• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

5080 Reviews

Comixbooks

Comixbooks

Fully [H]
Joined
Jun 7, 2008
Messages
26,275
https://videocardz.com/newz/alleged-geforce-rtx-5080-3dmark-leak-15-faster-than-rtx-4080-super


Looking like 15% over the 4080 Super which isn't bad but surprisingly it's not as fast as a 4090 :p
I'm sure the Benchmarks will be all over the place when the reviews hit on in a few days or the 30th same as launch day.



View: https://youtu.be/0L1Uyw22UAw?si=gVHVVmiMKv_5pxFF


View: https://youtu.be/nShh_j4s2YE?si=I9e2LkKfpPIKhKX0


View: https://youtu.be/meekBr-ZB1E?si=lToRFz2txaHBOoYI


View: https://youtu.be/sEu6k-MdZgc?si=OExGjMprSUHU3agC
 
Last edited:
If the above report is to be believed, I think 5070 Ti alleged 20% improvement over 4070 Ti for $50 cheaper in msrp and with additional 4GB vram will be worth it than say 5080 versus 4080 super for the same msrp price and same vram. Just my one cent.

1737913849239.png
 
No FE for the 5070ti, if it delivers, will have to see how many to buy at MSRP there will be out there...
 
Still a better story than 5090. At least priced at 1000$ but the AIBs messed up the pricing. I really think nobody will be buying these for 1400-1600$ when they can probably get a used 4090 for the same price.
 
KickAssCop said:
Still a better story than 5090. At least priced at 1000$ but the AIBs messed up the pricing. I really think nobody will be buying these for 1400-1600$ when they can probably get a used 4090 for the same price.
Click to expand...
AIBs didn't mess up the pricing. This is what they need to charge to survive based on what Nvidia is giving them the GPU for. Nvidia charges less for FE because they ARE the manufacturer so there is no middle man for them.

I find it interesting that AIB get singled out to blame when Nvidia is the one who is charging the AIBs for the GPUs and undercutting their own partners.
 
vjhawk said:
AIBs didn't mess up the pricing. This is what they need to charge to survive based on what Nvidia is giving them the GPU for. Nvidia charges less for FE because they ARE the manufacturer so there is no middle man for them.
Click to expand...
Nvidia should not be doing that though.
MSRP price should be price you can buy cheapest reference designs like Palit for.
In the past Nvidia sold reference coolers to AIBs and you got slew of reference designs and some fancy models.
Now we have Nvidia charging less for FE and no one else has coolers/design like that and instead they have different coolers/design and higher prices.

Nvidia still does it better than 3dfx with their whole "we buy STB and only make cards ourselves" but it is similar situation. Except Nvidia has monopoly so they can do that.
The saving grace for AIBs is that availability of FE models is so poor almost no one can buy them. And if someone buys FE it will most likely be scalper who will boost price to be even higher than e.g. Palit. So it is less of an issue than it could be if Nvidia flooded market with their FE models.

vjhawk said:
I find it interesting that AIB get singled out to blame when Nvidia is the one who is charging the AIBs for the GPUs and undercutting their own partners.
Click to expand...
Nvidia just has it too good having almost monopoly and stupid thinking started. Just like 3dfx - who again did it even worse, failed to develop VSA-100 chip to have real Riva TNT2 competitor and didn't really have cash to buy STB to begin with.

Jensen not being an idiot he would stop doing these "FE" shenanigans the moment he had actually viable competition.
Having poster GPU is ok but they would charge at least as much as cheapest AIBs do.

The whole thing is to make RTX cards appear cheaper than they are anyways.
RTX 5090 is $2200 GPU and not $1999 GPU. Just because you can in theory get FE for $1999 doesn't mean many people will.
 
Comixbooks said:
https://videocardz.com/newz/alleged-geforce-rtx-5080-3dmark-leak-15-faster-than-rtx-4080-super
Looking like 15% over the 4080 Super which isn't bad but surprisingly it's not as fast as a 4090 :p
I'm sure the Benchmarks will be all over the place when the reviews hit on in a few days or the 30th same as launch day.
Click to expand...
It was already known from revised Nvidia charts for 5080.
It looks like at CES and few days afterwards before Nvidia changed charts they showed incorrect performance.
It was fairly obvious this is unrealistic because 5080 over 4080 showed similar boost as 5090 over 4090 despite 5090 having massive advantage in additional cores and memory changed from 384-bit to 512-bit versus almost the same core count and the same memory bit depth between 4080 and 5080.

All in all 4090 should still be fairly faster than 5080. Same is true even for hypothetical revised 24GB 5080 Super if Nvidia decides to release 24GB Super card - which will be a possibility with announced 3GB memory modules vs 2GB modules used now. This is because GB203 chip used by 5080 is already maxed out. There can be a small change in base/boost clocks on Super cards. Alternatively information we got about real core count of GB203 is incorrect and there are few cores which are disabled on 5080.

BTW. Weren't benchmarks supposed to be moved to 29th Jan?
 
vjhawk said:
AIBs didn't mess up the pricing. This is what they need to charge to survive based on what Nvidia is giving them the GPU for. Nvidia charges less for FE because they ARE the manufacturer so there is no middle man for them.

I find it interesting that AIB get singled out to blame when Nvidia is the one who is charging the AIBs for the GPUs and undercutting their own partners.
Click to expand...
I don't agree with this. AIBs used to sell the TUF equivalent cooler for like 30-50$ on cheaper models whereas, charged 200-300$ on higher end models. It was the exact same cooler.
So whereas, there might be areas where Nvidia is charging the way it is I do think this is not the only reason why prices are inflated. There is no justification for a ROG ASTRAL for 800$ more than FE price. The cooler isn't worth 800$ by any stretch of imagination.
 
KickAssCop said:
I don't agree with this. AIBs used to sell the TUF equivalent cooler for like 30-50$ on cheaper models whereas, charged 200-300$ on higher end models. It was the exact same cooler.
So whereas, there might be areas where Nvidia is charging the way it is I do think this is not the only reason why prices are inflated. There is no justification for a ROG ASTRAL for 800$ more than FE price. The cooler isn't worth 800$ by any stretch of imagination.
Click to expand...
People will still buy it because scalpers will make other cards almost as expensive.
It might be that this Astral will be the only thing in-stock in normal shops.
Marketing at Asus knows what they are doing 🥲
 
XoR_ said:
It might be that this Astral will be the only thing in-stock in normal shops.
Marketing at Asus knows what they are doing 🥲
Click to expand...

Or the opposite, and they'll be the most sought after, at least if the trend of the last 3 gens continues: Strix (the flagship moniker before Astral) was the most sought after, and the highest priced among all AIB's, especially the white edition.

Scientists can't explain it.
 
Last edited:
Rumors are the 5090 will have very limited launch supply, but lots of 5080's. I heard Microcenter guy here in the Chicago burbs say they will maybe get 30 to 50 RTX-5090's, but get like 150 to 200 RTX-5080's.

Latest benchmark reviews on the 5080, it's 10% to 15% faster than the 4080, which is a joke as the 4080 was 50% faster than the 3080, and the 3080 also 50% faster than 2080. Man nVidia WTF? Same with the 5090 being only 30% faster than the 4090, where the 4090 was 70% faster than the 3090, and the 3090 was 40% to 50% faster than the 2080 Ti.
 
KickAssCop said:
The cooler isn't worth 800$ by any stretch of imagination
Click to expand...
To go one step further, i'm sure the FE cooler costs a whole hell of a lot more in R&D time and costs than what Asus invested with the Astral.

Me over here wondering why AIB's need to exist at all anymore.... don't mind me.
 
Last edited:
They need to exist because both AMD and Nvidia haven't made a quiet cooler in decades. Anyone who says otherwise probably is not sensitive to sound but if your cooler is doing 35 db+ then dammit, we need better coolers.

Of course I am generalizing and assuming tests are performed at a reasonable distance etc. blah blah.
 
The prices for the 5080 AIBs has started to trickle in here in Norway, and they start at around $1600,- . :eek:
 
KickAssCop said:
They need to exist because both AMD and Nvidia haven't made a quiet cooler in decades. Anyone who says otherwise probably is not sensitive to sound but if your cooler is doing 35 db+ then dammit, we need better coolers.

Of course I am generalizing and assuming tests are performed at a reasonable distance etc. blah blah.
Click to expand...
Maybe. I couldn't even tell you what my 3090FE sounded like, I never heard it. Cranked up the power level to max for two years.
 
Good for you then. I can even hear my TUF 4090 even though it is one of the quietest cards out.
 
Zorachus said:
Rumors are the 5090 will have very limited launch supply, but lots of 5080's. I heard Microcenter guy here in the Chicago burbs say they will maybe get 30 to 50 RTX-5090's, but get like 150 to 200 RTX-5080's.

Latest benchmark reviews on the 5080, it's 10% to 15% faster than the 4080, which is a joke as the 4080 was 50% faster than the 3080, and the 3080 also 50% faster than 2080. Man nVidia WTF? Same with the 5090 being only 30% faster than the 4090, where the 4090 was 70% faster than the 3090, and the 3090 was 40% to 50% faster than the 2080 Ti.
Click to expand...
The 5080 is only 10-15% faster than the 4080 because it was nerfed really hard in terms of cores vs the 90 series. As I've said before it's a 70 series badged as an 80 series.

The 5080 has only 50% the cores of the 90 whereas previous gen 80 series had around 70% the cores of the 90 series. As we can see the performance has logically suffered as a result.

My prediction for months now is that Nvidia will later release a 5080 Ti for $1200 that will give you the actual performance and cores (~70%) that you would have expected from the 80 series.
 
pututu said:
If the above report is to be believed, I think 5070 Ti alleged 20% improvement over 4070 Ti for $50 cheaper in msrp and with additional 4GB vram will be worth it than say 5080 versus 4080 super for the same msrp price and same vram. Just my one cent.

View attachment 706459
Click to expand...
Yeah honestly 5080 doesn't seem that compelling an upgrade anymore. If it was as fast as a 4090, maybe. At this point with the 5070 Ti at 16GB, the 5080 would have maybe done better as an offering with 24GB.

vjhawk said:
My prediction for months now is that Nvidia will later release a 5080 Ti for $1200 that will give you the actual performance and cores (~70%) that you would have expected from the 80 series.
Click to expand...
I doubt it. The 5080 is already using the full GB203 and I highly doubt Nvidia sees any reason to sell GB202 silicon for cheaper than the 5090, especially with no competition at that performance level. 4080 Ti never happened either. I expect a repeat of that. I think at best you'll see a 24GB version of the 5080 whether they call it a "Super" or "Ti" or what have you, but even then, maybe not with how much they are pushing the texture memory compression marketing.
 
Considering the 4080 was called a 4070, maybe we should be calling the 5080 a 60 series card and maybe we will be saying that AMD is not even able to compete with the performance of a 60 series card in 2025...

vjhawk said:
My prediction for months now is that Nvidia will later release a 5080 Ti for $1200 that will give you the actual performance and cores (~70%) that you would have expected from the 80 series.
Click to expand...
Selling 750mm GB202 die in $1200 gaming card, maybe (if they have terrible yield and a big change happen in the market), but a $900 5080 24GB super with a 3-5% performance boost over the 5080 sound almost more likely... and it is not like that would not beat AMD offering

70% of a 5090 cores, that getting really close to the $1600 4090 core count and a 50% upgrade from the regular 5080.

The last time the xx80 shipped on a bigger die than the previous gen halo was the 2080 I think, but it had a 700-800 msrp price tag versus the $700 1080ti, it was not cheaper.
 
I have an unopened ASUS TUF 4080 Super...if I can sell that and get back the full retail price I paid for it then I'll get the 5080...otherwise I'm keeping the 4080 Super
 
polonyc2 said:
I have an unopened ASUS TUF 4080 Super...if I can sell that and get back the full retail price I paid for it then I'll get the 5080...otherwise I'm keeping the 4080 Super
Click to expand...

You ever open your mail? sometimes I'll leave magazines sealed for a few months then read them a few months later. Ever since the digital boom it's hard to shut off my phone and just read a magazine.
 
MavericK said:
I thought it was the 29th, and the launch was the 30th?
Click to expand...

I thought the 5080 embargo initially was announced as lifting on the same day (January 30th) but it looks like you're right that it's now the day before...
 
polonyc2 said:
I thought the 5080 embargo initially was announced as lifting on the same day (January 30th) but it looks like you're right that it's now the day before...
Click to expand...
I seem to recall something similar. I think they figured the 30th was kind of shitty, so they pushed it back to the 29th.
 
it is complicated because msrp vs non-msrp have different embargo date its seem, and the release date moved from 21 to 30 making it hard to track, so both date could be true...

https://thinglabs.io/rtx-5090-review-embargo-ends-jan-24th-rtx-5080-jan-29-30th
CardReview Embargo Lift Date
RTX 5090January 24th
RTX 5080 (MSRP)January 29th
RTX 5080 (Non-MSRP)January 30th

Those were the date announced january 17, maybe it changed since...

Drivers-software stack was not ready for the 24 reviews that seem to have been rushed (maybe to avoid big deception-backlash from people already in physical line when they hit the web), will see this time around if Reflex2, cuda Raytracing, a lot of games with native transformer DLSS support, PyTorch and what not are all ready to go.
 
Last edited:
pututu said:
If the above report is to be believed, I think 5070 Ti alleged 20% improvement over 4070 Ti for $50 cheaper in msrp and with additional 4GB vram will be worth it than say 5080 versus 4080 super for the same msrp price and same vram. Just my one cent.

View attachment 706459
Click to expand...
My impression is 5070Ti provides 20% performance gain for less price, 5080 provides 15% performance gain for the same price, 5090 provides 27% performance gain with 25% higher price. 5090 price sucks, but I have to buy one.
 
LukeTbk said:
Selling 750mm GB202 die in $1200 gaming card, maybe (if they have terrible yield and a big change happen in the market), but a $900 5080 24GB super with a 3-5% performance boost over the 5080 sound almost more likely... and it is not like that would not beat AMD offering

70% of a 5090 cores, that getting really close to the $1600 4090 core count and a 50% upgrade from the regular 5080.

The last time the xx80 shipped on a bigger die than the previous gen halo was the 2080 I think, but it had a 700-800 msrp price tag versus the $700 1080ti, it was not cheaper.
Click to expand...
Not sure where they would get an extra 3-5% more performance. 5080 is already using the full GB203. I don't see them selling GB202 for cheaper either. Maybe there are terrible yields, but I doubt it. Seems to be a pretty mature node at this point.

Yeah Turing using 12nm TSMC was basically a half-stepping of Pascal's TSMC 16nm and there wasn't any transistor density gain on Turing vs Pascal, hence the giant dies. You are absolutely correct...as a result sure wasn't cheaper.
 
Sir Beregond said:
Not sure where they would get an extra 3-5% more performance.
Click to expand...
A little bit more clocks on the memory and die was my thought how they would, with the process just getting better and better (and them having bin the best of the best for a little bit in prevision for a super).

4080 super did a little 2505->2550, 1400->1438, that said that gave them a ~1% performance boost... even 3% could be optimistic here.
 
LukeTbk said:
A little bit more clocks on the memory and die was my thought how they would, with the process just getting better and better (and them having bin the best of the best for a little bit in prevision for a super).

4080 super did a little 2505->2550, 1400->1438, that said that gave them a ~1% performance boost... even 3% could be optimistic here.
Click to expand...
Oh...yeah could be.

The 4080 -> 4080 Super also saw it go from a 5% cut down AD103 to a full AD103. 9728 -> 10240 shaders.
 
Last edited:
Sir Beregond said:
I doubt it. The 5080 is already using the full GB203 and I highly doubt Nvidia sees any reason to sell GB202 silicon for cheaper than the 5090, especially with no competition at that performance level. 4080 Ti never happened either. I expect a repeat of that. I think at best you'll see a 24GB version of the 5080 whether they call it a "Super" or "Ti" or what have you, but even then, maybe not with how much they are pushing the texture memory compression marketing.
Click to expand...
How do we know that the current version of 5080 is the full GB203? That's what I find doubtful. Very unlikely that Nvidia create the current 5080 without allowances for defects which means that the 10,752 cores is unlikely to be the max count.

Or maybe the plan B is to turn the very defective 5090's into the 5080 Ti, who knows. But definitely I feel GB203 has a higher full core count to allow for defects. Since even with mature nodes you run into defects and this is a brand new manufacture node at the moment.
 
vjhawk said:
How do we know that the current version of 5080 is the full GB203? That's what I find doubtful. Very unlikely that Nvidia create the current 5080 without allowances for defects which means that the 10,752 cores is unlikely to be the max count.
Click to expand...
It is a relatively small enough die on a quite mature node, with what should a popular product still sold at an high price getting the lesser GB203 in the 5070Ti and the 5090 laptop seem to be getting the almost perfect one (10,496 out of 10,752).

According to this:
https://www.techpowerup.com/gpu-specs/nvidia-gb203.g1073

They get the full node, 10752, being the max, the 4080 super was also a full node on Lovelace at virtually the same size and the same price. Maybe they had "little" choice, if they wanted to keep the same size to fit in laptop and be faster than a 4080super to go full enabled, but considering the 4080super size-price, maybe something they would have done anyway.
 
Last edited:
vjhawk said:
How do we know that the current version of 5080 is the full GB203? That's what I find doubtful. Very unlikely that Nvidia create the current 5080 without allowances for defects which means that the 10,752 cores is unlikely to be the max count.

Or maybe the plan B is to turn the very defective 5090's into the 5080 Ti, who knows. But definitely I feel GB203 has a higher full core count to allow for defects. Since even with mature nodes you run into defects and this is a brand new manufacture node at the moment.
Click to expand...
Well AD103 was 10240 full die and it was 379mm2. 4080 Super is full die, so yields were fine towards the latter half of the generation.

GB203 is marginally more shaders at 10752 and appears to be around the same size die. And it's on the same node.

As for GB202. The die is so massive at 750mm2 I just don't see any world where that gets cut down further to be sold at a $1200ish price point. Remember the full die is 24576 shaders. Its already cut down to make the 5090 which has 21760 shaders...and that's for $1999.
 
Last edited:
Saw prices on Zephyrus G16 5090 here in Europe. €5700 or almost $6000.
Might go for a 4090 instead.
 
Hi. Here 5080 is much slower than 4090, on 1440P. But with MFG 5080 will be faster?
MM.jpg
 
You must log in or register to reply here.
Back
Top