May I infer what monitor you would pair with this beastly card sir?
Wendell has been loving his new 240hz 4k monitor shown here:
this one is limited to 4k/120
the one in the video is the aorus one I mention in this video but the review of the aorus one isn’t out yet. it’s full fat dp2.1 also
Couldn’t agree more. I’m reasonably sure the reason why every single AIB 5090 we’ve seen so far, even the $2,800 Astral only has a single 12V 2x6 connector is because NVIDIA still is limiting how much board partners can modify the reference design.
If even cheap AIB cards with a factory OC are reaching 600W, then it makes even less sense for cards that cost $500-800 more with higher clocks and plenty of OC options to still be using a single connector. Some overclocking testing on regular AIB cards was already getting disturbingly close to the combined 675W limit of the board slot + 12V 2x6 connector. ASUS is even putting a thermal pad on the backside of the power connector on some 5090’s.
While most PSUs only have a single 12V 2x6 connector, all 5090’s still ship with a PCIe to 12V 2x6 adapter cable. So anyone buying a 5090 could easily run a dual 12V 2x6 connector card that’s load balancing the power draw, had even a single AIB released one.
1 Like
FEs are on the low side. The limited set of AIC for which data’s currently available are showing 620-640 W. So figure probably 650+ W.
Looks like at least MSI is pulling ~615 W through 12V-2x6 at stock. So like a -16% safety margin with a standard aging allowance.
Assuming the usual ~10% TDP exceedance and tendency not to pull much slot power, 5070 Ti’s the most I’d consider running on a single 12V-2x6 unless measurement shows substantial decreases in connector temperatures from 40 series (which, so far, it hasn’t). Probably I’d rather keep it to 5070.
That’s just the GPU… How about the rest of the system?
Kilowatt PSUs time has come…
From discussion I’ve seen so far looks like 1200 or 1300 W, if not 1500 or 1600, will be a more common pairing. ~850-900 W system power’s on the high side for a 1000.
1 Like
240v, 30amp circuit braker, Hela 2050w or Hela 2500w.
Just give in
Good luck buying one, the staggered launch is generally a good idea to push up the margins and have less silicon taken away from the AI dies.
I’m really not sure how much pent-up demand there will be from people for this card. Some people will buy the best no matter what, but I think there is more than that when it came to the 4090’s success. And I don’t believe a lot of 4090 users will be selling their cards for a 5090 any time soon, especially when there is simply nothing new to play.
1 Like
AFAIK the 5090 die has a few cores disabled vs. the datacenter version. So the better the yields, the fewer 5090s will be available.
Not sure, and hard to tell if it’s sold out any way. The 5090 is a dream for amateur AI workloads since I would guess the dual slot flow through would be pretty good for multi-GPU setups.
I’m actually guessing more 4090 owners will upgrade than one would expect. Most 4090 owners are probably wealthy gamers for whom 2000+ is a fraction of disposable income or AI people for whom just the VRAM bump will be worth it.
2 Likes
While on the surface that may seem true, “knowledgeable” 4090 owners may not be so keen to upgrade.
Example, personally, any one of my 4090s in single mode, seem to have higher raw scores (no DLSS) than the benchmarks from all the review sites.
This could be due to configuration, OS optimizations, system optimizations, raw hardware, BIOS optimizations, or all of the above, but still a reproducibly true fact.
then add in, poor build quality on several 5090 and AIB constructs, PCIE 5 issues, driver issues, chromatic issues, fan issues, etc…
…and the PRICE GAUGING even by AIB manufacturers and non existent nvidia 5090 FE stock…
… a "knowledgeable: "4090 owner, not fooled by DLSS gimmicks… considering buying a 5090, …for now, has definitely lost its shine.
with non stellar raw performance increase (questionable, remember using the term raw, not DLSS), price gauging, driver issues, pcie 5 issues, physics issues on pcie 5, build quality issues, instability… etc etc. Its a hard argument to make on the purchase.
…I’m definitely happy on my dual 4090… almost feeling like the lack of availability did me a favor…
…and then the hype on a Ti model for August release…
… add in deafening silence from nvidia on the launch debacle…
very tough to even consider at this point the 5090, but what do I know?
IMHO
a reasonable, knowledgeable 4090 owner, would have doubts on the upgrade.
some example data points.
I usually test with games that have a built in static benchmark such as Chernobylite, Cyberpunk, Talos Principle I + II, and Gravity Mark. When no DLSS is used, with max settings / Ultra ray tracing, 4k, no Film Noise, no DoF, no vignetting, no screen reduction, and no Shader Cache (Disabled)
Take the Chrnobylite score reported by DSOG on the 5090 ay 4k, no DLSS, raw performance.
NOW, look at the Chenobylite on one of my 4090s, including min max and lowest 1%
The same scenario repeats on Cyberpunk, and Talos Principle II etc. So, I a 4090s owner, have some doubts on the 5090
I hope I presented a logical case
for giggles:
here the same bench when using dual 4090s:
2 Likes