RTX 3080 3D Mark Scores And Other Benches Leaked

[nick name]

I'm sure a lot of you have seen the leaks over at videocardz, but for those that haven't yet I'll post the link:

 

https://videocardz.com/newz/alleged-geforce-rtx-3080-graphics-card-test-leaks-online

 

Of interest to me are the Fire Strike and Time Spy graphics scores which are double my best scores with an overclocked 1070 Ti.  And that's precisely the sort of performance I was waiting for.  Unfortunately, the games that were benched I don't own so I can't compare directly against my own card. 

 

I was very unsure about buying a  2000 series card (so I never did), but if these 3080 numbers are real then I am almost certainly going to buy one.  

[xg32]

now turing users just need a trashcan, looking for that demand spike.if the 3070 does indeed come out faster than a 2080 ti, oh boy.

[VeganJoy]

Just now, DutchGuyTom said:

The Firestrike and TimeSpy scores are not double my scores though (2600x and 2070S). Since the 2080 and 2070S are very much on par with each other, these scores suggest that NVidia's claim that the 3080 is double the performance of the 2080 must be limited to specific use cases.

does nvidia making statements that are only tEcHnIcAlLy TrUe surprise anyone

[nick name]

9 minutes ago, DutchGuyTom said:

The Firestrike and TimeSpy scores are not double my scores though (2600x and 2070S). Since the 2080 and 2070S are very much on par with each other, these scores suggest that NVidia's claim that the 3080 is double the performance of the 2080 must be limited to specific use cases.

The performance from the 2070 (and variants) and 2080 (and variants) actually had me looking for cheap 1080 Ti deals for a little bit.  And I was close to pulling the trigger on a 1080 Ti when they were still in production, but trips to the strip club quickly relieved me of the cash for it.  

 

So if a 3080 is gonna get me double performance of my current card then I'm glad I didn't buy a 1080 Ti and instead spent quality time with strippers.

[DildorTheDecent]

Compared to my RTX 2080Ti (2130MHz core, 8000MHz memory), this 3080 is:

• 14.2% faster in Firestrike
• 13.5% faster in Firestrike Extreme
• 16.4% faster in Firestrike Ultra
• 9.9% faster in Timespy

Mature drivers for the RTX 3080 would slightly increase the gap. Knowing what the clock speed of the RTX 3080 was would be beneficial though.

[nick name]

8 minutes ago, DildorTheDecent said:

Compared to my RTX 2080Ti (2130MHz core, 8000MHz memory), this 3080 is:

• 14.2% faster in Firestrike
• 13.5% faster in Firestrike Extreme
• 16.4% faster in Firestrike Ultra
• 9.9% faster in Timespy

Mature drivers for the RTX 3080 would slightly increase the gap. Knowing what the clock speed of the RTX 3080 was would be beneficial though.

Ooooh.  That really makes me want to grab a 3080 at launch just so I can be in the top 10 (briefly) on as many benchmarks as I can run over at the bot.  

[Rohith_Kumar_Sp]

That DLSS on vs off comparison though... 

[Rohith_Kumar_Sp]

1 hour ago, DutchGuyTom said:

The Firestrike and TimeSpy scores are not double my scores though (2600x and 2070S). Since the 2080 and 2070S are very much on par with each other, these scores suggest that NVidia's claim that the 3080 is double the performance of the 2080 must be limited to specific use cases.

the tests are with 2080s across has 43 to 64% increase, so given nvidia compared it with 2080 on stage, it's a fair indication of what they said right? that too with updated drivers which technically hasn't been out yet 

[VeganJoy]

1 minute ago, DutchGuyTom said:

I am considering trying to grab a cheap 2080 ti sometime soon. I am anticipating it to be generally better in performance than the 3070 and occasionally people are selling their 2080 Tis as though they're worse.

i havent seen any under 500, but if i did id grab that shit. even if its worse by a bit its still got 11 gb vram and all that. unless the new cuda core layout turns out to be super awesome the 2080ti shouldnt go much below 500 for a while.

[jones177]

 

My TimeSpy scores are:

 

GTX 980 ti    =  5930(i7 8086k)

GTX 1080 ti  =  9720(i7 8086k)

RTX 2080 ti  =  14548(i7 8086k)

 

GTX 980 ti  to  GTX 1080 ti  gain  =  3790

GTX 1080 ti to RTX 2080 ti  gain  =  4828

 

RTX 2080 ti  =  15919(i9 10900k)*

RTX 3080    =   17428(i9 10900k) article score

 

RTX 2080 ti to RTX 3080  gain = 1509

So a RTX 3080 is not a good upgrade for a RTX 2080 ti.

 

The most disappointing of the 3DMark scores is Port Royal at 11455. my 2080 ti does 10061. An improvement but all it tells me is that it will be a little easier to maintain 60fps at 1440p with RT on without DLSS in Control.    

 

My AC:O score at 4k ultra is 59fps with a i7 8086k average so the 67fps on the 3080 is good.   

 

If the SOTTR average frame rate is 83 at 4k highest settings that is not a big leap from my 77fps average.

 

If these scores are real that means the 3070 will only be as fast as a 2080 ti. Still fantastic for the money.

 

*My i9 10900k is not overclocked yet so it only did a score of 14084. The i9 10900k in the article had a score of 15466.

My RTX 2080 ti is overclocked so I ended up with a much higher score than their 2080 ti in the comparison chart.

 

 

[jones177]

6 hours ago, DutchGuyTom said:

The scores don't say anything about the 3070 so there is no way to know if it will perform the same as a 2080ti. The 3070 has less than 70% of the CUDA cores found in the 3080 so if that is any indication, it won't do as well as the 2080 Ti.

it is only speculation on my part and I put it here so I can get some feedback so thanks.

[SolarNova]

A disappointing (vs Nvidias claims)  but not unexpected result.

 

We knew Nvidias benchmarks will have been cherry picked.

 

What this does suggest is that a possible 3080ti with closer specs to the 3090 will be the real upgrade path to aim for, for 4k 120hz gaming, however it will depend entirely on Nvidias chosen path for pricing. Do they go down the historical route of replacing the current 3080 price point with the 3080ti ..or do they go the newer route started with the 20 series and place in a 3080ti between the 3080 $700 and the 3090 $1500. The later will be very telling of Nvidias attitude towards its consumers.

 

 

The silver lining here however is that this opens up the race to the real possibility that AMD can actually challenge the 3080.

[MadPistol]

Double the score of my 2070 Super in Port Royal. Hard to argue with that.

[nick name]

I wouldn't be surprised if Nvidia secretly built these new cards to excel at crypto mining.  They'd never have to advertise to it as miners would try it and find out themselves.  Then they have cards to please gamers and stockholders.  

[MadPistol]

13 minutes ago, DutchGuyTom said:

So almost twice as good as 2070 Supers in a raytracing benchmark but not in non-RT benchmarks.

 

I like RT but it's not a selling point for me. I am happy to go without so long as games offer the option to turn off RT (I see no reasons why they won't).

Just ran Fire Strike Ultra, and it's right at 2x performance as well.

I keep telling people this, but they don't listen, so I will say it again; the 3080 is a 4K GPU. You will not get linear scaling at lower resolutions. The reason for this is simple. It's much harder to feed that many CUDA cores at lower resolutions than it is at higher resolutions.

[Yoshi Moshi]

13 minutes ago, MadPistol said:

Just ran Fire Strike Ultra, and it's right at 2x performance as well.

I keep telling people this, but they don't listen, so I will say it again; the 3080 is a 4K GPU. You will not get linear scaling at lower resolutions. The reason for this is simple. It's much harder to feed that many CUDA cores at lower resolutions than it is at higher resolutions.

Do you disagree with my logic? I think 3080 at 1440p 144 Hz is the best for ray traced games.

 

Control at 4k with DLSS on with ray tracing was ONLY 100 FPS (not close to 144 Hz). Control is the most graphically demanding game as far as I know. Cyberpunk 2077 is set to overtake that with even more demanding ray tracing features than Control. Wouldn't be surprised if you hit 60-80 FPS in that game at max settings with DLSS on at 4K with ray tracing. Meaning a 144 Hz 2K monitor is still the sweet spot for max setting ray tracing on. 3090 will be another story maybe you can get more than 144 Hz on that card at 2K in cyberpunk 2077, and a 144 Hz 4K panel would be best with that.

[MadPistol]

2 minutes ago, DutchGuyTom said:

Yeah, looking at average performance it is almost 2x the performance as well. Honestly had not considered that 4K would have vastly different performance gains than at 1440p or 1080p (I only have the free version of 3DMark). Personally, that makes it an interesting comparison. If you're like me and content with 1080p, Ampere is not as big of an improvement as NVidia claimed it would be. If you're playing games on a 4K panel, 30-series seems to be a great buy (again, assuming these results are representative and accurate).

First time you've told me so no need to take your frustration out at me. Also consider that, as per Steam's recent analysis, most people are still using 1080p. So if some people are frustrated that the 3080 is not twice the performance of the 2080 at their resolution, then that is understandable. 

Sorry, that post was not aimed at you. There are a lot of people that are claiming that the RTX 3xxx cards are the same crap that Nvidia gave us for the RTX 2xxx series.

The RTX 3xxx series of cards is a different beast all together. The 3080 is GA102, whereas the 2070 Super/2080/Super were all TU104. Also, the memory on the RTX 3080 is significantly faster than the RTX 2080 and other cards, which further leads to its ability to drive higher resolutions.

Based on what I've seen thus far,

RTX 3090 = beyond 4K
RTX 3080 = 4K GPU
RTX 3070 = 1440P GPU
Everything else = 1080P or lower.

I will readily admit if I'm wrong on this, but we've seen 2x performance scaling on most games @ 4K as long as the CPU is not a limiting factor. This isn't a fluke. The 3080 is a 4K beast.

[Zombiep00pZ]

Hey guys, here is another leak: 

 

[nick name]

10 minutes ago, Zombiep00pZ said:

Hey guys, here is another leak: 

 

That's the same leak.  That's the video all the information was pulled from.

[Zombiep00pZ]

17 minutes ago, nick name said:

That's the same leak.  That's the video all the information was pulled from.

Aw hell, at least the vid itself wasnt posted here. Makes it easier for people to watch rather than read.

[nick name]

15 minutes ago, Zombiep00pZ said:

Aw hell, at least the vid itself wasnt posted here. Makes it easier for people to watch rather than read.

Honestly, I didn't watch it yesterday.  I watched it earlier today simply because I ran out of things to watch already.  

[Zombiep00pZ]

1 minute ago, nick name said:

Honestly, I didn't watch it yesterday.  I watched it earlier today simply because I ran out of things to watch already.  

Yeah same, I was waiting for this but completely missed it.

[xg32]

2 minutes ago, Zombiep00pZ said:

Yeah same, I was waiting for this but completely missed it.

all the leaks are pretty consistent so far but this was the most comprehensive one. running outta vram in horizon lol.

[nick name]

Just now, xg32 said:

all the leaks are pretty consistent so far but this was the most comprehensive one. running outta vram in horizon lol.

Whoa I didn't even notice that.  

[xg32]

Just now, nick name said:

Whoa I didn't even notice that.  

i saw the same thing with a 2080 ti, think gtav with mods was another offender, also RE3 maxed out would crash the game, same with bl3 when alot of weapons drop, but 95% of the time 10gb should be enough, some tightroping by nvidia.