Nvidia ,Rtx2080ti,2080,2070, information thread. Reviews and prices September 14.

Page 50 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

zinfamous

No Lifer
Jul 12, 2006
111,994
31,557
146
Because Die size is the primary determinate of production costs, and in the end, retail pricing.

Look at right now, with falling GTX 1080 prices. Have Vega 64 prices followed?

GTX 1080 cards have better performance, are from the leading company (mindshare) and are selling for less.

Why haven't competitive pressure dropped Vega 64 prices?

Everything from a consumer perspective says they should be selling for less than GTX 1080. But they aren't.

You have to look beyond the consumer perspective, to understand why, and it's the production cost, driven primarily by die size economics.

It isn't economically viable to sell that big chip card for the price of the small chip competitor. Ultimately they may be forced to sell remaining stock cheaper, but that will be at a loss, which isn't economically viable. They will probably soon end production of Vega consumer cards (if it already hasn't ended).

When dies get larger, production costs increase, and production cost increases are normally passed on to the consumer.

No, the profits are passed on to the shareholders in these cases, because they choose to pass the costs onto consumers. At what is it, an unprecedented climb to ~67% margins, the decisions that they are making is clear. The cost to produce is always a cost to consumer is a tired, and unsupported argument. Nvidia would still be printing $$$$ and still be running full steam ahead with normal, sane pricing, but the board is dead set on not shaving off ~3% of their record margins.

This is a symptom of a company driven by short-term, bubbly, 10-fold share price gains. Absolutely nothing more re: pricing is going on beyond a determined board to keep that momentum exactly where it is.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Bottom line, dont know about gaming, but for pro-apps like Octane, this is gonna be absolutely awesome, at least it looks that way. .

There is no chance, all the dedicated RT HW, isn't going to have a similar multiplier effect on Ray Tracing in Games. Dedicated RT HW is the only way to go for the foreseeable future.

ub4ty just likes writing huge hyper critical essays about anything that competes with AMD.
 

Hitman928

Diamond Member
Apr 15, 2012
6,754
12,500
136
Because Die size is the primary determinate of production costs, and in the end, retail pricing.

Disagree. Consumer demand and it's relationship to price elasticity is what ultimately determines retail pricing. It doesn't matter what your production costs are if no one will pay for it.

BTW, you still didn't answer my question. I already acknowledged your reasoning in my post but you still didn't say why I should care, as a consumer, how big Nvidia wants to make their die if it doesn't give me any benefit in perf/$.

Look at right now, with falling GTX 1080 prices. Have Vega 64 prices followed?

GTX 1080 cards have better performance, are from the leading company (mindshare) and are selling for less.

Why haven't competitive pressure dropped Vega 64 prices?

Everything from a consumer perspective says they should be selling for less than GTX 1080. But they aren't.

GTX 20xx line hasn't launched yet so there's no need for AMD to drop prices yet. I'm sure Nvidia and partners want to clear as much inventory before the 20xx series launches as possible, hence the price cuts. AMD doesn't have inventory to clear, they have no new cards coming out, so they'll keep the cards at their price as long as possible (i.e. when the RTX series becomes readily available). Additionally, Vega cards were the last cards miners were buying for their farms as the crypto crash happened, so the demand might still be trailing off from that.

You have to look beyond the consumer perspective, to understand why, and it's the production cost, driven primarily by die size economics.

Not necessarily as I explained above. You're also completely ignoring the "fixed" costs that have built up leading to the release of a product. All the R&D expenses are a very significant cost of a product. The mask set is also significantly more expensive than an actual wafer run. Both of these costs are "fixed" going into production, are orders of magnitude higher than the cost of the actual die, and must be amortized over the lifetime of selling the product.

It isn't economically viable to sell that big chip card for the price of the small chip competitor.

Sure it is, your margins will take a hit, but it's certainly viable at current costs. But again, why should I as a consumer be concerned with a company having lower margins?

Ultimately they may be forced to sell remaining stock cheaper, but that will be at a loss, which isn't economically viable. They will probably soon end production of Vega consumer cards (if it already hasn't ended).

When dies get larger, production costs increase, and production cost increases are normally passed on to the consumer.

Once again, depends on the market conditions and demand. Production cost means nothing to the demand of a product, performance does. You're basically arguing that we should want Nvidia to have as high of margins as possible. That, to me, is an insane perspective as a consumer, it's a consumer consciousness equivalent of the meme, "thank you sir, may I have another." (google it for those who may not know the reference).
 
Last edited:

arandomguy

Senior member
Sep 3, 2013
556
183
116
When Nvidia was selling Fermi their gross margins I think were in the 30% range. Nvidia's gross margins for around Pascal's generation were around 60%.

Cost (and there more to cost than die size) may dictate a price floor but there is no way these things are priced according to cost. Consumer electronic/computer parts in these categories are luxury goods, they are almost certainly priced relative to market demand and company strategy (eg. "loss leaders").

I don't remember where, but I read that TSMC's own comment on 12 nm was that it didn't give any performance benefits versus it's 16 nm process. So I am assuming the max OC on the new generation will be roughly the same as the old.

12 nm FFC according to TSMC is supposed to offer better performance/power characteristics and density (if taken advantage of) vs their 16nm variants. I don't believe there is any official word on what exactly 12nm FFN offers that Nvidia uses.

https://www.semiwiki.com/forum/content/6662-tsmc-talks-about-22nm-12nm-7nm-euv-e.html

12nm FFC offers a 10% performance gain or a 25% power reduction. 12nm also offers a 20% area reduction with 6T Libraries versus 7.5T or 9T.
 

crisium

Platinum Member
Aug 19, 2001
2,643
615
136
GTX 770 2GB release day May 2013 , die size 294mm2 at 28nm, launch price $399

GTX970 3.5GB release day September 2014, die size 398mm2 at 28nm , launch price $329

GTX970 at $329 was as fast as GTX780Ti that launched at $699 one year earlier.

This is the normal thing to happen in computer hardware. We always get higher performance at lower price 1-2 years later.

Wait, you mean to tell me we used to get bigger die cards with higher performance, more VRAM, and for a lower price? I have it on authority that as consumers we should just count our lucky stars that we got bigger dies and needn't worry about those other bothersome metrics.
 

arandomguy

Senior member
Sep 3, 2013
556
183
116
I don't think we can write off perf/$ increases yet. If RTX 2070 does match 1080ti than we did go from $650 - > $599 (if we want to use the FE price) along with new features.

Sadly the issue is also one of competition. AMD 7970 was a competitor to the GTX 770. 7xx series also had another price drop when AMD launched Hawaii and 2xx later that year. Kepler arguably at that time was also suffering some mind share issues due the speculated impact of the console generation changeover (which ultimately did prove somewhat true).

9xx (Maxwell) did come off an earlier mining crash. At that point comparable performance Hawaii cards were in the sub $300 range. This crash even made AMDs own 3xx launch the following year look poor as there was regression in perf/$ vs what existed for 2xx in terms of street prices.

Currently Vega itself is a poor perf/$ for gaming and may not even reach those performance levels depending on where the RTX 2070 actually lands. While Pascal cards have not yet experienced the deep price crash of the previous mining boom/bust as the 2xx series did.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
Disagree. Consumer demand and it's relationship to price elasticity is what ultimately determines retail pricing. It doesn't matter what your production costs are if no one will pay for it.

BTW, you still didn't answer my question. I already acknowledged your reasoning in my post but you still didn't say why I should care, as a consumer, how big Nvidia wants to make their die if it doesn't give me any benefit in perf/$.

You don't have to care.

But if you are ranting that they are indiscriminately increasing prices for no reason other than price gouging, then that is a position of ignorance. Production cost have increased and they are passing those on.

Don't like it, don't buy it. But don't pretend this is merely price gouging.
 
  • Like
Reactions: TheF34RChannel

arandomguy

Senior member
Sep 3, 2013
556
183
116
Obviously there is multiple factors going into end pricing but it's pretty doubtful that price increases were simply scaled to cost increases.

When AMD dropped Fury Nano $150 and Nvidia dropped GTX 780 $150 was it because costs all of sudden went down by that amount in a manner of months? These things all have large margins built in.

If Nvidia wanted to increase the rate of adoption of their platform I'm sure they would be pricing these lower as well. It isn't part of their strategy and if the market is willing to pay then this is what we have.
 

Hitman928

Diamond Member
Apr 15, 2012
6,754
12,500
136
You don't have to care.

I know I don't have to care, but why should I care? I'm asking why you keep arguing that die sizes justify significant price increases; why should I care as a consumer? I agree it makes an interesting technical discussion and could make an interesting business discussion, but as a consumer, on a consumer forum, talking about consumer pricing, what is the point?

But if you are ranting that they are indiscriminately increasing prices for no reason other than price gouging, then that is a position of ignorance. Production cost have increased and they are passing those on.

Don't like it, don't buy it. But don't pretend this is merely price gouging.

I don't think I ranted once about price gouging, pretty sure I never even mentioned it and pretty sure I never will because I don't believe that price gouging is a thing (almost never). I've just asked why you think increased production costs should justify in the mind of the consumer a disproportionate increase in retail pricing regardless of value to the consumer? I've also explained that die size is not the only metric to be concerned with in production cost and is actually not even the majority of the cost, so even from a business perspective the justification of increased wafer cost to the increase in retail pricing doesn't seem to hold much water.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,787
136
Production cost have increased and they are passing those on.

Don't like it, don't buy it. But don't pretend this is merely price gouging.

Hate to repeat this, but since that seems to be your strategy,

No, its all about revenue. Modern corporations have anthropologists and psychologists trying to find out what will entice consumers - basically its the foundation of modern marketing. And naming is crucial to their strategy.

coercitiv said:
The "Turing Titan" in "2080Ti" clothing is pulling all prices up.

Step 1. Train and consumers about your brand
Step 2. Establish halo status
Step 3. Change naming, raise prices

They are all vitally important. "Oh, I'm buying a Ti". It gives them the ammo to justify all sort of price increases so they won't feel bad when they do buy it. It's not necessarily the consumer being idiots, its the corporation being manipulative.
 

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
136
I don't think I ranted once about price gouging...
Whether you specifically made the argument or not isn't the point.

Much of the RTX discussion is consumed with the argument that the price increase is completely unjustified. I am just pointing out that rising production costs are reasonable justification.
 

alcoholbob

Diamond Member
May 24, 2005
6,390
470
126
This timespy leak, if true, basically implies 0 IPC improvement, at least over Volta async improvements.

~10,000 score for GTX 2080 would mean around 14,000-15,000 score for 2080 Ti.

Titan V already scores >14,500 at around 1650MHz boost clocks. So a 2080 Ti with a open air cooler hitting even high clocks should be near 15k mark, even with 0 IPC improvement, which is what this leak implies.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,576
126
This timespy leak, if true, basically implies 0 IPC improvement, at least over Volta async improvements.

~10,000 score for GTX 2080 would mean around 14,000-15,000 score for 2080 Ti.

Titan V already scores >14,500 at around 1650MHz boost clocks. So a 2080 Ti with a open air cooler hitting even high clocks should be near 15k mark, even with 0 IPC improvement, which is what this leak implies.
Is Timespy able to fully utilize the features of Turin though?

This is probably a low result due to early drivers and a lack of Turin core support.
 

Malogeek

Golden Member
Mar 5, 2017
1,390
778
136
yaktribe.org
This timespy leak, if true, basically implies 0 IPC improvement, at least over Volta async improvements.

~10,000 score for GTX 2080 would mean around 14,000-15,000 score for 2080 Ti.

Titan V already scores >14,500 at around 1650MHz boost clocks. So a 2080 Ti with a open air cooler hitting even high clocks should be near 15k mark, even with 0 IPC improvement, which is what this leak implies.
Is there a reason why Turing isn't considered the consumer offering of Volta with the additional RT cores added? As far as the shaders are concerned. Because it's a different chip name and released last year therefore it's a "generation" behind?
 

Despoiler

Golden Member
Nov 10, 2007
1,968
773
136
Is Timespy able to fully utilize the features of Turin though?

This is probably a low result due to early drivers and a lack of Turin core support.

It's using the traditional raster pipeline that everything uses currently because that is what the benchmark measures. Why would it use tensor or RT cores for that? Early drivers is something talked about when showcasing early silicon. It's about 2 weeks from embargo lift. Drivers aren't going to shift performance in a large way this close to launch.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,576
126
It's using the traditional raster pipeline that everything uses currently because that is what the benchmark measures. Why would it use tensor or RT cores for that? Early drivers is something talked about when showcasing early silicon. It's about 2 weeks from embargo lift. Drivers aren't going to shift performance in a large way this close to launch.
What about Turin's DLSS and AI capabilities?
I doubt that NV has already made release-ready drivers available.
 

Eddward

Member
Apr 10, 2012
56
19
81
There is something 100% not right with this leak... along with TimeSpy there is also Firestrike Ultra score on the same card from the same guy, but this score is way below GTX 1080, around 60% lower than average GTX 1080, which is absurd...
 

alcoholbob

Diamond Member
May 24, 2005
6,390
470
126
Is there a reason why Turing isn't considered the consumer offering of Volta with the additional RT cores added? As far as the shaders are concerned. Because it's a different chip name and released last year therefore it's a "generation" behind?

Because people want to believe in IPC increases, despite the fact that a serious IPC increase only happened once, with Maxwell. Not just "hey we finally brought async compute to the consumer market."
 

Despoiler

Golden Member
Nov 10, 2007
1,968
773
136
What about Turin's DLSS and AI capabilities?
I doubt that NV has already made release-ready drivers available.

FutureMark is in the process of making a ray tracing benchmark. It's not going to use Nvidia's current proprietary implementation. It will use DXR and/or Vulkan. There won't be any benchmark that uses DLSS as it's proprietary. The only thing you will see concerning DLSS is a comparison between some existing form of AA and DLSS on Nvidia cards much like Nvidia has already provided.
 

psolord

Platinum Member
Sep 16, 2009
2,142
1,265
136
Because people want to believe in IPC increases, despite the fact that a serious IPC increase only happened once, with Maxwell. Not just "hey we finally brought async compute to the consumer market."

Is gpu ipc really that straightforward to tell apart, in different gpus? I mean we have FP16, FP32, FP64, tesselation, async compute and now raytracing. I do not know if I understand it correctly, but gpus have different performance, with different coding, do they not?
 

sze5003

Lifer
Aug 18, 2012
14,320
683
126
There is something 100% not right with this leak... along with TimeSpy there is also Firestrike Ultra score on the same card from the same guy, but this score is way below GTX 1080, around 60% lower than average GTX 1080, which is absurd...
They also say in the article it looks like it could be a 2080 or a 2070. These numbers don't look very persuasive.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
They also say in the article it looks like it could be a 2080 or a 2070. These numbers don't look very persuasive.
Man...its just sad. I hope this is fake or a 2070.
Its like 2 years of nearly no progression.
 
Status
Not open for further replies.