Nvidia Blackwell RTX 5000 GPUs may debut earlier than expected
- Thread starter emorphy
- Start date
Not for the 5000 series but they are working on it. They're having latency issues between chiplets which was always the biggest issue in a chiplet design. This is a large part of why AMD looks faster on paper but isn't in practice. It took AMD almost 10 years(they were working on it before zen 1) to work out all the kinks of a chiplet design before they took the gaming crown (ryzen 7000 series).
Nvidia will be releasing a chiplet design but that's because node shrinks are costing more per wafer and yields per wafer are dropping. bit I do have it on good athurority that the 5090 is going to be a mono style die
Neatfeatguy
Posts: 1,948 +3,831
My 980Ti (Zotac AMP! Omega), I should have used a support bracket to keep that sucker from sagging, but I didn't know any better. After a couple of years the sag of that GPU physically bent the PCIE slot it was in. I'm honestly surprised that the extreme amount the slot was bent that it didn't break
To prevent any future sag for GPUs I moved away from a traditional computer tower and started using one that the MB sits parallel to the ground. Sure, a GPU bracket could also rectify the issue or even laying the computer tower on it's side, but I find I get superior cooling over any other tower I've had from using the CM HAF XB Evo case. Only real drawback is the maximum length a GPU you can fit in the case and that is 13" (or 33cm), plus if you have a rad mounted on the front of the case that reduces the GPU length even more.
It's nothing recent in terms of GPU size that has required the use of GPU support brackets. That 980Ti was released 8.5 years ago.
I don't think that the 16pin would have been an issue if it was built correctly (clearly Nvidia dropped the ball on that one), but at the same time I really see no reason to go that route just because of one GPU (currently just one) needing 500W+. I guess we see what the future holds in terms of power and connectors on the next gen of GPUs.
I have a 7900XT that's back in its box. I was wholly underwhelmed with it, and happened to see a 4090 FE on the shelf. Although I am much happier with it, the fact that it takes up 3 full slots, has come back to bite me. I decided to build a 7950x3D rig with a TUF GAMING x670E-PLUS motherboard, later deciding I wanted the TB4 addon card. But for some reason unknown to me, the instructions say that this card can ONLY be plugged into the PCIe x4 slot... which makes the card block like 90% of the airflow to the 4090. I will NOT buy another card that takes up a full 3 slots. I would like to actually use this addon that I paid over $200 for (I removed it).
As for the cable melting issue, I don't have that problem. I picked up a proper power supply with its own proper cables. The power cable for the 4090 is rated at 600W, as it should be. SilverStone HELA 1200R Platinum, is what I got.
As for the cable melting issue, I don't have that problem. I picked up a proper power supply with its own proper cables. The power cable for the 4090 is rated at 600W, as it should be. SilverStone HELA 1200R Platinum, is what I got.
godrilla
Posts: 1,853 +1,287
Make sure that air cooled 4090 is secure, them pcbs be cracking. Also even if you have a top rated psu the problem is with the 16 pin connector/cable and not the quality of the psu so much. Some things you can do, 1) obviously make sure cable plugged in, 2) use techpowerup gpuz 16 pin 12 volt monitoring to observe fluctuations and possible bad connection with cable. 3) If possible take a temperature reading of the connector at the point of contact during load.
If possible try to get one of these new cables.
Seasonic Unveils 90-Degree Angled 12V-2×6 Power Connector Cables For Next-Gen GPUs
https://videocardz.com/newz/seasoni...angled-power-invites-rtx-40-owners-to-test-it
No need, my cable is perfect. As for air cooling, I never said that. I’m using a Corsair H150i ELITE. I’m more worried about the darn 4090 cracking. Considering getting a vertical mount for my case (Lian Li O11 Dynamic EVO).
PEnnn
Posts: 1,285 +1,854
And whatever is left of the polar ice caps!
m3tavision
Posts: 1,733 +1,510
This is my take as well^.
Seeing you already paid the RTX tax, so you might as well full send RTX ON and use DLSS & upscale and enjoy your 3080 for a few more years. (If... RTX is is why you bought your card.. so use it.)
Nothing about Blackwell is for gaming, just more of Ada architecture. 5000 series performance will come from node process and newer memory.
My prediction: 5090 few hundred more expensive, 5080 unchanged. Everyting got more expensive with rtx 4000. But as they will use same gen chips for 5000, it should be on the same level as 4000.
godrilla
Posts: 1,853 +1,287
The upcoming 4000 super gpus might forecast the potential pricing at least in those 5000 tiers succesions. If Nvidia prices them too high then the market will simply reject them outside a few sales like we seen with the 4080 and initially 4070ti at launch. I mostly agree on the 5090 predictions, if the 3090 launched at $1499, 4090 launched at $1599, so the 5090 might launch at $1699. Then again the current market conditions are pointing to closer to but below $2k. Anything can change within a year like when the 3090ti launched at $2k and 1 quarter later it fell by half lol. Although I have a theory on that Nvidia wanted the second hand market to plummet at the top to make way for 4000 series. For the 5090 to be successful at let's say $1699 to $1999 it needs to bring on the same delta gains as did the 4090 from 3090/ti imo. Lastly Nvidia will try to space themselves as far as they can from Intel and Next gen consoles to justify the sky is the limit premiums. They can't stagnate performance and justify higher premiums even with software smoke and mirrors tricks like frame generation.
LimyG
Posts: 204 +78
Yes, I heard they are still "working on a solution to stop meltage."Interesting that there is all the specs about performance improvements, but none about power requirements. Gee, I wonder what LeatherMan is up to?
High end models may have built in insta trigger safety solution. A kind of "extinguisher," Not confirmed.
But if they take that route: -
- It won't wait to burn before blasting whole GPU and whole case with gooey surpressent..
It's triggers at any plastic melt or temps over 120c. Which ever happens first it would seem.
Fear not. If it happens we have a one month warranty to mail (out cost) to NV and they will
clean it up.
"It will look like new!"
An NV leaker gushing praise said. [Shipping costs for returning repairs will also be paid by the customer. A deposit when sending in is required.
Same leaker praised NV, and went on to say;
Whatever the final (free) anti melt (patent pending) solution is NV put their customers first.
" NV customers are ensured full piece of mind. They will replace Memory thermal pads, "if they deem that they no longer cool the memory."
He said with genuine enthusiasm.
That's all. But more leaks are bound to surface.
d0x360
Posts: 19 +4
Accept it, upscaling is not going to go away for at another decade if it doesn't become permanent along with user defined frame gen gains. Need 10 fps get just 10 at essentially 0 latency cost or image quality loss assuming native frames rate is always hitting 60 min.
Nvidia doesn't need to worry about rdna4 but they should worry about 5. 4 is where amd finally adds dedicated hardware for RT and upscaling which should allow them to close the gap rapidly on the software side especially if they continue to use an open source model. If it's open Nvidia has no chance competing in the image quality arena vs literally every developer at every company able to enhance the code base.
Rdna5 is where AMD will likely come back strong with what was the original goal with rdna3. Latency penalties are essentially already solved when using chiplets and it will likely also be a stacked die for even higher levels of performance. From there is just a matter of adding more and more cores... in theory there is no limit and AMD has extensive experience with chiplet architectures and stacked dies. Their experience in the CPU market is about to significantly enhance their GPU market. Nvidia would be utterly incompetent if they don't see it coming and used the extra year they have to prepare.
Nothing launched already means Nvidia is either just sitting on their next design or like AMD decided that this would be a year of refreshed cards along the stack but AMD isn't just adding some cores and calling it a day they are coming to play with a better performance to dollar ratio, lower consumer side prices and a flagship card that will likely match if not outperform a 4090 across the board even in RT.
Add in open source software for image enhancement tech and an already solid and once a couple bugs are solved likely superior frame gen technology that's available to all which means higher game adoption rates and nvidia once again should be preparing themselves while also debuting something that can awe gamers... Right now. Waiting for AMD to steal market share will only lead to AMD being able to steal even more with rdna5 which should offer even better value than 4 but at much much higher levels of performance and scalability.
Yeah no kidding. I think between Nvidia price gouging and ridiculous electrical QC issues with the 4000 series; the RTX 3080Ti will be my last nvidia gpu; it's going to be AMD despite the annyoing driver qc issues they have. I have no interest in paying riduculously inflated prices to nvidia while simultaneously having a less than 2 year lifespan probability for that investment. I've got 8800 GTXs that are still kicking after over 17 years despite how ill designed they are in terms of resiliance; the fact 4090s are turning into the gotterdamerung of GPUs inside a year for many is a testament to how arrogant Nvidia has gotten towards its customers. This attitude is a result of essential monopoly and unbridled greed sustained by top tier buyers for whom money is no object.
b3rdm4n
Posts: 191 +134
Lots of...uhh.. funny? talk about power requirements, despite it not mentioned, and sorry I won't subject myself to MLID, sorry not sorry. I'd put money on 5090 @ 450w again though.
I can forgive many other things, but not price, and I bet RTX 50 will carry a hefty tag. Both companies will be competing for my upgrade from a 3080 and targeting 4k120, let the best GPU (at the most sane price... if such a thing exists) win I guess.
I can forgive many other things, but not price, and I bet RTX 50 will carry a hefty tag. Both companies will be competing for my upgrade from a 3080 and targeting 4k120, let the best GPU (at the most sane price... if such a thing exists) win I guess.
Vanderlinde
Posts: 642 +436
Mi300X tells a different story. Computational it blasts nvidia away. However with games which is a different workload, latency kicks in. I'm sure they will find a way to absolutely minimize the effect, but monolithic are going to be frigging expensive for both company's at this point.
2nm is twice the price already and it requires 28 billion investment to setup equipment for that. You pay per wafer and the amount of good dies you can extract from that will pretty much determine the price per chip.
Thing is, and AMD has always bin right about this: the money is not in ultra high end gaming cards. The money is in mid and low end range.
So, since 4000 super is pretty much hear, we won't see 5000 till 2025.
If so, sad. I was hoping for something a bit faster.
On the other hand.
October 12, 2022 was 4090 release date. October 2023 marks one year.
October 12 2024 for 5090? I hope so.
I will be back in October.
If so, sad. I was hoping for something a bit faster.
On the other hand.
October 12, 2022 was 4090 release date. October 2023 marks one year.
October 12 2024 for 5090? I hope so.
I will be back in October.
They will release the 5xxx series by this year's autumn guaranteed. It marks 2 years from the 4x series release.
Similar threads
