Nvidia reports record revenue of $1.43 billion, up 24pc year-on-year

Quote:

---

Strong demand for Pascal GPUs and Deep Learning interest behind the results says CEO.

---

Read more.

Awesome! ;)

How do the unit volumes compare to previous years?

Quote:

---

Originally Posted by kalniel

How do the unit volumes compare to previous years?

---

~35mil units of gpus combined (amd/nv) vs. ~100mil a few years back (don't think that counts workstation/grid etc). They are selling FAR more high end now which is why AMD should NOT have went bottom up. 1080 is the fastest selling high end gpu in history for NV and retailers (multiple retailers said this). Nv is still having issues keeping 1080 in stock at many places. More than 2/3 of newegg 1080's are out of stock. I hope AMD is aiming high with ZEN, meaning $400-1700 like HEDT. I'll gladly pay that if it beats Intel's $430 chip hands down, which it easily could if its 300mm^2 or larger. I'm hoping it's the size of an xbox soc (360+!). The cpu side is up even more than gpu side according to hdinsights, so AMD needs in on Intel's cash cows. MARGINS are king and there is a WHOLE lot of margin on $400-1700 chips (and high end gpus, look at NV margins vs. AMD). AMD hasn't had 58% margins since 2006 when they had $800+ chips.

Quote:

---

Originally Posted by nobodyspecial

~35mil units of gpus combined (amd/nv) vs. ~100mil a few years back (don't think that counts workstation/grid etc).

---

I thought as much, hence why the chase for margins is completely reasonable from a business perspective. Something that seems to have been forgotten in another thread around here ;)

Quote:

---

Originally Posted by kalniel

I thought as much, hence why the chase for margins is completely reasonable from a business perspective. Something that seems to have been forgotten in another thread around here ;)

---

Yeah,right unless those numbers can be verified,and even if they are,but also I like how people forget all of the following numbers and justify $1200 GPUs.

Now we see the fruits of the whole Titan rebranding saga,ie,large chips now pushed to excessive price levels(there are price increases and then there are price increases).

Margins

April 30, 2016 57.55%
Jan. 31, 2016 56.46%
Oct. 31, 2015 56.25%
July 31, 2015 54.99%
April 30, 2015 56.73%
Jan. 31, 2015 55.88%
Oct. 31, 2014 55.18%
July 31, 2014 56.12%
April 30, 2014 54.76%
Jan. 31, 2014 54.15%
Oct. 31, 2013 55.45%
July 31, 2013 55.82%
April 30, 2013 54.32%
Jan. 31, 2013 52.90%
Oct. 31, 2012 52.87%
July 31, 2012 51.78%
April 30, 2012 50.10%

Compare to the days of the 8800GTX,etc:

Jan. 31, 2008 45.70%
Oct. 31, 2007 46.21%
July 31, 2007 45.33%
April 30, 2007 45.03%
Jan. 31, 2007 43.89%
Oct. 31, 2006 40.70%

Revenue compared to 8800GTX days


https://ycharts.com/companies/NVDA/revenues

April 30, 2016 1.305B
Jan. 31, 2016 1.401B
Oct. 31, 2015 1.305B
July 31, 2015 1.153B
April 30, 2015 1.151B
Jan. 31, 2015 1.251B
Oct. 31, 2014 1.225B
July 31, 2014 1.103B
April 30, 2014 1.103B
Jan. 31, 2014 1.144B
Oct. 31, 2013 1.054B
July 31, 2013 977.24M
April 30, 2013 954.74M
Jan. 31, 2013 1.107B
Oct. 31, 2012 1.204B
July 31, 2012 1.044B
April 30, 2012 924.88M

April 30, 2008 1.153B
Jan. 31, 2008 1.203B
Oct. 31, 2007 1.116B
July 31, 2007 935.25M
April 30, 2007 844.28M
Jan. 31, 2007 878.87M
Oct. 31, 2006 820.57M

Plus it even gets better:

https://ycharts.com/companies/NVDA/cash_on_hand

April 30, 2016 4.754B
Jan. 31, 2016 5.037B
Oct. 31, 2015 4.728B
July 31, 2015 4.505B
April 30, 2015 4.792B
Jan. 31, 2015 4.623B
Oct. 31, 2014 4.241B
July 31, 2014 4.386B
April 30, 2014 4.348B
Jan. 31, 2014 4.672B
Oct. 31, 2013 3.033B
July 31, 2013 2.936B
April 30, 2013 3.713B
Jan. 31, 2013 3.728B
Oct. 31, 2012 3.435B
July 31, 2012 3.278B
April 30, 2012 3.131B


April 30, 2008 1.153B
Jan. 31, 2008 1.203B
Oct. 31, 2007 1.116B
July 31, 2007 935.25M
April 30, 2007 844.28M
Jan. 31, 2007 878.87M

Thats cash in hand. Cash in hand does not go up so much if you are struggling! ;)

$1.25 billion of that is down to Intel and I suspect there were losses down to Tegra,but it shows you selling smaller dies for more money has had a very postive impact as has gimping the midrange,which meant more people(probably me at this rate) will be either spending £200 to £300 on a midrange card or waiting longer and longer between upgrades under £200. Nvidia has made $2.3 billion+ in extra free cash and investments due to the new strategy.

I believe £200+ sales doubled last year,and I expect more will happen this year.

The excuse making is hilarious - especially the little side swipes.

Another hardware enthsuiast who is desperate to pay more because of PCMR.

No wonder card sales might have issues,when people justify $1200 graphics cards - I told you more and more people are moving to consoles due to cost.

But PCMR have their head in the sand as usual.

So to recap:
1.)Massively increased revenue our the last few years despite the losses with Tegra
2.)Over FIVE times the amount of free cash despite the increases to R and D and Tegra losses
3.)Margins massively up
4.)Tegra generating LOSSES for almost that entire time period




Whats hilarious,is even on some of the other tech forums,people are noticing the same trends too,and that includes a whole number who do buy higher end cards(are devs,etc) and also aware of those numbers I posted(even before me to add).

To add a term used on some other forums,they are milking PC gamers and hardware enthusiasts.

Its a sad day when Hexus has more excuses for this even more segmentation of cards than OcUK and other forums.

Yes,Nvidia is struggling we must pay more.

Reap what you sow.

Apparently! ;)

PS:

You can both have the last say on this as I am sure you will just ignore 99% of it and make up some metric to justify $1200 graphics card!

Teresa May should get some of the PR and marketing people companies like Apple and Nvidia use,as she would actually make people WANT to pay more tax and get relatively less.
;)

GPU pricing is just silly now, and the price/perf damage is twofold - first performance increases for a given sector are about the lowest I've ever seen them (regardless of how much they're hyped), and prices in dollars are continuing to creep up.

I don't buy that 16nm is as monstrously expensive as we've been led to believe, not after the Xbox One S SoC. That's a very price sensitive market and they could have gotten away with a minor change to the 28nm part if they needed to modify it.

I remember at the 680 release people were complaining about its price and positioning (small die sold as a top-end part). Nothing has improved since then. The 780 was what you'd expect for the 680, but it wasn't a new GPU anyway.

Quote:

---

Originally Posted by CAT-THE-FIFTH

Yeah,right unless those numbers can be verified,and even if they are,but also I like how people forget all of the following numbers and justify $1200 GPUs.

---

I don't think anyone is 'forgetting' anything - those numbers back up what we're saying. nVidia is chasing margins because by doing so revenue holds up despite falling unit volumes.

I don't get the justifying comment though - we're not in some party controlled country where offers to tender have to be justified for non-essential products. Yes, if we had to buy a titan X every day to survive then nVidia would be accused of abusing their position. But I've yet to feel any kind of need for a titan X, just like I don't for a Ford GT.

It's the cost of the products below the top-end that concerns me more. Let's just pretend the Titan is something more than an over-marketed gaming card, and hence ignore it for now - we're still seeing $699 pricing for the Gx104 cards, just comparing to Maxwell that's a $150 jump for the same market segment, we're still not seeing the MSRP prices with most cards hovering around that FE pricing.

The GM204 was a bigger die too, but we can perhaps put that down to the extended life of 28nm making it cheaper than is typical for a given die size.

GTX500 was the last series before Nvidia changed the positioning and pricing of GPU tiers, and back then the GF114 was $199-$249 - this more or less the same tier the 1070 and 1080 now reside in. Yes, development and production costs for a given die size have risen, but I really doubt it's in line with the product prices!

But is it in line with a shrinking market?

Either way, it doesn't affect you or I. We have a budget and we'll get the best we can for that budget regardless of what model number the card is. If that means more people are buying 'lower tier' products then so be it - devs will target their specifications accordingly.

But here's the thing - we're moving down model numbers, but getting better performance than we got last time we upgraded cards. If I buy a 480 or 1060 now for £230 then the performance is going to be many times higher than my 7870 which cost £250 however many years ago.

Better performance for a given price point now takes *much* longer to materialise than it did though - I remember when we got more than a doubling of performance something like every 18 months within the same price bracket.

I also tend to stick to the same ~£230 price bracket as you, but in my case that bought me a 280X, and after three years that same £230 doesn't come close to getting me double performance, which I find quite disappointing.

With regard to market size - how much is it really shrinking by overall? Lower overall desktop sales and faster IGPs are killing off lower-end GPUs but how big of a market were they anyway, and what were the margins like on the really cheap parts? We also keep hearing how well the mid-range/high end GPUs are selling now. I don't know the answers to these questions off-hand, but Nvidia has much higher revenue despite this supposed decline in sales, so you can't really argue it's a case of just balancing the books

Well it is balancing the books, but as a public company that also includes the aim to increase returns for their shareholders, so to balance that additional cost they need to increase revenue and/or decrease other costs ;)

We don't have to like that of course (unless we're a shareholder, disclaimer: it's quite likely that I'm indirectly one through a tech portfolio along with AMD, ARM/Softbank, Intel and the likes) but I don't share the doom and gloom sentiment as a result of them doing so. I am personally quite happy for tech to slow down a bit - rather than hoping for a doubling of performance every three years I'd like my purchases to remain viable for 5+ years, which means devs have to target 5 year old performance levels - the smaller the gap in those 5 years the easier that becomes for them.

Quote:

---

Originally Posted by watercooled

Better performance for a given price point now takes *much* longer to materialise than it did though - I remember when we got more than a doubling of performance something like every 18 months within the same price bracket.

---

The law of diminishing returns in action, we've been seeing the same thing with CPU's and you get the same effect with lots of other things.

Quote:

---

Originally Posted by kalniel

Well it is balancing the books, but as a public company that also includes the aim to increase returns for their shareholders, so to balance that additional cost they need to increase revenue and/or decrease other costs ;)

---

Quote:

---

Originally Posted by kalniel

We don't have to like that of course

---

That's basically the point we, or at least I, am getting at. They're not just raising prices because market conditions are somehow forcing them to, they're doing it because they can.

Quote:

---

Originally Posted by kalniel

I am personally quite happy for tech to slow down a bit - rather than hoping for a doubling of performance every three years I'd like my purchases to remain viable for 5+ years, which means devs have to target 5 year old performance levels - the smaller the gap in those 5 years the easier that becomes for them.

---

Interesting way of looking at it, swings and roundabouts I guess, but that doesn't really seem to be happening for a lot of games. We get plenty of games where performance is pretty terrible despite really not looking all that good. I'm still not sure who's to blame with Gameworks for instance, but it does seem more than coincidental that nearly every game using the libraries displayed those traits. It often disproportionately harmed performance on non-Nvidia hardware, but even on Nvidia hardware they hardly ran well.

I'm all for devs having a reason to put an effort into better optimising code, like how the 360 console generation encouraged better multithreading in games rather than just lumping everything into a single thread because they could get away with it on PC, expecting people to continually upgrade in order to run increasingly bloated code.

Quote:

---

Originally Posted by Corky34

The law of diminishing returns in action, we've been seeing the same thing with CPU's and you get the same effect with lots of other things.

---

Oh I'm not ignoring that, we're not getting new nodes and half-nodes they way we used to so I am of course considering that, it's just that the combined effect of slowing progression and rip-off pricing makes it a whole lot worse than it needs to be. Having said that, if you look at performance for a given die size, GPUs are still scaling fairly well and will likely continue to do so for a while given their highly parallel nature. Like we've said, dies that are now being sold as high-end cards in terms of marketing and pricing, are much smaller than they used to be, so it makes the slowing seem worse than it really is - even Nvidia's internal die naming basically acknowledges that, Gx104 are now being sold to the market the Gx100 dies used to address. It's not just that we're waiting longer for improvements because of less frequent lithography shrinks, which I could understand.

Yes, wafer costs are increasing and hence cost/mm2 is said to be increasing too, but not nearly to the extent as to offset the far smaller dies we're now seeing. Even if we go crazy high, much higher than Nvidia's doom-and-gloom graphs claimed, and assume it's now twice as expensive for a given die size, that still doesn't offset the reduction in die size/increase in cost we've seen at retail. And lets not forget that the die cost is only a part of GPU pricing - even if die cost had doubled, that should not lead to a doubling of graphics card pricing.

So bearing all that in mind, it just seems like a convenient thing to blame really - like I say, due to the embarrassingly parallel nature of graphics workloads, GPUs are continuing to scale just fine, they're not really affected by the same performance walls as CPUs where core sizes have shrunk to the point of being essentially pad-bound, having unused die area and throwing on huge (in comparison to CPU cores) integrated GPUs to fill them out. GPUs can still make use of all the transistors you throw at them, especially given the higher resolution/FPS/VR hype train. You can see this for yourself in how performance scales surprisingly well with die size/transistor count, in fact the smaller GPUs often do a bit less-well than you'd expect because of e.g. memory controllers and fixed function blocks taking up a more significant amount of space relative to the die as a whole.