Etern4l

39 minutes ago, Papusan said:

I don’t think you’ll save much power whatsoever PSU. I got the Dark Power 12 Pro (10 years warranty) early this year and happy with it. Even only 1500w it provides more power than the 1600 SuperNova T2. Best time to buy new high end PSU is next summer. Maybe too long to wait. The high end Corsair 1600 is what to get if money ain’t your problem (but what you save on the electricity bill won’t cover the higher price). 

 

See also 

Best Power Supplies of 2022 - Top PSUs for Gaming PCs | Tom's Hardware

Thanks, that's one of the first articles I've read, hence the Corsair, but now reading horror stories everywhere about it. Good points as always, I don't really need titanium efficiency, which opens more options.

Will look into the Dark Power Pro. 1500W should be fine frankly even with a crazy 2x4090 setup if I ever get there.

Why do you say that it provides more power than the SuperNova T2? That is supposed to deliver up to 2000W sustained power in practice. I've read this used to be the PSU of choice of miners, retailing for about $1k from scalpers 🙂 Anyway, the one place that had it around here no longer does (they offered me an open boxed unit "checked by their technician"...). And then there is this EVGA thing - will they even be around in a year or two?

Other than that, the Super Flower (same platform as the T2) seems weird, not sure about support. The Seasonic is fine, but not perfect. It has 12 years warranty, and 2x12VHPWR, but not fully ATX3.0 compatible and apparently only supporting up to 450W. Sounds idiotic TBH.

Will keep reading, probably ideal to wait as you say. In emergency I have 3 PCIe outputs in my current veteran Toughpower 1200W PSU so could even run a 4090 off that for a while, although it's dual rail and they don't recommend mixing rails (I guess it's just a reduced safety issue).

A few other options:

* Super Flower Leadex Titanium 1600W Fully Modular "80 Plus Titanium" Power Supply  - $360, looks similar to the SuperNova T2

* Seasonic Prime TX-1600 1600W 80 Plus Titanium - $479

Not sure people are following the desktop forum, posted a quick question on PSU options, would appreciate your advice.

Hello,

I'm shopping for a new 1600W high efficiency PSU for 24/7 operation under load. The main two criteria are:

* Reliability - must be able to handle constant operation under high load in non ACed space

* Efficiency - Titanium, given the electricity costs

Ideally it would also support 12VHPWR out of the box or a native adapter should be available.

So far I really only shortlisted two:

* EVGA SuperNova 1600W T2 - PSU of choice of miners, apparently super-reliable including dual ball bearing fans - cost about $330, very poor availability. The EVGA situation is a bit of a concern as regards the 10 year warranty...

* Corsair AX1600i - apparently best in class but $500 and some reliability concerns visible in the Comments section on Amazon

Had an EVGA 1600W T2 on order, but this got cancelled for me. Any other ideas, or should I take the plunge with the Corsair.

At the moment I'm tempted to wait for some new 12VHPWR options, since there is no immediate need to upgrade other than my veteran 1200W PSU's efficiency being 77% lol. Thoughts?

On 5/13/2022 at 10:49 PM, jaybee83 said:

now my first question: what does nvidia get out of it? with the green goblin, always follow the money, they never do anything out of the "goodness of their heart" 😋

Very simple - save on dev costs.

On 10/4/2022 at 11:26 PM, Papusan said:

This New Linux Kernel Update Can Damage Your Laptop Display tomshardware.com

 

According to a report by Phoronix, a new Linux kernel - version 5.19.12, is reportedly damaging laptop displays. 

 

This is one of the very few times we've ever seen a software update - not related to overclocking, have the power to damage a computer component physically. This issue effects all Intel-based notebooks where the integrated display is wired directly to the integrated graphics.

 

This means all Nvidia Optimus laptops, and some potential Intel + Radeon combined laptops (though we effectively never see those) are at risk as well - since the iGP is driving the primary display even while the discrete GPU is active.

 

The only exception to this, is if the notebook in question can disable Nvidia Optimus entirely or the internal display is not hooked up to an Intel integrated graphics solution at all. Such laptops are rare, but they do exist. Thankfully, this only applies with Intel graphics solutions, and means all AMD-based notebooks with AMD CPUs or APUs are not affected.

The Fix Is Already Here

Fortunately, the Linux community has already released a new kernel that reverts all the problematic issues found in the graphics driver back to what they were in previous kernel updates.

This new kernel is known as 5.19.13, and has already been released for everyone to use. However, chances are that most users will have to wait for their respected Linux distros to roll out the kernel update before the mainstream users get a chance to use it.

 

I'm sure there is some here that remember the black screen issues a few years ago with windows machines (Alienware and Clevo's). Not a fun experience. Now they are at risk with Linux as well. Not so sure it will be as easy this time to fix destroyed displays. So take your precautions before it's too late.

 

It seems Linux follow another well known OS.....

 

Good news is that 5.19 is cutting edge, common distros are on 5.15. Teething issue 😉

Hello,

Has anyone used this distro? On Alder Lake CPUs it now beats Windows 11 in Phoronix Benchmark by a respectable 8% on average:

https://www.phoronix.com/review/windows-linux-mid22adl/7

Any comments, issues? How did Nvidia driver installation go?

IMHO the only reason to use Win 11 is to better utilise Alder Lake/Raptor Lake thanks to Intel Thread Director, things can get ridiculous on Windows 10 (if they don't - stay on Win 10), although some more or less cumbersome workarounds exist.

Brave and Edge (both 64 bit) started crashing with "out of memory" on very large pages with plenty of system memory available, reverted to Firefox - so far so good.

Elon has a tendency to shoot from the hip with his AI comments. Plus it's fairly cynical of him to issue these "AI warnings", and then invest heavily in advanced AI research.

You need special Dolby Vision streams, you can't just enable this on any content. I'm not aware of YouTube supporting this, and on Netflix you need the top-tier subscription, otherwise it's not available. No special configuration should be required (unless Sony are very weird, which I'm sure they're not).

Brilliant, all we need is an extended family edition - a couple of people pedalling hard in concert should be able to power a high-end rig.

But more seriously, in this day and age it's not such a bad idea to generate power from exercise bikes etc.

Not too much into anime, but Cowboy Bebop and its Netflix adaptation were fun.

Finally caught up with "Manhunt: Unabomber" - surprisingly good (Paul Bettany did a particularly good job), although seems to romanticise Ted Kaczynski a bit too much.

"Dahmer (something something)" is also decent but grim.

"The Sandman" started well, but quickly went downhill, couldn't finish.

Can't go too wrong with "Breaking Bad" and "Better Call Saul".

Edit: The Expanse was awesome. I hope they shoot/CGI the rest soon rather than wait 30 years.

37 minutes ago, cylix said:

But what happens with the stock from Partners that is already build and in storage? 😄 Now they need to print new boxes with 4070 

That would be further cost of doing business with Nvidia. On the other hand were there even any leaks of these 12GB cards being in production? Might have been mostly vapourware at this point.

42 minutes ago, cylix said:

HAHAHAHAHAHAHHAHA

That's hilarious, but I am starting to seriously wonder if Nvidia are planning to release any new video cards under $1k, or is it just not worth the hassle for Jensen&co 🙂

Edit: actually, the answer is obvious - they can't really release the lower tier models until they sell the Ampere stock, otherwise they would have to discount it very hard... probably more of a reason why they cancelled the 12GB 4080, under the pretence of a bad name.

43 minutes ago, Mr. Fox said:

Yes. And, I cancelled my 13900K to wait for the KS release. Either way, I win. I buy the KS and have the better CPU, or I buy a cheaper K later on.

All good if there is no need to get a new CPU soon. Those who do are in a disgusting situation of being forced into the trap. It is what it is, at least there is relative clarity over the opponents' malign strategy vs the even worse situation where the upcoming Ti/KS releases would turn out to be a complete surprise.

Best interest of customers in general would dictate they should be just opening with genuine top one or two models simultaneously, and go from there. I guess the main defence is to avoid purchasing the fake flagships as much as possible to penalise this annoying to borderline exploitative behaviour.

17 minutes ago, Mr. Fox said:

 

In other words they are not doing anyone any favors by delaying release of the Ti. They are not losing any sales or cannibalizing 4090. All they are doing is delaying the profit and making the Ti hold-outs wait. 

Makes sense. They want people who are willing to pay 20%++ more for roughly the same silicon (albeit binned) to hold out, rather than settle for 4090 and leave Nvidia effectively out of pocket. Probably very few would get the 4090, then the 4090 Ti again, and even if they did there would be a 4090 on secondary market, which they don't want.

Also they don't want to release the 4090 Ti straight away (although that would be the best option for the consumer) because that would generate only half of the combined 4090+4090 Ti hype, and they probably need to amass enough binned chips first anyway. Intel is probably executing the same strategy with the KS.

1 hour ago, tps3443 said:

My EVGA 1300 P+ power supply is now screaming with coil whine so bad it sounds like an outdoor commercial bug zapper or something. My last PSU was an EVGA 1300G+, that one popped and smoked after just a few weeks of owning it. I’m really losing faith in EVGA power supplies. This coil whine is louder than my water chiller. 

Sorry to hear. I'm hoping the apparently tried and true 1600 T2 I just ordered will hold up for me. It was a great deal vs Corsair AXi 1600, about which I've read just too many horror stories.

11 minutes ago, Reciever said:

Holy hell, and they are pre-order only.

That's more than can be said about the 4090 itself at the moment 🤣

28 minutes ago, Reciever said:

It wouldn't surprise me that they just want to keep the hype going and with 4090 supply seemingly low they can probably get away with a higher price tag to boot. Just my 2 cents though

Supply shenanigans + hype + RDNA 3 fail = NVIDIA profit. The first two terms are already in place, now they also need AMD to flop. If that happens, and if the device is a Ti, I would expect launch price to be $2000+ for FE; if it's the full fat professional Titan, they will be asking at least $2500+, probably $3000k+.

41 minutes ago, electrosoft said:

 

Nvidia was shook the last time around with Ampere vs RDNA2. They brought the heat this time around and left nothing on the table Ada. I was hoping AMD had mastered their chiplet approach to bring effectively 2x the performance as RDNA2 theoretically and really shake Nvidia up even more at least in pure rasterization.

 

If AMD can only shake things up to let's say the 4080 16GB that leaves the 4090 to roam free and Nvidia to keep back the 4090ti for much longer as they have no reason to release it or scale down their expectations for Titan as previously tried and go full fat full FP on a Titan card with the ~18k core dies and enable SLI since there is no competition at the top for them.

 

 

 

 

4090 Ti rumours keep piling up:

https://www.notebookcheck.net/Nvidia-reportedly-preparing-RTX-4090-Ti-cards-with-up-to-20-increased-performance-over-the-RTX-4090.661441.0.html

Why would NVIDIA rush to release this Ti variant now and kind of cannibalise the 4090? One possibility is that they are afraid 4090 wouldn't be enough to best RDNA 3. The other, of course, is that they are planning to charge a massive premium for the Ti card, so it's purely about money rather than the battle against AMD per se 🙂

51 minutes ago, johnksss said:

I'm running it again now stock. Lets see how much of a difference there is.🙂

Since you are so thorough, I took a look at the free version of Geekbench 5 and in fact it has a CUDA benchmark as well (just need to change OpenCL to CUDA in the dropdown). Would you care to run that as well?

I'm quite sceptical of Geekbench, and indeed the CUDA benchmark is a bit short, but OTOH it should take just a moment of your time. Probably the best thing about it is that that it is cross-platform, so I/we can later boot to Ubuntu or Pop live USB and see if there is any difference.

Just now, johnksss said:

No problem. Sorry it took so long though.

 

Also to add I will need to rerun it as I think that is over clocked.

No worries about the impact of OC on my account, as I was just looking for ballpark numbers. Many thanks again.

18 minutes ago, Papusan said:

Mixing cables between differnt brands PSU's is a bad idea. As for etc the Bequiet's new 12VHPWR PCI-E ADAPTER CABLE, you can clearly see the difference in the socket design on the PSU side of the cable. People need to do their due dilligence. You can't etc put in an 12 pin socket into an 8 pin🙂 Not all cables is vired exactly the same way into the PSU sockets. Neither do they use same socket design. All quality PSU manufacturers will deliver this cable if the PSU  is strong enough for Nvidia's specs.

 

 

 

You can find a lot more info about this topic if you gooogle....... 

PSA: Mixing Modular PSU Cables Can Kill Components

Thank you again bro, very useful. Hopefully EVGA doesn't exit the PSU business before they offer a 12VHPWR cable, since I just ordered a SuperNova 1600 T2...

+ karma and virtual reps 🙂

Edit: I was talking specifically about reusing the 2x8pin PCIe->12VHPWR adapter, but still probably best to contact EVGA support and not assume their 8pin PCIe sockets are industry standard.

8 hours ago, Papusan said:

I have the Dark Power Pro 12 PSU. Nice seeing BeQuiet finallly is out with the new 12VHPWR PCI-E ADAPTER CABLE. I have to order it because the cable adapter from Nvidia and the AIB partners is so damn ugly. This will look a lot nicer in the chassis than the big cluster with cable mess🙂 And they offer 3 years manufacturer warranty for this cable. Not 10 years as for the PSU🙃

 

Wonder if EVGA will come up with an equivalent product, although presumably any cable, including the ones from BeQuiet or Corsair would work with EVGA PSUs.

1 hour ago, cylix said:

There are some rumors that RDNA 3 cannot keep up with 4090, not sure if rasterization or in RT, hope only RT.. And that they will be delayed until December. 😐  I mean ill be happy if the 7900xt will be 80% of 4090 but around 1000 dollar price. Well see, but December launch is rally bad..

 

Well, it's not going to be as easy as vs Intel in the CPU space (and it's not actually even that easy there), since they don't have the process advantage. But of course, Nvidia can take away rational customer choice, but they can't take away our effective free market competition dreams 😉

4 hours ago, johnksss said:

Thanks a lot bro, very interesting. So looking at the single precision FLOPS, theoretically there should be about 10.5x improvement over a single 2060S, however, it's "only" 7.8x faster. Probably the limitation of the WDDM Windows driver model. Almost certainly the fast TCC mode is still locked on this $2k card ($2k is peanuts in terms of Nvidia pro GPU pricing, which is the point), I won't even bother asking someone to check this. Well, more indirect evidence it's time to move to Linux, as even Ubuntu is now 4% faster than Win11 on ADL, and Clear pulls 8% ahead on average (ex any CUDA performance improvements).

https://www.phoronix.com/review/windows-linux-mid22adl/7

+8% and no more Capability Access Manager Service spinning up randomly taking up 25% CPU? I'll take that, installing this weekend.

BTW SHA-1 Hash is only +3x vs 2060S - so Nvidia are still continuing with the anti-mining throttle in the driver, even though mining is dead and buried and someone might want to use their $2k card to compute hashes for other purposes. Must be mining boom nostalgia 😁 I mean AMD please step up, we need a plan B here and fast.

39 minutes ago, electrosoft said:

 

I'm pretty sure the cancelled Titan (at this moment)  is different than the 4090ti. Still hilarious. 🙂

 

This was on Videocardz dated the 12th in regards to a 4090ti so it's in the works:

 

https://videocardz.com/newz/according-to-the-rumor-nvidia-is-already-saving-ad102-gpus-for-geforce-rtx-4090-ti

 

My fav comment: