What GPU are you using and whats your next GPU?

[Decathect]

I'm using the Evga 3060 XC, and I'm going to wait until the 40 series and try (try) to get a 4070 next year.

[Decathect]

引用自 ZeekAncient：

My 3070 Ti will crush a 2080 Ti.

Absolute truth.

2080ti isn’t coming near a 3070ti.

[ZeekAncient]

引用自 Monk：

Again I am saying a stock 2080ti beats a 3070 and is close to a 3080 under proper conditions.

A modified 2080ti is trading blows with a 3080.

Sorry, but this not what I have seen. I don't go by youtube videos, I go by every professional publication out there. On average, a 3070 is 6% more powerful than a 2080 Ti, and a 3070 Ti is 13% more powerful than a 2080 Ti.

Well, maybe if you have a 'modified' GPU. But I am not exactly sure what you mean by modified. I mean, if you do some extreme overclocking, and/or using liquid cooler, and having the power limits go as high as they can go, sure I am sure you can get your 2080 Ti to perform where you say it is performing.

But that isn't every 2080 Ti is it? If you modify and overclock the 3070 and the 3070 Ti in the same way, the results will be the same as if you where comparing stock GPUs. So the key term there is 'modified'. Not every 2080 Ti is going to be a 'modified' GPU.

[ZeekAncient]

引用自 Monk：

If a system us running the ever popular ddr4 3200 ram its losing out.

If the cpu isn't overclocked it cannot properly feed the gpu to the fullest.

Yeah maybe if you are just playing or testing at 1080p. At that point you are introducing CPU limitations. Go to 4K, and some cases even 1440p, and you will see setups with different memory speeds and CPUs begin to level out. Especially at 4K. This is the great CPU equalizer, and different CPU speeds and memory speeds will all produce the same results with a given GPU.

[r.linder]

A power modded 2080 Ti can crush a 3080, that's pretty much the key point of why Ampere is even as big of a power increase as it was, the combination of the efficiency increases from the die shrink, and the subsequent increase in power draw, is why the 3070 can even match a lot of stock 2080 Ti models. The 2070/S used substantially less power than the 3070.

Also, when Ampere was branched over to mobile, they were barely any better than Turing mobile GPUs, because they couldn't increase the power draw as it'd draw too much power from the battery and power supply for the laptop, and it'd run way too hot.

Power is the primary reason for Ampere being as good as it is, which is why a 2080 Ti with a shunt mod can match and beat a 3080, while it normally trades blows with the 3070 and 3070 Ti depending on the title, model, and OC. As long as you can cool it, you could do the same for a 3090 (or 3090 Ti) and get it running well above spec. That's pretty much all NVIDIA has done here, just expand the limits of how much power they can use.

[Monk]

引用自 ZeekAncient：

引用自 Monk：

Even throttling your video shows it easily beating the 3060ti, there's no 3070 but it'll be slightly ahead as the 3070/ti is a bad card compared to the 3060ti / 3080.

I have my 3070 Ti boosting to 2130 mhz during gaming. And I have it going even higher than that in benchmarks. Temps are usually in the 60sC. Honestly, the results I have with my overclocked 3070 Ti have it performing very close to a vanilla 3080.

Almost every publications I have seen has the 3070 and the 2080 Ti trading blows, with the 3070 winning more often than not. A 3070 Ti typically performs better than a 2080 Ti, in pretty much every benchmark I have seen. My 3070 Ti will crush a 2080 Ti.

Not a power modded one it won't and that's a nice clock, but given the 3090 tends to clock lower than the lower models (as does every top card model nvidia has released for years), as a pure example of what power modding and proper cooling will get you, my 3090 runs at 2190MHz during gameplay / benchmarking as at stock heat and power limit is a big issue for the 102 dies, always has been.

To gunstar.

No, overclocking is not hard, to do it properly is time consuming, to go the extra mile to watercool and power mod is an extreme option but that's where you get the full potential.

You may if been building oc's longer than me, I'm closing on 40, built my first pc at 12 or 13 but dudnt get into pc's properly until early / mid 20's, not that much from back then is relevant.

Again as I have explained, many reviewers do not test stuff under ideal conditions, they normalise stuff so they can have a big ol comparison list reaching back, it's why even gamers nexus gpu reviews are lacking as the system is held back to normalise stuff.

I sold my 2080ti over a year ago and have been running 3090's since, along with sli 1080ti's in my second system.

You clearly do not understand half of what I am talki g about as you somehow compared the power draw abd psu listed requirements if the 5700xt vs the 2080ti which coukdnt be less relevant.

I have explained the issues with the tests you have put forward along with the benefits of fully modding and running a card optimally vs stock.

As I said there are only a few on this forum that go as far as I do to get every little bit out of the card.

[ZeekAncient]

引用自 尺.ㄥ丨几ᗪ乇尺：

A power modded 2080 Ti can crush a 3080, that's pretty much the key point of why Ampere is even as big of a power increase as it was, the combination of the efficiency increases from the die shrink, and the subsequent increase in power draw, is why the 3070 can even match a lot of stock 2080 Ti models. The 2070/S used substantially less power than the 3070.

Also, when Ampere was branched over to mobile, they were barely any better than Turing mobile GPUs, because they couldn't increase the power draw as it'd draw too much power from the battery and power supply for the laptop, and it'd run way too hot.

Power is the primary reason for Ampere being as good as it is, which is why a 2080 Ti with a shunt mod can match and beat a 3080, while it normally trades blows with the 3070 and 3070 Ti depending on the title, model, and OC. As long as you can cool it, you could do the same for a 3090 (or 3090 Ti) and get it running well above spec. That's pretty much all NVIDIA has done here, just expand the limits of how much power they can use.

Again, the key term here is 'power modding'. Sure you start increasing the power limits on these GPUs, and they can perform a lot higher than their stock variants. But you can't make an argument saying that a 2080 Ti is better than a 3070/Ti because my power modded 2080 Ti will outperform a stock 3070/Ti.

[Monk]

Oh and the 3070/ti is no where near as restrained by power limits or heat as the 1080ti, 2080ti, or 3090 were/are at stock.

As I explained a stock 2080ti is around 250w, if you bios swap or shunt mod it it will pull 320 to 400w and can then fully stretch its legs, mind you this needs adequate cooling.

This is not a restraint the 3070ti you own has.

As the 3090 is what I've used most recently, I had 1 FE and 3 strix's to use. The FE capped at 320w stock. The strix a much higher 480w stock. Modified with an xoc bios the strix pulls upto 540w.

The same was true of the 1080ti and 2080ti, their stock design highly limited their potential, but as I said, this stuff is beyond pure enthusiast level and us not covered in most YouTube or websites as urs a very niche group willing to both watercool then either bios swap or physically mod their £1000+ cards.

[Monk]

引用自 ZeekAncient：

引用自 尺.ㄥ丨几ᗪ乇尺：

A power modded 2080 Ti can crush a 3080, that's pretty much the key point of why Ampere is even as big of a power increase as it was, the combination of the efficiency increases from the die shrink, and the subsequent increase in power draw, is why the 3070 can even match a lot of stock 2080 Ti models. The 2070/S used substantially less power than the 3070.

Also, when Ampere was branched over to mobile, they were barely any better than Turing mobile GPUs, because they couldn't increase the power draw as it'd draw too much power from the battery and power supply for the laptop, and it'd run way too hot.

Power is the primary reason for Ampere being as good as it is, which is why a 2080 Ti with a shunt mod can match and beat a 3080, while it normally trades blows with the 3070 and 3070 Ti depending on the title, model, and OC. As long as you can cool it, you could do the same for a 3090 (or 3090 Ti) and get it running well above spec. That's pretty much all NVIDIA has done here, just expand the limits of how much power they can use.

Again, the key term here is 'power modding'. Sure you start increasing the power limits on these GPUs, and they can perform a lot higher than their stock variants. But you can't make an argument saying that a 2080 Ti is better than a 3070/Ti because my power modded 2080 Ti will outperform a stock 3070/Ti.

You can though as one us highly restrained at stock the other is not and can get the results with a bios swap and alot of fans at high speed, though I'd recommend watercooling, we are talking about potential here.

The 3070/ti at stock is using all of its potential with no real gains to be had.

[Soulreaver]

Oh my comment created a discussion >.<

I don't have that much time rn so I try to make it short.

3070 bios flashed / 2080ti powermodded (starts at 3:40)
https://www.youtube.com/watch?v=u6wrsKZzIo4&t=416s

The powermodded 2080ti outperforms a bios flashed 3070. He said powermodding the 3070 wouldn't gain him much as it already ran at 2175Mhz. So it's relatively save to assume that this is as far as you can push a 3070 and therefore this should atleast kill the argument if a 2080ti at all can outperform a 3070
_____________________________________________________________

This is mine. Not at it's fastest as this was with my old PSU where I couldn't fully max it out but yeah. But there's not a lot more I gained by with the new PSU.
https://imgur.com/a/MXxFKc7
_____________________________________________________________

This might also be an indicator
https://www.youtube.com/watch?v=LOiBySoBF_U&t=167s
_____________________________________________________________

There's not a lot of bios flash/ powermodding content on youtube so sorry for not finding more appropriat videos as I'm in a rush.

I don't argue that a stock 2080ti can reach my GPU's performance. The 2080ti is just one of the most overclockable cards we had in the last few years. They left a lot of performance on the floor with it just to keep temps in check and to not have a lot of RMA's

What I should also mention. When I said between 3070ti and 3080 performance I was refering to their stock performances.
I didn't say mine is better. As better is rather subjective. If one doesn't like more extreme overclocking the 3070ti is obviously more powerfull. Didn't know this had to be said.

[Monk]

Cheers, that's the frame Chasers vid I didn't think to link as I was trying to find his 3080 version, but yeah that pretty much makes my point.

[ZeekAncient]

引用自 ZeekAncient：

This is mine. Not at it's fastest as this was with my old PSU where I couldn't fully max it out but yeah. But there's not a lot more I gained by with the new PSU.
https://imgur.com/a/MXxFKc7

Very good score. Nice!!

If you look on my profile, I think I had better runs with Firestrike Extreme and Ultra, and Timespy Extreme, than I did with my Timespy runs. Meaning that I had more constant clocks on those runs than my TImespy run I will post. And I also had a better graphics score on some of my other Timepsy runs, if I remember correctly. But this is one of my better 'overall' Timespy runs. This is with the CPU(10700K) at stock and my GPU fans are NOT at 100%, so I didn't maintain the best temps. But eventually when I OC my CPU I will rerun all the benchmarks with my GPU fans at 100%.

https://www.3dmark.com/3dm/66725329?

[Monk]

Thing to remember about benchmarks is they dont always line up with actual fane play, but, they are a good, well, benchmark :)

[Monk]

To back up my point, here's a 1080ti (sli) score trading blows with 3090 numbers.
https://www.3dmark.com/spy/2742994
Shame sli died off :(

[ZeekAncient]

引用自 Monk：

To back up my point, here's a 1080ti (sli) score trading blows with 3090 numbers.
https://www.3dmark.com/spy/2742994
Shame sli died off :(

Wow that is pretty sick for two 1080 Ti! Pretty amazing actually. I think.

I miss SLi as well. I didn't have two 1080 Ti, but I had two 1070s at one point. And I remember that the performance, when the game or application made use of it, was way better than just one 1080 Ti. Well, a stock one, lol.

I llke my 3070 Ti. I ended up with a good one. But originally I really wanted a 3080, but its all good.

[Soulreaver]

引用自 ZeekAncient：

引用自 ZeekAncient：

This is mine. Not at it's fastest as this was with my old PSU where I couldn't fully max it out but yeah. But there's not a lot more I gained by with the new PSU.
https://imgur.com/a/MXxFKc7

Very good score. Nice!!

If you look on my profile, I think I had better runs with Firestrike Extreme and Ultra, and Timespy Extreme, than I did with my Timespy runs. Meaning that I had more constant clocks on those runs than my TImespy run I will post. And I also had a better graphics score on some of my other Timepsy runs, if I remember correctly. But this is one of my better 'overall' Timespy runs. This is with the CPU(10700K) at stock and my GPU fans are NOT at 100%, so I didn't maintain the best temps. But eventually when I OC my CPU I will rerun all the benchmarks with my GPU fans at 100%.

https://www.3dmark.com/3dm/66725329?

Thanks yours seems also very good.
I'm not that knowledgeable with 3070ti OC performance :D
Your 3070ti clocks to 2160Mhz on air??? =O

Edit:
Haha ZeekAncient "plays" 3DMark right now. Someone wants to show us some numbers xD