
Mod callouts are not allowed. Why did you go and delete every post you made in this thread?
esquared
Anandtech Forum Director
I said the list was based on price & performance.
The question asked was are we paying more for mid range performance, not die size.
I asked the die size, power consumption, wafer cost argument stays out of the thread.,
Now even the mods ruin threads..
thanks
good day
Mod callouts are not allowed. Why did you go and delete every post you made in this thread?
esquared
Anandtech Forum Director
Last edited by a moderator:
Despite a surge in revenue over the past few years, Nvidia's gross margins have been extremely stable. Going back to Q3 of 2011 (500 series released in 2011) it has fluctuated between 50.1% and 59.03% including 12 out of 21 quarters between 54.0% and 57.0%. In their just released Q3 results, revenue was up 53.6% while gross margin only increased 2.7%. If Nvidia is just arbitrarily raising prices to screw the customers over instead of adjusting prices to account for COGS, why are their gross margins so stationary?
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
What you and many others don't understand is the following,
If you payed $500 to get 150% performance tier (Fastest Single Chip Solution on the market) in 2010 with the GTX 480
For the same 150% performance tier in 2016 you pay $1200 for the TITAN X Pascal
If you payed $250 to get 100% performance tier (Middle-End Single Chip Solution on the market) in 2010 with the GTX 460
For the same 100% performance tier in 2016 you pay $600-700 for the GTX 1080
If you payed $139 to get 67% performance tier in 2010 with the GTX 450
For the same 67% performance tier in 2016 you pay $250 for the GTX 1060
So yes, TODAY for the mid range tier (GTX 460 - GTX 1080) we pay 2.4x to 2.8x more than we did for the same performance tier in 2010.
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
Just to point out that Margins and ASP are not the same
You can have the same Margins selling a 100$ product or a $1000 product.
StrangerGuy
Diamond Member
- May 9, 2004
- 8,443
- 124
- 106
Having a more efficient perf/die area shouldn't be rewarded by the market because it's a heinous crime. *sarcasm*
If the tables were turned, we will be endlessly hearing how amazingly efficient AMD's designs are from the usual loudmouths in these parts.
Which totally misses the point and doesn't answer the question. If Nvidia was selling medium dies for $250, and the cost to produce said dies remained the same, then selling them for $500+ that they do now should obviously increase their gross margins significantly. Why have Nvidia's gross margins been so stable if they are not adjusting MSRP's in line with the increase in costs to produce the products?
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
The ASP and sales Volume went up significantly in Q3. As I have said before, NVIDIA managed to make people step up one or two tiers selling them more expensive cards and thus increasing the ASP of its products significantly. Margins stayed relatively the same but the ASP got ballistic.
Still not answering the question. ASP has nothing to do with what I am asking. Answer the question I asked.
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
Lets see NVIDIAs margins,
http://investor.nvidia.com/results.cfm
FY 2015
Q1 = 54,8%
Q2 = 56,1%
Q3 = 55,2%
Q4 = 55,9%
FY 2016
Q1 = 56,7%
Q2 = 55,0%
Q3 = 56,3%
Q4 =56,5%
FY 2017
Q1 = 57,5%
Q2 = 57,9%
Q3 = 59,0%
FY 2017 margins are the highest the company had the last 3 years. They have an increase in Margins from FY Q3 2016 to FY Q3 2017 for every Quarter.
Now lets see Q2 and Q3 FY 2017
Q2
Revenue = 1428M
Margins = 57,9%
COGs = 601,188M
Profit = 1428 - 601,188 = 826,812M
Q3
Revenue = 2004M
Margins = 59%
COGs = 821,640M
Profit = 2004 - 821,640 = 1182,36M
Q3 FY 2017 had,
Higher ASP
Higher Margins
Higher Volume
= Higher Revenue = Higher profits
Q2 to Q3 COGs increased by 36,6% while Revenue increased by 40,33%.
This answer your question
For reference, Margins NVIDIA had in 2010 (FY 2011) when they released Fermi GF100/GF110
FY 2011
Q2 = 16,6% - Fermi GTX 480 (GF100) Launch
Q3 = 46,5%
Q4 = 48,1% - GTX 580 (GF110) Launch
Edit: It is amazing that NVIDIA increased their Margins (since 2010 when they released Fermi GF100) in a time when they complained about the increase of Node/manufacturing cost and some people here insist that performance/cost is the same as before.
No people, NVIDIA is selling at higher prices per tier today than they did back in 2010, simple as that.
Last edited:
No dude. You're straight up, 100% wrong on this. Trying to artificially limit parameters to fit a perplexingly strange desired outcome does not make it any more correct.
Here's how logical reasoning works: Facts -> Analysis -> Conclusion. Here's how you want this thread to work, apparently: Conclusion -> Facts -> Analysis.
I can say everyone in the world speaks english if I define "the world" as "people living in my house" but it doesn't mean you can go around saying "Everyone in the world speaks english"
ultimatebob
Lifer
- Jul 1, 2001
- 25,135
- 2,445
- 126
Is it just me, or is the GeForce GTX 1060 not really an "Low End" card? It seems more like a mid-range card both in pricing and performance.
Hmm there is one argument that I can think of that would support AtenRa's narrative:
Fermi launched into a time where 1080p started to take over the market. Pascal launches into a time where 1440p, 21:9 resolution and even 4k start to take over the market. If we're talking about these higher resolutions, then yes, the 1060 would look more like these smaller Fermi cards. At 1080p though I would absolutely agree that it's not low-end performance.
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
GTX 460 was a Middle-End card both in performance and price over the High-End card GTX 480 (for Single Chip Graphics Cards)
GTX 480 at $500 was 50% faster than GTX 460 at $229
Today TITAN X Pascal at $1199 is 2x faster than the GTX 1060 at $250
Today GTX 1060 at $250 is in the same tier ($130) as GTX 450 was back in 2010.
2010
GTX 480 = High-End Single Chip GPU = $500
GTX 460 = Middle-End Single Chip GPU = $229
GTS 450 = Low-End Single Chip GPU = $130
GT 440 = Entry-Level Single Chip GPU = $100
Today
TITAN X Pascal = High-End Single Chip GPU = $1199
GTX 1080 = Middle-End Single Chip GPU = $600-700
GTX 1060 = Low-End Single Chip GPU = $250
GTX 1050Ti = Entry-Level Single Chip GPU = $139
You're forgetting the GTX 590 which was $699 at launch. And the GTX 690, and Titan Z, which were $999 and $2999 respectively. What category do those fall in?
Corrected!
My corrections make a lot more sense, and reflect not only pricing but actual performance as well..
Last edited:
Corrected!
I agree that people that bought a 1080 probably want to feel like they paid a lot of money for a high end card, even though it's more of a mid-ranger historically (code name, die size etc). They did pay $600-$700. In addition, it's quite clear that AMD has nothing close to a 1080 currently, so it's not like they really have any other options.
However, the performance characteristics of the 580/560ti/550ti/545 basically line up perfectly with the titan xp/1080/1060/1050ti.
Giving the Titan 100% (4k):
Titan XP 100%
1080 76%
1060 44%
1050ti 26%
Giving the 580 100% (2560x1600):
580 100%
560ti 73%
550ti 43%
545 26% (@ 1080p)
It's possible that I made a mistake in a my calculations somewhere.
Source:
https://www.techpowerup.com/reviews/NVIDIA/Titan_X_Pascal/24.html
https://www.techpowerup.com/reviews/ASUS/GeForce_GTX_550_Ti_Direct_Cu/23.html
https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_560_Ti/27.html
https://www.techpowerup.com/reviews/MSI/GTX_1050_Ti_Gaming_X/27.html
https://www.techpowerup.com/gpudb/627/geforce-gt-545
Comparing things with the 590 is just wrong, though. It seems like SLI is being abandoned/neglected (at least that's what I understand from posts on this board), there's just no dual-gpu high end card anymore. If Nvidia suddenly releases a dual Titan XP tomorrow, will the 1080 move from your "high end" tier to your mid-range tier?
Last edited:
AtenRa
Lifer
- Feb 2, 2009
- 14,001
- 3,357
- 136
GTX 590 = DUAL Chip GF110 - there is no such tier today
GTX 690 = DUAL CHIP GK104 - there is no such tier today
TITAN Z = DUAL Chip GK110B - there is no such tier today
Todays GTX 590 equivalent would be a DUAL Chip Graphics Card with 2x GP102 like TITAN Z
Regarding dual GPUs, yes they aren't around today; at least for NVidia.. But they were around for years before, for several successive generations, so you can't just sweep them under a rug as much as you'd like to.. These constituted the ULTRA HIGH END category which I referred to earlier.
However, NVidia has seemingly abandoned dual GPUs in favor of making larger and faster single core GPUs, which is actually better if you ask me, because dual GPUs don't have the reliability that single GPUs have. These bigger GPUs like the GTX 980 Ti, Titan X, Titan XP etcetera have now displaced those dual GPUs and are the new ultra high end category..
However, NVidia has seemingly abandoned dual GPUs in favor of making larger and faster single core GPUs, which is actually better if you ask me, because dual GPUs don't have the reliability that single GPUs have. These bigger GPUs like the GTX 980 Ti, Titan X, Titan XP etcetera have now displaced those dual GPUs and are the new ultra high end category..
Except the fact that Nvidia are not making larger single core GPUs (faster, yes). They're mostly the same size as they were before, even a bit smaller. The Titan XP is 471mm^2 (actually, effectively less as it is a cut GP102). For reference the 580 was about 520mm^2, 780ti was 580mm^2, 480 was also 580mm^2, 280 was 576mm^2. The 980ti/Titan x was much larger, at 601mm^2.
As already presented here before, by me just a few posts ago, the current performance differential between the Titan XP/1080/1060/1050ti is essentially identical to the differential of the 580/560ti/550ti/545. There are no "new larger" single core GPU. Just new GPUs that were pushed up the stack. The old midrange GPU has become "high end" and the old "high-end" has become "ultra high-end".
EDIT: Added 900 series. Also, Just for reference, the 1080's 314mm^2 is a bit larger than the 294mm^2 760, but smaller than the 360mm^2 and 332mm^2 560ti and 460 respectively, and much smaller than the 398mm^2 980.
Last edited:
What's your point? Of course there are more transistors, thanks to improvements in process technology and R&D. This has been a true fact in every new generation of GPUs, and (in this regard) there's nothing special in the newer GPUs.
As already stated, carfax's claim is false. The Titan XP is not a larger chip, and nvidia has not been making larger GPUs. They've been making newer smaller GPUs and pushing them up the stack, thanks to a lack of competition in the high end (the 980ti still beats every AMD card almost a year and a half later), great marketing, and a very very loyal fanbase.
As already presented above, both the 1080 and Titan XP fit right in with mid-range and high-end division of the 2,4,5,7 series (and also G80, although I did not include it). Their performance characteristics,physical dimensions and even code names place them very nicely in other historical high-end Nvidia GPUs as the x60 and x80 cards, respectively.
But as 96Firebird stated, transistor count has obviously increased. What this tells me is that die size isn't as important for determining performance as you think. Performance per watt, has increased TREMENDOUSLY since Fermi, whilst die size obviously hasn't.
This goes to show where NVidia is focusing the brunt of their R&D effort; into increasing performance per watt, or performance per mm^2. Compared to AMD, they have a massive lead in those two areas.
I thought it was quite obvious. Instead of looking at die size, look at transistor count. Guess which one effects the performance of a GPU? Or do we not care about performance anymore?
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 10K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
-
-
Facebook
Twitter