MSI GeForce RTX 3090 Suprim X

[R0H1T]

But there Is real data and it says you are wrong every time. Sources were provided, did you even get into them or? Do you read reviews? We know how efficient the arch and the node is, the products are out... We know Ampere has more efficient RT, but less efficient raster. We know the end FPS, bandwidth/memory/power usage, we know die size and we know the architectural improvements. If you still didn't figure it out by now, you never will. Your opinion is irrelevant, the facts are there. Its like a graph with lots of dots, you just need to connect them with straight lines and you've got your comparison well laid out.

What real data? Do you have reviews of Ampere's consumer lineup on 7nm TSMC?

Sources like? Do you read reviews, actual reviews?

No we only know how efficient the node is for the products that are out, we have "0" data on how (much more or less) efficient it is for the same chip on SS' 8nm.

Yeah I'm not in the business of extrapolating such complex data to form a conclusion based on sometimes vast uarch differences, heck the differences between chips on the same node can vary greatly!

Graphs with what exactly? TSMC could be 5% more efficient for Ampere or 35% ~ what you & many others are doing is complete BS! Let's check the last real world data we had for the same chip on similar nodes ~

Conclusion

Based on the results of our testing, it's clear that both versions of Apple's A9 SoC deliver the same level of performance, but Samsung's 14nm FinFET process appears to offer slightly better power efficiency, extending battery life between 3.5-10.8 percent. This is a little more than the 2-3 percent quoted by Apple, but not much, and it equates to only about 5-15 minutes of runtime under the most extreme conditions.

iPhone 6s: Samsung And TSMC A9 SoCs Tested

The A9 SoC inside the iPhone 6s comes from two different vendors (Samsung and TSMC) using two different FinFET processes (14nm and 16nm, respectively). We test both versions to see if there's any power or performance differences.

www.tomshardware.com

Next time don't bring waffles to a biscuit party

 

[Vayra86]

What real data? Do you have reviews of Ampere's consumer lineup on 7nm TSMC?

Sources like? Do you read reviews, actual reviews?

No we only know how efficient the node is for the products that are out, we have "0" data on how (much more or less) efficient it is for the same chip on SS' 8nm.

Yeah I'm not in the business of extrapolating such complex data to form a conclusion based on sometimes vast uarch differences, heck the differences between chips on the same node can vary greatly!

Graphs with what exactly? TSMC could be 5% more efficient for Ampere or 35% ~ what you & many others are doing is complete BS! Let's check the last real world data we had for the same chip on similar nodes ~

iPhone 6s: Samsung And TSMC A9 SoCs Tested

The A9 SoC inside the iPhone 6s comes from two different vendors (Samsung and TSMC) using two different FinFET processes (14nm and 16nm, respectively). We test both versions to see if there's any power or performance differences.

www.tomshardware.com

Next time don't bring waffles to a biscuit party

You need to bring ARM to prove something in an x86 GPU comparison? Grasping at straws.

 

[R0H1T]

What does ARM (ISA) have anything to do with it & x86 GPU really? You said we have "real world data" such as?
The data I provided is real world for the exact same chip on two different nodes, if you don't like facts then please don't present your assumptions as "graphs with lots of dots" that show exactly how efficient 7 nm is!

 

[the54thvoid]

The passion is great in this discussion but seriously, stop derailing the review thread. Discuss the card--pros and cons--without going into petty squabbles. This isn't the playground.

 

[kapone32]

80'c ? pretty bad. ASUS RTX 3090 STRIX OC = 68'c

Yes but not pulling 470+ watts.

 

[lexluthermiester]

474w peak in games
....oooookay then
fermi 2.0

Except that that this card is a HUGE jump in performance over the RTX2000 series cards, something that didn't happen with the Fermi gen cards. So not really. Yes, this generation of cards eat a ton of power but there is excellent performance trade off.

MSI pulled out all the stops with this card. It is a freaken beast of a card in physical size, performance and quality of build! Impressive!

As always, great review W1zard!

 

[Nima]

Oh wow, how the mighty have fallen! Don't remember such a massive dip in perf/W with a new uarch+node since ~ well perhaps ever

Power consumption is only measured in 4k and the obtained number is also used to calculate performance per watt for 1080p and 1440p resolutions. RTX 3090 is not fully utilized in lower resolutions and consume much less power than 4k so the actual efficiency is much higher than what these charts are suggesting.

In my opinion 1080p and 1440p numbers should not be included in performance per watt page. showing these charts is just pointless and can only misinform people.

 

[ratirt]

This card is not my thing though. Somehow people are justifying the power consumption by better performance. I think there's boundaries in that regard and this card has crossed them definitely.

 

[Sovsefanden]

Except that that this card is a HUGE jump in performance over the RTX2000 series cards, something that didn't happen with the Fermi gen cards. So not really. Yes, this generation of cards eat a ton of power but there is excellent performance trade off.

MSI pulled out all the stops with this card. It is a freaken beast of a card in physical size, performance and quality of build! Impressive!

As always, great review W1zard!

Yes this is NOTHING like Fermi

Ampere HAS the performance

 

[spnidel]

Except that that this card is a HUGE jump in performance over the RTX2000 series cards, something that didn't happen with the Fermi gen cards. So not really. Yes, this generation of cards eat a ton of power but there is excellent performance trade off.

MSI pulled out all the stops with this card. It is a freaken beast of a card in physical size, performance and quality of build! Impressive!

As always, great review W1zard!

it's fermi 2.0

 

[lexluthermiester]

it's fermi 2.0

And your comment makes as much sense.

 

[spnidel]

And your comment makes as much sense.

not here to argue, 500W is 500W

 

[lexluthermiester]

not here to argue, 500W is 500W

Yet that is exactly what you are doing, poorly. And as the testing showed a MAXIMUM of 482w under the HIGHEST load possible, your point isn't one. Additionally, this card is a literal top tier card that grants maximum performance. It's expected to have a high power draw. That does NOT make it Fermi 2.0. Such a comparison is has zero merit.

 

[ratirt]

Yet that is exactly what you are doing, poorly. And as the testing showed a MAXIMUM of 482w under the HIGHEST load possible, your point isn't one. Additionally, this card is a literal top tier card that grants maximum performance. It's expected to have a high power draw. That does NOT make it Fermi 2.0. Such a comparison is has zero merit.

I'm curious. If next year NV releases a card like 4090 and it will gobble 650W would you be OK with it if the "performance is there"? Is there a limit to power usage/performance for you?

 

[lexluthermiester]

I'm curious. If next year NV releases a card like 4090 and it will gobble 650W would you be OK with it if the "performance is there"? Is there a limit to power usage/performance for you?

I'll assume you're being serious and not a smart-alec. Given that SLI configurations of the past(Dual, Triple and Quad SLI) as well as CrossFire(also Dual, Triple and Quad) could easily top 750W, for people who want the best performance, 650W is a reasonable power level. If that is a problem for certain people, hard cheese, don't buy such a card. However, complaining about it is not going to stop people from buying it. The upcoming RX6900XT is rumored to be near 550W requirement. Is everyone going to whine about that too? Seriously, find something better to complain about.

 

[ratirt]

I'll assume you're being serious and not a smart-alec. Given that SLI configurations of the past(Dual, Triple and Quad SLI) could easily top 750W, for people who want the best performance, 650W is a reasonable power level. If that is a problem for certain people, hard cheese, don't buy such a card. However, complaining about it is not going to stop people from buying it. The upcoming RX6900XT is rumored to be near 550W requirement. Is everyone going to whine about that too? Seriously, find something better to complain about.

Oh I'm smart that's for sure but not like smart-alec. Honest question with that power usage. Just want to see how much people's perception of graphics card's power usage changed over the years.
It's hard to compare 2 cards vs one in performance per watt. If you do, you do realize that 2 cards eat up a lot of Watts and the performance gain is not that spectacular. Also, I though the SLI is dead at least a lot of people say it and SLI support supports that theory. So 650w for 2 cards or one is a great deal of Watt for me. Is that the boundary for you with a single card 650W? For me that's way overboard to be honest. I only hope, this is not a trend to get more performance. If it is and we end up with 1 cards sucking down over 1k Watt that would be a horrible thing and you would say innovation in that regard is dead just like SLI. Still there but nobody cares.
Rumors say strange things and 6900XT is spec'ed as 300W card. We'll know soon what the actually card power consumption is, the reviews are around the corner. 550W is simply not true for a stock card. Especially if you look at 6800XT's power draw.

 

[lexluthermiester]

It's hard to compare 2 cards vs one in performance per watt.

Except that today's GPU's leave past, high wattage, GPU's in the dust performance-wise. Perspective within context is important. Whether you have a single card giving top-tier performance or multiple cards doing the same is irrelevant. Again, if buyers don't want cards which use that much powere, no one is forcing them to buy them.

 

[ratirt]

Except that today's GPU's leave past, high wattage, GPU's in the dust performance-wise. Perspective within context is important. Whether you have a single card giving top-tier performance or multiple cards doing the same, is irrelevant. Again, if buyers don't want cards which use that much wattage, no one is forcing them to buy one.

Yeah the performance is there but I disagree with the concept that the only way to get more performance is to go overboard with the power consumption. This is not an advancement but rather pushing to the limits of inferior architecture.
Your argument about, don't want don't buy it is irrelevant here. The advancement in computing is not about raising the power over and over. If that's the case, that means the company is just going inefficient in terms of advancement. It's like it doesn't have a plan how to get it done which and it is slow and clumsy. We don't have anything to offer now so bump the price and release whatever.
AMD did it right though with Ryzen. The power consumption is lower or stays at the same level and the performance uplift is great.
I wonder, what would be the difference if you have locked the 2 gen NV cards with power 250Watts and compared the performance. Maybe someone did it already. How would that comparison look like. You may disagree but that's how I see it.

 

[spnidel]

Yet that is exactly what you are doing, poorly. And as the testing showed a MAXIMUM of 482w under the HIGHEST load possible, your point isn't one. Additionally, this card is a literal top tier card that grants maximum performance. It's expected to have a high power draw. That does NOT make it Fermi 2.0. Such a comparison is has zero merit.

nah, I'm not arguing, 500w is 500w, and these cards are fermi 2.0

 

[lexluthermiester]

This is not an advancement but rather pushing to the limits of inferior architecture.

Are seriously calling brand new, high performance designs "inferior". What a laughable notion!

Your argument about, don't want don't buy it is irrelevant here.

Your opinion, not supported by common sense.

The advancement in computing is not about raising the power over and over. If that's the case, that means the company is just going inefficient in terms of advancement. It's like it doesn't have a plan how to get it done which and it is slow and clumsy.

Once again, context is important. You are talking about brand new designs. Pushing them to their limits is not slow, lazy, clumsy or stupid. It is perfectly logical and is exactly what literally everyone has been doing since computers were invented.

We don't have anything to offer now so bump the price and release whatever.

How is it you don't get the way things work? Seems like you need to brush up on history just a little bit.

 

[ratirt]

Are seriously calling brand new, high performance designs "inferior". What a laughable notion!

Your opinion, not supported by common sense.

Once again, context is important. You are talking about brand new designs. Pushing them to their limits is not slow, lazy, clumsy or stupid. It is perfectly logical and is exactly what literally everyone has been doing since computers were invented.

How is it you don't get the way things work? Seems like you need to brush up on history just a little bit.

Listen. I'm not gonna argue with you because you take whatever companies bring before your eyes. If you look closer, Turing vs Ampere is not that different one to the other. So brand new, just because NV calls it that way and it has been released, doesn't mean it actually is brand new by design but by release date.
NV can say new RT cores, gen 2 RT cores or gen "whatever" tensor cores but you wont be able to verify the difference anyway and there's new node to consider which is not NV achievement. The sole clue for the performance is bump the number of cores and power limit and get a price bump. If that is how "the way things work" for you then you are a lucky man. hope it works out for you the same way in every other aspect.

 

[lexluthermiester]

Listen. I'm not gonna argue with you because you take whatever companies bring before your eyes.

And yet, you continue?

If you look closer, Turing vs Ampere is not that different one to the other.

You keep thinking that. Ignorance is bliss after all.

So brand new, just because NV calls it that way and it has been released, doesn't mean it actually is brand new by design but by release date.

So are you saying the same thing about AMD's RDNA2? Because such opinions would be silly and without merit of any kind.

NV can say new RT cores, gen 2 RT cores or gen "whatever" tensor cores but you wont be able to verify the difference anyway and there's new node to consider which is not NV achievement.

You're still arguing and failing to bring merit to your position.

If that is how "the way things work" for you then you are a lucky man.

Seemingly so.

Do you wish to "not" argue more?

 

[HenrySomeone]

474w peak in games
....oooookay then
fermi 2.0

It's not great, no, but just like with Fermi, they are the undeniably faster cards and in this price bracket that matters (far) more than a couple dozen W of power consumption. If Ampere was on the same 7nm TSMC node though, there would once again be absolutely no competition...however the already bad availability would likely reach epic levels...

 

[spnidel]

It's not great, no, but just like with Fermi, they are the undeniably faster cards and in this price bracket that matters (far) more than a couple dozen W of power consumption. If Ampere was on the same 7nm TSMC node though, there would once again be absolutely no competition...however the already bad availability would likely reach epic levels...

oh, so if nvidia does it, then power consumption doesn't matter? got it, thanks!

 

[ratirt]

It's not great, no, but just like with Fermi, they are the undeniably faster cards and in this price bracket that matters (far) more than a couple dozen W of power consumption. If Ampere was on the same 7nm TSMC node though, there would once again be absolutely no competition...however the already bad availability would likely reach epic levels...

I wouldn't be so sure about that. The 8nm Samsung and 7nm TSMC are fairly similar in density. There are pros and cons on both sides and NV went for better yields and lower price per wafer instead of slightly better power efficiency. Either way it would have been better for NV to go TSMC but saying absolutely no competition if they would have gone 7nm is over exaggerated.