Geforce GTX 1060 Thread: faster than RX 480, 120W, $249

[DamZe]

Newegg has the $249 EVGA ACX 2.0 in stock. BTW the 3GB model (GP106-300-A1) was listed again in Germany, should be close to launch.

A 3GB card in 2016? What were NVidia thinking?

EDIT: To think they have the audacity since the 1060 has 980-performance level.

 

[antihelten]

BTW the 3GB model (GP106-300-A1) was listed again in Germany, should be close to launch.

Has there been any confirmation about the shader count of GP106-300-A1?

 

[Mondozei]

In hindsight that is absolutely correct, but the fact is the 980 cost 550$ at the time, so a 980 cut in half for 200$ seemed OK for most casual users.

It was obvious even at launch. A 280X was a much better buy from day one. Good on you for realising it from the start, but far too many people did not and insisted in random arguments revolving around watt usage when the performance was just not there, ever.

This time it seems AMD is doing worse with the 480. The 460 is also doing worse than the 950.

This is a change, since AMD has done better in the last few years in the sub-$250 segment, largely thanks to a very strong legacy of the 280X and some crazy firesales of the 290.

 

[Sonikku]

Was doing WoW invasions last night and performance dropped to 12fps... Come on 1060. An i5 2500 isn't that much of a slouch.

 

[poofyhairguy]

This time it seems AMD is doing worse with the 480.

What is wrong with a 480? There isn't a better card for less than $250.

 

[antihelten]

What is wrong with a 480? There isn't a better card for less than $250.

Whilst I wouldn't really say that there is anything wrong with the 480 as such, but one could argue that it is just as good/bad as the 960.

The 960 had essentially the exact same perf/$ as the 970 when it launched, which was generally somewhat poorly received, since we expect to get better perf/$ as we go down the product stack (and vice versa), plus you had the 280X which provided slightly better perf/$.

The 480 also has essentially the same perf/$ as the 1070, but this time it's Nvidia that has the better competing product in the form of the 1060 which has slightly better perf/$ than the 480.

So the roles have basically been swapped between AMD and Nvidia, with the 480 taking the place of the 960. The one notable difference is that this time around the competing card (1060 vs. 480) doesn't come with the caveat of worse efficiency (280X vs 960), so in that sense the 480 is doing worse than the 960 relatively speaking (if you care about efficiency).

 

[Erenhardt]

Whilst I wouldn't really say that there is anything wrong with the 480 as such, but one could argue that it is just as good/bad as the 960.

The 960 had essentially the exact same perf/$ as the 970 when it launched, which was generally somewhat poorly received, since we expect to get better perf/$ as we go down the product stack (and vice versa), plus you had the 280X which provided slightly better perf/$.

The 480 also has essentially the same perf/$ as the 1070, but this time it's Nvidia that has the better competing product in the form of the 1060 which has slightly better perf/$ than the 480.

So the roles have basically been swapped between AMD and Nvidia, with the 480 taking the place of the 960. The one notable difference is that this time around the competing card (1060 vs. 480) doesn't come with the caveat of worse efficiency (280X vs 960), so in that sense the 480 is doing worse than the 960 relatively speaking (if you care about efficiency).

I feel like you bend the reality just a bit more than it is necessary
480 offers 25% better perf/$ than 1070 (or 50% if you look at 4GB version, which you should't)

While 960 was behind 280x and 290:

So it is very much not the same situation

 

[AtenRa]

I feel like you bend the reality just a bit more than it is necessary
480 offers 25% better perf/$ than 1070 (or 50% if you look at 4GB version, which you should't)

And those TPU slides include 10 GameWorks games and only a single DX-12 (RoTR). If we look again perf/$ at the end of 2016 or early 2017, things will be completely different with all those DX-12 games.

 

[antihelten]

I feel like you bend the reality just a bit more than it is necessary
480 offers 25% better perf/$ than 1070 (or 50% if you look at 4GB version, which you should't)

snip...

While 960 was behind 280x and 290:

snip...

So it is very much not the same situation

My estimate was based on MSRPs of $240 and $380 respectively, with which the RX 480 ends up as having only 3% better perf/$ than the 1070 at 1440P. The fact that the 1070 isn't really available at MSRP at the moment obviously skews this a bit, but then again the 970 wasn't really available at MSRP for quite a while after launch either.

The 960 may have been behind the 290 and the 280X, but only by about 5%, which is the same amount that the RX 480 falls behind the 1060 (this isn't reflected in TPUs current perf/$ graphs since they are for some weird ass reason setting the price of the 1060 to $300 even though it is readily available at it's MSRP of $250).

So the RX 480/1060 situation is very much the same as the 960/280X situation (with the notable of difference of the 1060 beating the 480 in both perf/$ and efficiency). The 480/1070 situation is arguably a bit different from the 960/970 situation, but only because the 1070 is currently hard/impossible to find at its MSRP.

 

[Bacon1]

The 480/1070 situation is arguably a bit different from the 960/970 situation, but only because the 1070 is currently hard/impossible to find at its MSRP.

Might as well quote the 480 @ $200 then since neither are available at those prices.

 

[antihelten]

Might as well quote the 480 @ $200 then since neither are available at those prices.

Tell you what, we'll just go with the lowest prices currently available from Newegg. That put's the 1070 at $425 and the RX 480 at $400 (all lower priced 480s are currently out of stock).

So with current prices from Newegg the perf/$ of the 480 is 30% worse than the 1070.

 

[Thala]

And those TPU slides include 10 GameWorks games and only a single DX-12 (RoTR). If we look again perf/$ at the end of 2016 or early 2017, things will be completely different with all those DX-12 games.

You can already look today. Issue is that TPUs set of benchmarks consists of overwhelmingly many GameWorks titles as you pointed out. If you take for instance the benchmark suite of computerbase.de, the GTX1070 come out only 40% faster while being almost 60% more expensive. Of course that will get worse as more DX12 titles enter the sample set.

 

[Bacon1]

Tell you what, we'll just go with the lowest prices currently available from Newegg. That put's the 1070 at $425 and the RX 480 at $400 (all lower priced 480s are currently out of stock).

So with current prices from Newegg the perf/$ of the 480 is 30% worse than the 1070.

Oh come on, you can't compare sold by 3rd party overpriced as the standard price. No one would buy that and for good reason, it is over $100 over "overpriced" and $150+ over retail.

Only use prices directly sold from Amazon / Newegg, not 3rd party sellers

 

[antihelten]

Oh come on, you can't compare sold by 3rd party overpriced as the standard price. No one would buy that and for good reason, it is over $100 over "overpriced" and $150+ over retail.

Only use prices directly sold from Amazon / Newegg, not 3rd party sellers

There aren't any 480s sold directly from Amazon or Newegg, they are all out of stock, that's the whole point. The $400 480 I linked is literally the only 480 currently available on Newegg

If you prefer you can also find a $325 RX 480 on Amazon from a 3rd party seller (with some rather hefty shipping costs). That would put the RX 480 as having only 15% worse perf/$ than the 1070.

Edit: As I was writing this a $300 RX 480 sold directly by Newegg just came into stock. That would put the RX 480 as having 7% worse perf/$ than the 1070.

 

[Thala]

There aren't any 480s sold directly from Amazon or Newegg, they are all out of stock, that's the whole point. The $400 480 I linked is literally the only 480 currently available on Newegg

If you prefer you can also find a $325 RX 480 on Amazon from a 3rd party seller (with some rather hefty shipping costs). That would put the RX 480 as having only 15% worse perf/$ than the 1070.

Edit: As I was writing this a $300 RX 480 sold directly by Newegg just came into stock. That would put the RX 480 as having 7% worse perf/$ than the 1070.

Of course you can construct any argument, when choosing from the worst set of benchmarks (TPU) combined with the highest possible prices for RX480. But even TPU was not as crazy to put GTX1070 in front of RX480 perf/$ wise despite their skewed benchmarks.

 

[antihelten]

Of course you can construct any argument, when choosing from the worst set of benchmarks (TPU) combined with the highest possible prices for RX480. But even TPU was not as crazy to put GTX1070 in front of RX480 perf/$ wise despite their skewed benchmarks.

TPU is not even remotely skewed for the RX 480. They have it at 66% the performance of the 1070, which is pretty much the average of sites out there (3Dcenter got 65% across 15 different sites).

And I'm quite literally using the lowest possible prices that are currently available.

But obviously unless you use sites that show AMD in a better light than the average out there and use prices that are not currently available, you are biased against AMD, right?

 

[Thala]

TPU is not even remotely skewed for the RX 480. They have it at 66% the performance of the 1070, which is pretty much the average of sites out there (3Dcenter got 65% across 15 different sites).

And I'm quite literally using the lowest possible prices that are currently available.

If you use actual retail prices, that is going to be expected due to market dynamics. The product with better relative value will be pushed up in price and the product with worse relative value will be pushed down. So nothing to see here.
Either you use MSRP or leave it if you want to make any tangible and relevant statement, because the actual retail situation changes as we speak.

But obviously unless you use sites that show AMD in a better light than the average out there and use prices that are not currently available, you are biased against AMD, right?

That is not what i have said. You should first decide if the set of benchmarks is a reasonably unbiased reflection independent of the average of different review sites. The fallacy here is the assumption that if you take into account more review sites bias is minimized. Bias is not necessarily distributed with expectation 0.

 

[Bacon1]

And I'm quite literally using the lowest possible prices that are currently available.

And amazingly the price has fallen 25% just today. Going by your logic of pricing the card should be free for anyone purchasing it by this weekend.

Just because the cards are selling out almost instantly doesn't mean you can use third party pricing as MSRP.

 

[antihelten]

If you use actual retail prices, that is going to be expected due to market dynamics. The product with better relative value will be pushed up in price and the product with worse relative value will be pushed down. So nothing to see here.
Either you use MSRP or leave it if you want to make any tangible and relevant statement, because the actual retail situation changes as we speak.

I was using MSRP for both cards to begin with, but apparently that wasn't good enough for some of the posters here, so then I used the lowest in stock price for both cards instead, but apparently that isn't OK either.

Apparently the only acceptable thing is to use MSRP for AMD and lowest in stock price for Nvidia.

That is not what i have said. You should first decide if the set of benchmarks is a reasonably unbiased reflection independent of the average of different review sites. The fallacy here is the assumption that if you take into account more review sites bias is minimized. Bias is not necessarily distributed with expectation 0.

You're absolutely right that you should first determine if a set of benchmarks is biased/skewed, but you seem to forget that you were the one who claimed that TPUs numbers were skewed with zero evidence to back up that claim.

And amazingly the price has fallen 25% just today. Going by your logic of pricing the card should be free for anyone purchasing it by this weekend.

Just because the cards are selling out almost instantly doesn't mean you can use third party pricing as MSRP.

"My logic"? What the hell are you talking about? I made no claims about prices continuously falling. Two data points does not a trend make.

And I never used third party pricing as MSRP, I used MSRP as MSRP and the lowest in stock price as the lowest in stock price. I never claimed that the $400, $325 or $300 prices were MSRP.

Seriously if all you have left is strawman attacks them I'm not going to bother responding anymore.

 

[Headfoot]

So the roles have basically been swapped between AMD and Nvidia, with the 480 taking the place of the 960. The one notable difference is that this time around the competing card (1060 vs. 480) doesn't come with the caveat of worse efficiency (280X vs 960), so in that sense the 480 is doing worse than the 960 relatively speaking (if you care about efficiency).

The 960 was never bad because of the 970 or its absolute performance or features. It was always because it cost too much money, and the 50+% faster 290 was only $50 more. That's the difference between MSRP 480 4GB and MSRP 1060 6GB, except the 290 had double the VRAM massively more performance where the 1060 today costs the same amount more but its only a bit faster with a bit more VRAM -- which is more normal for competing cards.

The 960 4GB was such an epic turd pretty much because it was only $50 less than a vastly better card.

 

[Bacon1]

I was using MSRP for both cards to begin with, but apparently that wasn't good enough for some of the posters here, so then I used the lowest in stock price for both cards instead, but apparently that isn't OK either.

Apparently the only acceptable thing is to use MSRP for AMD and lowest in stock price for Nvidia.

Who does this? It is incorrect. You should use the actual manufacturer MSRP or actual selling price directly from Amazon / Newegg and not small time shops massively overcharging. You trying to say the 480 costs $400 was completely ridiculous.

"My logic"? What the hell are you talking about? I made no claims about prices continuously falling. Two days points does not a trend make.

I was trying to point out how insane your pricing was.

And I never used third party pricing as MSRP, I used MSRP as MSRP and the lowest in stock price as the lowest in stock price. I never claimed that the $400, $325 or $300 prices were MSRP.

https://www.nowinstock.net/computers/videocards/amd/rx480/

You seem to have ignored the $240 8GB version that sold out at amazon earlier today

Or even the $280 Red Devil from newegg

The $250 XFX OC'd w/ backplate was also in stock earlier but yet you only found the $300 one because it fit your narrative better.

 

[antihelten]

The 960 was never bad because of the 970 or its absolute performance or features. It was always because it cost too much money, and the 50+% faster 290 was only $50 more. That's the difference between MSRP 480 4GB and MSRP 1060 6GB, except the 290 had double the VRAM massively more performance where the 1060 today costs the same amount more but its only a bit faster with a bit more VRAM -- which is more normal for competing cards.

The 960 4GB was such an epic turd pretty much because it was only $50 less than a vastly better card.

Cheap 290s were without a doubt the biggest spoiler for the 960 and a similar spoiler doesn't exist for the 480 currently. However cheap 290 were generally only available in the US, and even then only somewhat sporadically.

Who does this? It is incorrect. You should use the actual manufacturer MSRP or actual selling price directly from Amazon / Newegg and not small time shops massively overcharging. You trying to say the 480 costs $400 was completely ridiculous.

I was trying to point out how insane your pricing was.

I'm not the one saying the 480 costs $400, Newegg is, since that was the only card in stock when I looked.

And of course the price is insane, but that doesn't change that it was also the only one available, so if I can't use MSRP then that price was the only one available.

https://www.nowinstock.net/computers/videocards/amd/rx480/

You seem to have ignored the $240 8GB version that sold out at amazon earlier today

Or even the $280 Red Devil from newegg

The $250 XFX OC'd w/ backplate was also in stock earlier but yet you only found the $300 one because it fit your narrative better.

I "ignored" them because they were all sold out at the time I made my post I made my post, I don't understand why this is such a hard concept for you to understand. They may very well have popped into stock for a short while after I made my post, all I know is that they weren't in stock when I made the post, and they aren't in stock now.

And remember you were the one who first complained about using prices that weren't available, so the only reason why I started using those prices was because of you, but apparently you suddenly changed you're mind, and now it's okay to use prices that aren't available, at least as long as we're talking about AMD.

 

[Erenhardt]

Tell you what, we'll just go with the lowest prices currently available from Newegg. That put's the 1070 at $425 and the RX 480 at $400 (all lower priced 480s are currently out of stock).

So with current prices from Newegg the perf/$ of the 480 is 30% worse than the 1070.

I swear I went trought newegg prices after your replay in #2059 and noticed all cards out of stock and only a single $400 480 listing. Wanted to troll you a bit by linking it here to support your claim of superior perf/$ of nv GPUs, just for giggles, no one would take it seriously.

But yet, here we are... comparing $400 rx 480 to other cards.

Hilarious!

Cheap 290s were without a doubt the biggest spoiler for the 960 and a similar spoiler doesn't exist for the 480 currently. However cheap 290 were generally only available in the US, and even then only somewhat sporadically.

Not true. 290 was cheap worldwide. And there was a huge used market at a bargain prices after mining bubble.

 

[mohit9206]

A 3GB card in 2016? What were NVidia thinking?

EDIT: To think they have the audacity since the 1060 has 980-performance level.

I don't understand why do people have problem with Nvidia releasing a $200 card with 3GB in 2016?
Need i remind you that AMD launched R9 380 with only 2GB in May 2015?
May 2015 - AMD launches 2GB card for $180 with 4GB model at $220
Aug 2016- NV launches 3GB card for $200 with 6GB model at $250
Seems pretty similar strategy by both AMD and Nvidia here.

 

[Erenhardt]

2GB cards were DOA since 2014.
We are now at the edge of calling 4GB rx470 memory crippled and we recommend 8GB models to be on the safe side.
Yesterday I tested 8GB 480 vs 4gb 290 in ROTR. On max settings it was a boodbath for 290 as it spilled from VRAM to RAM, and then to HDD (8GB system memory).
Dropping textures to high fixed memory usage (down to less than 3GB WTF?).

If we get more clunky dx12 implementations like the one in ROTR, expect VRAM requirements to skyrocket, or prepare to drop texture quality.

Also, fun fact:
Dropping texture quality did almost nothing for 480 performance wise, while 290 improved something like 50% . Which shows that texture quality is almost free as long as you have enough VRAM to support it. You can't overpay VRAM and arguing the GPU core is to slow to use all that VRAM is nonsense when it is used for high quality textures and not ultra rendering resolution with demanding AA settings.