• 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

The easy gains are gone. It is not an act of laziness that there is a ton of emphasis on "alternative techniques" to make things faster, from all vendors.
 
Sadly. However, I have some friends that lost everything when the bubble burst, so I'm not terribly unhappy in that respect.
I had a friend that was heavily leveraged into Luna, but he's a rich lawyer and I don't think he cares anymore. My only position were coins he gave me that had like 50k in value and went to zero.
 
Worst launch since the 2080?
4080s owner, wouldn't upgrade if it only cost me $50. Just not worth the effort

I'd say somewhere between the GTX 4xx and 2xxx series, yeah. Not great (or even all that good), but the new software features are a nice look at future things to come.
If you're an RTX 3xxx or RX 6xxx user tho, RTX 5080/90 looks to be a decent upgrade.
 
I'd say somewhere between the GTX 4xx and 2xxx series, yeah. Not great (or even all that good), but the new software features are a nice look at future things to come.
If you're an RTX 3xxx or RX 6xxx user tho, RTX 5080/90 looks to be a decent upgrade.
It's not a good upgrade from Nvidia's 3xxx or AMD's 6xxx because, in hindsight, very nearly anybody who would have wanted to make such an upgrade already would have done so with a 4080 super. Artificial scarcity and artificial price premiums are not good reasons to buy.
 
Word on the street is you have too much memory bandwidth on the 5080, so going -500 or -1000 on the vram and giving power budget to the core can get you up to ~3300mhz and within 15% of the 4090.
 
It's not a good upgrade from Nvidia's 3xxx or AMD's 6xxx because, in hindsight, very nearly anybody who would have wanted to make such an upgrade already would have done so with a 4080 super. Artificial scarcity and artificial price premiums are not good reasons to buy.
Artificial scarcity? Did people forget that Nvidia was royally reamed by investors for overproducing GPUs and having to buy back a metric-fuckton of unsole stock?

Nvidia won't be producing large quantities of GPU for any release for the foreseeable future. Get used to it.
 
Word on the street is you have too much memory bandwidth on the 5080, so going -500 or -1000 on the vram and giving power budget to the core can get you up to ~3300mhz and within 15% of the 4090.
That's like the earliest days of graphics card bitcoin mining when downclocking the memory freed up additional power for the GPU cores. I was curious if using GDDR 7 memory would be a power hog like the switch from PCIe 4 to PCIe 5 m.2 drives' memory controllers.

Artificial scarcity? Did people forget that Nvidia was royally reamed by investors for overproducing GPUs and having to buy back a metric-fuckton of unsole stock?

Nvidia won't be producing large quantities of GPU for any release for the foreseeable future. Get used to it.

When you purposefully make less of a product than you otherwise readily can so as to further spur demand, that is artificial scarcity.
 
Worst launch since the 2080?
4080s owner, wouldn't upgrade if it only cost me $50. Just not worth the effort
4080 also gets the DLSS improvements so you essentially have the 5080 already -- just a slightly lower power limit and no useless multi frame gen.

I understand now why nvidia decided to stop 4080 production way back in like October. A good deal on a 4080 Super would have been like buying a 8GB version of 1080 Ti back in 2017.
 
Considering how similar (the exact same ? according to TPU density at least), is it not some big decision to stop making Lovelace die once you start pumping the Blackwell one, it is not like they could have made more cards by continuing to make 4080s in any ways...

When you purposefully make less of a product than you otherwise readily can so as to further spur demand, that is artificial scarcity.
Do we known for sure considering the B100-B200-H100 demands that Nvidia is not fully using his TSMC volume it can realistically get ? unstop for many months in a row now ?
 
Word on the street is you have too much memory bandwidth on the 5080, so going -500 or -1000 on the vram and giving power budget to the core can get you up to ~3300mhz and within 15% of the 4090.
Sure and if you sacrifice a goat. Stand in a dark room and spin around 90 times and chant Jensen. Your 5080 will grow horns and the built in ai will generate 3.666 frames for every prime reality frame.

I'm pretty sure if that was true in any way nv would have released it that way rather then release a 5080 that is only a few % faster then AMDs xtx never mind the single digit bump over the 4080s.
 
Sure and if you sacrifice a goat. Stand in a dark room and spin around 90 times and chant Jensen. Your 5080 will grow horns and the built in ai will generate 3.666 frames for every prime reality frame.

I'm pretty sure if that was true in any way nv would have released it that way rather then release a 5080 that is only a few % faster then AMDs xtx never mind the single digit bump over the 4080s.

Why would they bother when everyone is going to slurp up 4x frame gen like its the ambrosia of gods? Nobody plays on native any more.
 
Artificial scarcity? Did people forget that Nvidia was royally reamed by investors for overproducing GPUs and having to buy back a metric-fuckton of unsole stock?

Nvidia won't be producing large quantities of GPU for any release for the foreseeable future. Get used to it.
While true, don't forget that Nvidia had to shift all their production windows with TSMC to accommodate the flaws in Blackwell data center parts.
TSMC production windows are scheduled months if not years in advance, and if you have to stop something you can't just start another with the flick of a switch, and TSMC will not extend your window if it means delaying somebody else's. So who knows what Nvidia has been required to scrap and move around to facilitate the late discovery of the Blackwell flaw?

Nvidia very well could have been forced to shred a whole run of Blackwell, take the production run that was for the gaming components, and re-allocate them to the data center parts, so now what was the second data center run will have to become the main consumer-run and that won't happen until some time in early March most likely.

So that could easily force Nvidia into a paper launch because they took a portion of the run that was putting out the flawed Blackwell components turned it into their GB202 and GB203 components, and then took the run that was intended for GB202 and GB203 and put it back onto the Blackwell parts to get their enterprise parts back on track. But it would result in a 3-4 month switch in their shipping dates with initial low availability of both.
 
Word on the street is you have too much memory bandwidth on the 5080, so going -500 or -1000 on the vram and giving power budget to the core can get you up to ~3300mhz and within 15% of the 4090.
Is that significantly different ?, TPU stock result the 5080 is at 13% of a 4090 at 4k pure raster, 19% with RT on....
 
While true, don't forget that Nvidia had to shift all their production windows with TSMC to accommodate the flaws in Blackwell data center parts.
TSMC production windows are scheduled months if not years in advance, and if you have to stop something you can't just start another with the flick of a switch, and TSMC will not extend your window if it means delaying somebody else's. So who knows what Nvidia has been required to scrap and move around to facilitate the late discovery of the Blackwell flaw?

Nvidia very well could have been forced to shred a whole run of Blackwell, take the production run that was for the gaming components, and re-allocate them to the data center parts, so now what was the second data center run will have to become the main consumer-run and that won't happen until some time in early March most likely.

So that could easily force Nvidia into a paper launch because they took a portion of the run that was putting out the flawed Blackwell components turned it into their GB202 and GB203 components, and then took the run that was intended for GB202 and GB203 and put it back onto the Blackwell parts to get their enterprise parts back on track. But it would result in a 3-4 month switch in their shipping dates with initial low availability of both.
My point simply meant to say that Nvidia isn't artificially limiting the supply to drump up some FOMO.

Anyone who even remotely considered this to have a hint of truth hasn't been paying attention and at this point intentionally remaining ignorant for whatever reason.
 
Considering how similar (the exact same ? according to TPU density at least), is it not some big decision to stop making Lovelace die once you start pumping the Blackwell one, it is not like they could have made more cards by continuing to make 4080s in any ways...


Do we known for sure considering the B100-B200-H100 demands that Nvidia is not fully using his TSMC volume it can realistically get ? unstop for many months in a row now ?

TSMC is openly, brazenly, and happily mercenary and will sell to whomever meets their price, first come, first served unless ordering company is too much of a pain to work with or too openly insults Taiwan... with Intel running afoul of both problem areas. Nvidia almost certainly could have ordered more wafers earlier. However, it also looks good on Wall Street to have a long back order of pent up demand in both Enterprise/faux-AI and gaming. We know that gaming gets the leftovers from the bubble in faux-AI and that the 5000 series is even more explicitly focused on that market even than the 4000 series was. Just like how Zen 5's cores are enterprise focused, with AMD even cheapening out on re-using the same or a nearly identical memory controller from Zen 4.

But, Nvidia knows it can shape demand and since AMD has at least temporarily abandoned the nearly-high to high end gaming market, team Green can now do so quite freely because people have no other options but to buy Nvidia or not buy at all.

My point simply meant to say that Nvidia isn't artificially limiting the supply to drump up some FOMO.

Anyone who even remotely considered this to have a hint of truth hasn't been paying attention and at this point intentionally remaining ignorant for whatever reason.
Nvidia blatantly did that with the RTX 3000 series. Do you not remember people posting picture next to hundreds of retail packaged cards? Of MSI scalping its 3090 stock on Ebay? Part of the problem with the 4000 series launch was that there were so many 3000 cards sitting in warehouses that had been held back for too long the 4000 series was delayed a bit to stave off a market flood and price drop.
 
Nvidia blatantly did that with the RTX 3000 series. Do you not remember people posting picture next to hundreds of retail packaged cards? Of MSI scalping its 3090 stock on Ebay? Part of the problem with the 4000 series launch was that there were so many 3000 cards sitting in warehouses that had been held back for too long the 4000 series was delayed a bit to stave off a market flood and price drop.
Seeing as you believe Nvidia is artificially limiting stock to create fomo, I have an amazing investment opportunity for you in the form of beach front property out in Nevada. Trust me, bro.
 
TSMC is openly, brazenly, and happily mercenary and will sell to whomever meets their price, first come, first served
Not sure what would be the alternative to be happily mercenary has a seller, but I would think they let other outbid the buyer instead of accepting first offer without trying to see what the max that can get, the competition here being entity like Apple that sell their chips per MM even higher than nvidia.

Nvidia is the sole customer for the "TSMC 4N" node and we do not have much indication they are not buying everything TSMC can make on it (they outbid everyone on it, it seem, always), maybe, but is that part 100% speculative ? Same goes for GDDR6X and 7 now or optical networking, it seems like they buy everything the market is able to produce and have been for a while.

back order look good, but actual cash flow going in (deposit are nice, but full payment better), specially if it is a bubble like you say, you open the door to giant order cancellation scenario.

All those timings type manipulation to let 3060 stock happen once an unexpected if will this happen in advance crypto end, sure.

Now that they artificially limit supply by doing less on purpose of selling less of them versus they do not mind selling less of them because they want to allocate a lot of those TSMC 4N toward B100-B200 gpus, that a completely different affair and regardless of what they pick, you will accuse them to create artificial short supply on everything else.

5000 series is even more explicitly focused on that market
If only, not sure if it is anything special:
https://www.phoronix.com/review/nvidia-geforce-rtx5080-linux/5

Seem to be the same very disappointing performance there as well.
 
Feels like how the GTX 580 was against the GTX 480. 15 years later, and we now have the RTX 5080 vs the RTX 4080. If we're on a repeat cycle, then my money's on the 10080 Ti.
 
View attachment 707134View attachment 707135

The raster is even uglier then I expected. Launching this as a 4080 ti would have been disappointing.
This slide really threw me for a loop. As I said the market is nuts right now. Although I will say if you were one of the lucky few that jumped on a 4090 when it dipped down to $1500 you made out.
1738191605404.png
 
  • Like
Reactions: ChadD
like this
AMD has a huge opportunity to garner good value with their next releases.
AMD's GPU division is run by knuckle dragging retards, they have squandered every opportunity handed to them on a silver platter for the last decade. If they had anyone with brains left there they would have continued producing 7900XTX's and significantly undercut the 5080 to retake high tier market share.
 
Lol. Called it.
"When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works."
At least they put the disclaimer at the top of the page....
lol
 
This is essentially a RTX 5070. If priced at $550, it would be a good card though with a high power draw for a mid range card.

The actual "RTX 5070" may only perform as good as an RTX 4070 Ti, or be between the 4070 Super and 4070 Ti. Would be kind of sad if the 4070 Ti Super was faster than the 5070.
 
As usual the frame chasers lunatic has different results from everyone else and claims he's the only one that knows what he's doing.
 
AMD's GPU division is run by knuckle dragging retards, they have squandered every opportunity handed to them on a silver platter for the last decade. If they had anyone with brains left there they would have continued producing 7900XTX's and significantly undercut the 5080 to retake high tier market share.
Right, they have the potential, but they just never have the hardware.
 
AMD hasn't been able to release a decent high end RT card, that's been their Achilles heel...in the age of ray tracing they are stuck with rasterization
 
This is essentially a RTX 5070. If priced at $550, it would be a good card though with a high power draw for a mid range card.

The actual "RTX 5070" may only perform as good as an RTX 4070 Ti, or be between the 4070 Super and 4070 Ti. Would be kind of sad if the 4070 Ti Super was faster than the 5070.
Yep. It trades blows with the 4080, so it's exactly what we'd expect from a 70 series card. The VRAM also isn't enough for newest titles at highest settings. Which is kinda acceptable for a midrange card, but not high end.

But it's named and priced like a 80 series card. Lol.
 
AMD hasn't been able to release a decent high end RT card, that's been their Achilles heel...in the age of ray tracing they are stuck with rasterization
Raster will always be primary. RT feels to me like a fading trend. Being the first noticeable visual upgrade in some years, they grabbed hold of it and ran with a single feature that requires extra hardware to do. It doesn't look like it will keep scaling upward and as I've said before, it just isn't that impressive, not to me at least. Not skyrocketing the price for cards with the performance gains we've been getting each cycle impressive anyway.
 
AMD hasn't been able to release a decent high end RT card, that's been their Achilles heel...in the age of ray tracing they are stuck with rasterization

The problem is even in ray tracing the performance uplifts are equally dismal. It would be one thing of raster performance increased by 15% and ray tracing increased by 30-35%, but the uplift with ray tracing is still only around 15%.

relative-performance-rt-3840-2160.png


rt-cyberpunk-2077-3840-2160.png
rt-alan-wake-2-3840-2160.png
 
is it really "the age or RT" if there is only like 40 games and 8 GPUs that have it?

yes...it's where the games industry is headed...rasterization performance peaked with the 30 series from Nvidia...what's pushing GPU technology forward is now RT...if you don't play RT enabled games then a GPU from 2 generations ago will be fine

I'm sure there are more than 40 games with RT support
 
its not there yet.
sorry, its crept up to about 50, out of thousands and thousands and thousands...

it's not thousands and thousands...you need to look at AAA games released in the past few years (since the 3080)...you can't count indie games...almost every major AAA title has featured some sort of RT support...not to mention the older games like Quake 2, Half Life, Witcher 3 getting it as well
 
WTF is going on here?
At those speed some driver overhead issue could be big impact on FPS, 0.2 millisecond become a lot at 600 something.

Hardware canucks result were quite different, with the more expected 5090 > 4090:
enchmarks-source-hardware-canucks-v0-w8kgvre86yee1.png


Hardware unboxed tested the 5090 with the 566.36 drivers it seem, Canucks with the 571.86
 
Back
Top