• 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.

Falcon Northwest pulled the 5090

E

Epyon

[H]ard|Gawd
Joined
Oct 25, 2001
Messages
1,522
I went to see what a Tiki with the 5090 would be and i saw that they pulled the 5090 from the options. (for all computers) And the build time shot up to 45 days. So I guess we know what the supply of the 5090 will be......low as hell

well, that sucks ass.
 
Epyon said:
I went to see what a Tiki with the 5090 would be and i saw that they pulled the 5090 from the options. (for all computers) And the build time shot up to 45 days. So I guess we know what the supply of the 5090 will be......low as hell

well, that sucks ass.
Click to expand...
They probably sold out as they're going to be extremely popular for high end gamers and prosumers doing rendering, Cuda, and AI work. Early bird gets the worm!

Supposedly Nvidia is allocating most of the stock to local retailers and not allowing them to be sold by the pallet to businesses. We shall see...
 
Epyon said:
I went to see what a Tiki with the 5090 would be and i saw that they pulled the 5090 from the options. (for all computers) And the build time shot up to 45 days. So I guess we know what the supply of the 5090 will be......low as hell

well, that sucks ass.
Click to expand...
Shoot them an email and ask if they have pre sold 5090 builds and by doing so have sold out of their initial shipment of 5090s already. Maybe you get some kind of answer from the source anyway.
 
I would expect Jensen to only use faulty chips for gaming cards as apparently demand on datacenter/AI side of things is absolutely insane.
So in other words expect to buy 5090 from your local scalper for at least $2500 and otherwise see "out of stock" messages in online shops pages.

BTW. 5090 could literally be Ada Lovelace arch on 5nm with zero improvements other than more cores and power limit and it would sell like cup cakes.
As it is with some new features and such... yeah, whole initial stock will sell in seconds.
 
I know Falcon Northwest is pricey af but their PC with RTX 5080 and AMD 9800X3D is over $4500...

Who actually pays that much for a PC? My planned build with RTX 5080 and 9800X3D is about $2500 and I've seen similar pre-builds from other competitors for a little over $3K.
 
Well, They are worth over 3mill as a company or maybe that for a single year when i looked into it. I know they also have contracts with facebook. Not everyone is like us. And you can't buy there cases. I was able to get a hold of a old tiki case that had a facebook tag on it but you can't buy them new.

I really really like the tiki set up. But I have a good paying job so i can have some nice things. I bet this is is going to be another 1 year sold out thing going on because of 32gb of memory.
 
XoR_ said:
I would expect Jensen to only use faulty chips
Click to expand...
Considering the 5090 has 21760 of the 24576 core enable that a sensible expectation, will have to see the rest of the skus that use GB202 (rtx6000, x40, etc...) but 99.9% chance like Lovelace, Ampere, Turing, etc... the top gaming card will be the binned down one (A6000-A40 were getting the perfect die GA102 there way for a long time, then we got a 3090ti for them).

But at those size of die, faulty chips will normally be a giant percentage of them.
 
MistaSparkul said:
What happened to "$2500 5090 I'm out"? :ROFLMAO:
Click to expand...
rtx-50s-series-prices-announced-v0-euuvn9hemhbe1.png

Nvidia made an announcement since that it would not be a $2500 msrp it was the january 6, weeks ago...
 
LukeTbk said:
Could you explain your point of someone being out if the 5090 is announced at 2500, with someone trying to buy one after the announcement is at $2000 ? Not sure to follow.
Click to expand...

1737396782780.png


People thought 5090 will be $2500, turned out to be BS, now people do a 180 and are suddenly in for one. Hence "what happened to 5090 for $2500 I'm out". Need more explanation?
 
MistaSparkul said:
People thought 5090 will be $2500, turned out to be BS, now people do a 180 and are suddenly in for one. Hence "what happened to 5090 for $2500 I'm out". Need more explanation?
Click to expand...
How is 2500 im out, it is not 2500 it is less so I am in, a 180 degree, I do not see any change in view or contradiction going on.

Code: 
if (5090 >= $2500)
    exit
else
   will see
 
I have to run 2x video cards in my system The speed of the rendering is double but i only have 1 system. the $2500 5090 I am out for me is I need to buy 2 of the damn things. and I am 100% out of ANY AIB model because of:

Base price+tax+shipping

My M4 macbook pro is faster then i thought so I am moving over to running that via a hdmi cable and only stopping my main pc when i want to pay a game. If the Founder card would have been $2500 yes i would be out.

I just need to go back to the 2 computer set up i had before the 13900k fucked up my whole world and made the whole damn system worthless. I STILL have that cpu on the mother board in a box. I need to get off my ass and rma that and see if i can get my $700 back
 
Ok called F NW and the dude answered after a SINGLE RING no menu options or anything just straight up fucking picked up the phone. So Nvidia gave them the go ahead to pre sale all the cards they gave them but they ran out so fast they had to de list the 5090 and falcon and nvidia did not anticipate all the cards being sold out so fast or they would have sent more. So there build time is 14 days and now its up to 45 days so they must have sold hundreds of systems with the 5090 already. He did say it would return to the site either at launch or 10 days after to place a real order with a 5090.

Holy shit batman. But at least i know I order one and have it ready when i get back from being offshore.
 
Epyon said:
Ok called F NW and the dude answered after a SINGLE RING no menu options or anything just straight up fucking picked up the phone. So Nvidia gave them the go ahead to pre sale all the cards they gave them but they ran out so fast they had to de list the 5090 and falcon and nvidia did not anticipate all the cards being sold out so fast or they would have sent more. So there build time is 14 days and now its up to 45 days so they must have sold hundreds of systems with the 5090 already. He did say it would return to the site either at launch or 10 days after to place a real order with a 5090.

Holy shit batman. But at least i know I order one and have it ready when i get back from being offshore.
Click to expand...

Offshore? You’re an oilman?
 
Epyon said:
I went to see what a Tiki with the 5090 would be and i saw that they pulled the 5090 from the options. (for all computers) And the build time shot up to 45 days. So I guess we know what the supply of the 5090 will be......low as hell

well, that sucks ass.
Click to expand...
So i could make a quick buck scalping...never done it, feels sleazy
 
Epyon said:
I thought the 5090 dies are the defective ones that can not be used in data centers. What you said the die is so big.
Click to expand...
I honestly don't know much about how Nvidia decides what chip goes where etc...some people here seem to really have the insider view on that. Not sure how they know...

However, a bigger chip has a higher probability of having a defect and these chips are huge.
 
You can have a good idea with core count for a die and their skus for example
AD102 (what the 4090 was made from)

NameChipMemoryShadersTMUsROPsBase ClockBoost ClockMemory Clock
NVIDIA GeForce RTX 4070 Ti SUPER AD102AD102-175-KEF-A116 GB8448264962340 MHz2610 MHz1313 MHz
NVIDIA L20AD10248 GB117763681281440 MHz2520 MHz2250 MHz
NVIDIA RTX 5000 Ada GenerationAD10232 GB128004001761155 MHz2550 MHz2250 MHz
NVIDIA RTX 5880 Ada GenerationAD10248 GB14080440176975 MHz2460 MHz2250 MHz
NVIDIA GeForce RTX 4090 DAD102-250-A124 GB145924561762280 MHz2520 MHz1313 MHz
NVIDIA GeForce RTX 4090AD102-300-A124 GB163845121762235 MHz2520 MHz1313 MHz
NVIDIA L40AD102-895-A148 GB18176568192735 MHz2490 MHz2250 MHz
NVIDIA L40 CNXAD102-???-A124 GB181765681921005 MHz2475 MHz2250 MHz
NVIDIA L40GAD102-???-A124 GB181765681921005 MHz2475 MHz2250 MHz
NVIDIA L40SAD102-???-A148 GB181765681921110 MHz2520 MHz2250 MHz
NVIDIA RTX 6000 Ada GenerationAD10248 GB18176568192915 MHz2505 MHz2500 MHz

That why people say the perfect die went to the L40-RTX 6000 product line and the 4090 got lesser one (when yield get good they could end up artificially selling good die down the product stack), for such a big die they could still make 4070ti super out of those that did not even had half of it working correctly.

5090 is similar to the 4090, just 88-89% or so of the core need to work, we can suspect that again this gen X40-RTX 6000 will have the very high quality one and that will find X20-RTX5000 and maybe one day some 5070ti or 5080 super product for the not good enough to make it as a 5090 one.
 
Falcon NW is charging $300 more AND the build time went UP a day to 46 days. Holy shit people must really want that 5090 in there gaming rig. me too tho lol
 
LukeTbk said:
You can have a good idea with core count for a die and their skus for example
AD102 (what the 4090 was made from)

NameChipMemoryShadersTMUsROPsBase ClockBoost ClockMemory Clock
NVIDIA GeForce RTX 4070 Ti SUPER AD102AD102-175-KEF-A116 GB8448264962340 MHz2610 MHz1313 MHz
NVIDIA L20AD10248 GB117763681281440 MHz2520 MHz2250 MHz
NVIDIA RTX 5000 Ada GenerationAD10232 GB128004001761155 MHz2550 MHz2250 MHz
NVIDIA RTX 5880 Ada GenerationAD10248 GB14080440176975 MHz2460 MHz2250 MHz
NVIDIA GeForce RTX 4090 DAD102-250-A124 GB145924561762280 MHz2520 MHz1313 MHz
NVIDIA GeForce RTX 4090AD102-300-A124 GB163845121762235 MHz2520 MHz1313 MHz
NVIDIA L40AD102-895-A148 GB18176568192735 MHz2490 MHz2250 MHz
NVIDIA L40 CNXAD102-???-A124 GB181765681921005 MHz2475 MHz2250 MHz
NVIDIA L40GAD102-???-A124 GB181765681921005 MHz2475 MHz2250 MHz
NVIDIA L40SAD102-???-A148 GB181765681921110 MHz2520 MHz2250 MHz
NVIDIA RTX 6000 Ada GenerationAD10248 GB18176568192915 MHz2505 MHz2500 MHz

That why people say the perfect die went to the L40-RTX 6000 product line and the 4090 got lesser one (when yield get good they could end up artificially selling good die down the product stack), for such a big die they could still make 4070ti super out of those that did not even had half of it working correctly.

5090 is similar to the 4090, just 88-89% or so of the core need to work, we can suspect that again this gen X40-RTX 6000 will have the very high quality one and that will find X20-RTX5000 and maybe one day some 5070ti or 5080 super product for the not good enough to make it as a 5090 one.
Click to expand...
All that makes sense, but where are you learning what the die names and specs are?
 
1.1.2.3.5... said:
All that makes sense, but where are you learning what the die names and specs are?
Click to expand...
On techpowerup when you look at a gpu spec like a 4090:
https://www.techpowerup.com/gpu-specs/geforce-rtx-4090.c3889

In the graphic processor section, you can see a gpu name with a link (AD102) and the architecture name (Ada Lovelace) with a link, those wiill show you a list of the product launch and specs per die:
https://www.techpowerup.com/gpu-specs/?architecture=Ada+Lovelace&sort=generation
https://www.techpowerup.com/gpu-specs/nvidia-ad102.g1005

You can rank them by enable shaders count, memory clock, etc... the AD102 page will tell you the full core specs.

I think in general when people talk about datacenter-AI-giant demand and volume now a day, they have much more the H100(and the upcoming B100-200) of the world now and those are significantly different die, not sure if they can turn those into 4090-5090, different size, not sure they even have RT cores in them gpu start to be a bit different even if they can share tensor and a lot of their SM rt part aside can be similar, they have H100 CNX, regular PCI version of them to bin down toward.
 
You must log in or register to reply here.
Back
Top