NVIDIA GeForce RTX 2060 Super

[xorbe]

Anyone remember the good ol' days?
GTX 660 $214
GTX 760 $259
GTX 960 $232
GTX 1060 $250

GTX 1660 $210 (pcpartpicker)
GTX 1660 Ti $265 (pcpartpicker)

RTX 2060 Super $400 <= not really in the GTX x60 family, just sayin' ...

 

[Turmania]

We need same system specs to compare new AMD cards with these. We can't compare different configurations and make conclusions about it. In reality the games I play have been Intel/Nvdia favoured games but that issue is for me to look at. There is a lot of unknowns so far so best bet is to wait for each tech side to do their tests and compare when AMD launches their cards. Then we can talk about performance, Price and power consumption figures.

 

[raptori]

Thank you for the review.

"Overclocking the RTX 2060 Super is a breeze, with a massive 29 percent GPU overclock (+200 MHz) without tinkering with the voltages, and an easy-peasy memory overclock to 16 Gbps. "

I think there is a typo : the massive 29% is for the memory OC , the GPU OC is +200MHz which = ~ 13%.

 

[tajoh111]

Anyone remember the good ol' days?
GTX 660 $214
GTX 760 $259
GTX 960 $232
GTX 1060 $250



yeah every NV review from various sites seem to just give negreedia a pass for this these days. I wonder when people will start voting with their wallet? People complain about prices going up but still buy nv products anyway

Lol how come there isn't a peep out of you about Radeon 5700xt or RX590 pricing?

As far as your excuse that AMD cards don't see well when they undercut, they sell incredibly well to the point where they were able to obtain nearly 50% marketshare going by the 4870 and 5870 successes.

The problem is AMD does not undercut at launch and only take a reactionary measure when it is too late. That is they kill the marketing opportunity of reviews by launching at the same pricing as Nvidia cards which they tend to lose to in reviews at launch.

The RX 570 is not selling well because the market has become too saturated from the length of the polaris/pascal generation. For most of it's life, the RX 570 has been massively overpriced because of mining which prevented it from being a gaming card which is why it does so badly in steam hardware survey. The RX 570 was priced $300+ during the mining crisis which turned cards like the GTX 1060 and even the 1070 the only options for many gamers. During this time, Nvidia was able to scale production mostly to meet demand and this allowed them to take AMD gaming marketshare on top of their own. This on top of selling some of the above cards to miners allowed them to create record profits which were normally not obtainable. Since the GTX 1050 ti series was not effected by the mining craze, unlike the GTX 1060, it's marketshare was allowed to flourish during the mining craze. On top of this, the GTX 1050 ti is amoung the most popular gaming GPU's in laptops because of its performance and power consumption, AMD had nothing comparable. Considering new laptops outsell desktops nowadays, this is a tremendous advantage. Similarly huge desktop system builders like Dell, lenovo and HP use 300-350watt power supplies, which due to this increases the likelihood to see a gtx 1050 ti in such a system(these companies don't get game bundles).

By the time the RX 570 fell to the current steal pricing we see today, the market has moved on and people desiring this level of performance has purchased a GTX 1060/1070 or even a GTX 1050 ti. Basically there were no purchasers left because pascal/polaris has been on the market for 3+ years. This is why the card is not selling very well today even at todays pricing. Nvidia cards are equally effected by the consequences of the post mining boom and still sit on hundreds of millions of pascal inventory.

 

[bug]

Yeah I also remember how these cards barely made a jump in performance until we saw the 1060 - and only the 6GB at that.

660, 760 and 960 were pretty bad cards, and the 2060Super now joins that fantastic line up as another weak price/perf option. Its just not high end, and its just a bit too expensive.

But its wrong to compare Turing line up by model name. And not just Turing either. We ARE entering the realm of diminishing returns by now, and 2060-70-80 are closer together than ever before.

True, I held on to my 660Ti until I switched to the 1060. The other mid rangers were too meh.
And while I can see why people love their mid range cards to remain in the same price segment (the 1660Ti is still $250), you have to account for inflation (Kepler was over 7 years ago). For those who remember, AAA games used to be $50 for years; not anymore.
So don't be surprised when prices go up a bit, but rather be thankful when they don't.

 

[Dristun]

Inflation over last 7 years in the US is ~11% + costs of manufacturing everything (not just electronics) in China have also increased, wages there are steadily rising, etc.
Though it's slightly offset by a more favourable usd/yuan exchange rate.
But anyway - there you go, what cost 250$ in 2012 should now cost at least 280$, which, coincidentally, is the price of 1660ti (yknow, the card with a comparable die size to 560/560ti).

 

[Totally]

This card is a bit lacklustre. They delivered with the 2070s but fell short quite a bit with this one. It's too close in pricing to a regular 2070 which dulls the appeal of the 2060s quite significantly.
For context the 2080 outperforms the 2070s by 7-10% and costs $200 dollars more were only $80 separates the 2060s and the 2070.

 

[eltano06]

i like how nvidia trolls every asshole who bought and non super rtx 2060/2070. (me included - 2060 user).

 

[Fluffmeister]

Definitely, once you buy a card, the market isn't allowed to move forward until you wish it.

 

[Vayra86]

i like how nvidia trolls every asshole who bought and non super rtx 2060/2070. (me included - 2060 user).

But... its standard operating procedure, and it was very clear Turing wasn't the optimal price/perf. Though your 2060 still is one of the highest bang/buck cards in the range.

 

[John Naylor]

So ...

13.64% faster than 2060 FE outta the box @ 1440p, 12.44 % faster when both OC'd.
12.36% faster than Vega 64 outta the box @ 1440p,
96.15% as faster than reference 2060 outta the box @ 1440p

Can't talk about relative price as yet as no 2060 FE's around these days; will have to wait will AIB versions arrive before any of this is really worth talking about.

Thank you for the review.

"Overclocking the RTX 2060 Super is a breeze, with a massive 29 percent GPU overclock (+200 MHz) without tinkering with the voltages, and an easy-peasy memory overclock to 16 Gbps. "

I think there is a typo : the massive 29% is for the memory OC , the GPU OC is +200MHz which = ~ 13%.

Both core and memory Ocs are basically irrelevant since boost 3 ... what good is a 13 or 29 % OC when the fps increase is the only number that counts. This is significant that: 1) fps rarely proportional to the memory of GPU increase and 2) the highest core / memory never gets the highest fps in TPU testing. I'm guessing that the higher voltage needed to maintain those core and memory settings I imagine this is why undervolting has become a 'thng" of late. You can confirm this by looking at the TPU OC pages, the blue and white table lists the comparative core and memory OCs, but the highest core and highest memory never get the highest fps.

As far as the performance and cost comparisons, faulty reasoning as a) real costs aren't here yet, b) 5700's aren't here yet and have not been tested on same site on same box c) historically, at least since 2xx, AMD very aggressively clocks their card in the box and d) , we have not seen the AIB cards yet. I think I will wait till all of these happen before drawing any firm conclusions.

As for the future-proofing, I agree, that's a perception that doesn't sync very well w/ actual testing.

 

[raptori]

Both core and memory Ocs are basically irrelevant since boost 3 ... what good is a 13 or 29 % OC when the fps increase is the only number that counts. This is significant that: 1) fps rarely proportional to the memory of GPU increase and 2) the highest core / memory never gets the highest fps in TPU testing. I'm guessing that the higher voltage needed to maintain those core and memory settings I imagine this is why undervolting has become a 'thng" of late. You can confirm this by looking at the TPU OC pages, the blue and white table lists the comparative core and memory OCs, but the highest core and highest memory never get the highest fps.

As far as the performance and cost comparisons, faulty reasoning as a) real costs aren't here yet, b) 5700's aren't here yet and have not been tested on same site on same box c) historically, at least since 2xx, AMD very aggressively clocks their card in the box and d) , we have not seen the AIB cards yet. I think I will wait till all of these happen before drawing any firm conclusions.

As for the future-proofing, I agree, that's a perception that doesn't sync very well w/ actual testing.

Are you lost or something ?!! I was talking simply about a typo in the review , in the OC test page the 29% OC value was for the memory and on the conclusion page I think W1zzard mentioned that by mistake as a GPU OC value.

 

[Zareek]

Ngreedia strikes again!!!

 

[jmcosta]

It seems that Nvidia and AMD are gonna try to price fixing once again, which is a shame but thats how these corporations work. hopefully Intel will put a stop to it with the upcoming GPU

 

[laszlo]

both launched super version seems good question is who will buy them once they'll be forced to cut prices of previous generation

those who own 1080ti can still wait at least 2 years from now to upgrade...

 

[gridracedriver]

Good job for having updated the data of the old cards, in fact on famous sites like anandtech it has not been done and this shows the 2060S higher than 2070 which is technically impossible to do.
But as I said the other day there are some really badly optimized games on amd that kill the overall average of everything, from perf / watt to perf / cost to the summary performances, and it's a shame.

 

[londiste]

Good job for having updated the data of the old cards, in fact on famous sites like anandtech it has not been done and this shows the 2060S higher than 2070 which is technically impossible to do.

Why would it be technically impossible? 2070 and 2060S are close enough and due to higher frequencies 2060S actually minimally exceeds 2070 in some aspects like anything ROPs are doing.

 

[gridracedriver]

Why would it be technically impossible? 2070 and 2060S are close enough and due to higher frequencies 2060S actually minimally exceeds 2070 in some aspects like anything ROPs are doing.

The clocks are practically the same, 2060S is a 2070 cut, it has the same characteristics except precisely 128cc.

Average clock 2070 = 1860mhz

2060S is 1840mhz

, so tell me how it can go the same or even stronger as seen in anandtech or guru3d?

In fact, here on TPU and on HU italia it does not happen, it is 5% below as it is obvious.

but then you never realized that in the first 100 seconds it clocked at 1900mhz and then gradually went towards the 1800mhz? is it from pascal that they do so, specifically for reviews and doing a few more fps in the benches?

eheh, in nvidia are smart, not like amd ...

 

[londiste]

Different reviews, different circumstances?

Guru3D doesn't seem to show the clock speeds but Anandtech sees Super cards running at noticeably higher clocks:

The NVIDIA GeForce RTX 2070 Super & RTX 2060 Super Review: Smaller Numbers, Bigger Performance

www.anandtech.com

 

[bug]

Definitely, once you buy a card, the market isn't allowed to move forward until you wish it.

Yeah, the world will probably end if you bought a card and bam! a year later something faster was released.

 

[W1zzard]

is it from pascal that they do so

Yes, this was introduced with Pascal, I don't think it was changed for Turing.

specifically for reviews and doing a few more fps in the benches?

While I think this positive outcome for reviews is just a side-effect, it will affect many reviewers indeed, because they will start the benchmarks with a cool card, run a short test, and immediately record results. for my reviews i always include a warm up period in every test.

AMD does the same clocking, too, but to a lesser degree: https://www.techpowerup.com/review/amd-radeon-vii/34.html

Navi data for this will be interesting

 

[gridracedriver]

Yes, this was introduced with Pascal, I don't think it was changed for Turing.


While I think this positive outcome for reviews is just a side-effect, it will affect many reviewers indeed, because they will start the benchmarks with a cool card, run a short test, and immediately record results. for my reviews i always include a warm up period in every test.

AMD does the same clocking, too, but to a lesser degree: https://www.techpowerup.com/review/amd-radeon-vii/34.html

Navi data for this will be interesting

No, the working between amd and nvidia is not the same, those 20 mhz with the VII is the normal oscillation of the clock based on the management of the load and the turbo, the only card that had a similar operation made by amd is the Fury R9 Nano, where he had a maximum clock of 1000mhz to stay within 175watts, in fact the clocks were often lower during normal use, around 940mhz of average.
R9 Nano average clock:

I am pleased that you are using a "warm up" quote, this should be enough to make the bench real.

 

[W1zzard]

No, the working between amd and nvidia is not the same, those 20 mhz with the VII is the normal oscillation of the clock based on the management of the load and the turbo, the only card that had a similar operation made by amd is the Fury R9 Nano

Yeah, you are right. My mistake

R9 Nano average clock:

In case anyone wonders why they are so smeared out. It's because AMD reports averages over a short duration, whereas NVIDIA reports non-averaged values.

 

[c12038]

How many of you guys have shelled out a lot of money for the Series 10 range and RTX range now Nvidia have stuck this under your noses and said look newer cards come buy …..Erm NO why should I buy well you see what they don't tell you is they would prefer you to buy a new card once every year when new technology comes out Marketing tactics....

RTX range when first released was £1000 just shy of in the UK
10 Series top of the range still going strong £1,200 Just Shy of in the UK

and same goes for AMD they are just as guilty

Your paying now for what was called the Bitcoin gold rush


Start showing these companies your not going to stand being ripped off by using you spending power DON@T buy in to the biggest con yet......

 

[Aquinus]

So basically a rebranded 2080.