Sir Beregond

So taking a look, I'd probably opt for something platinum rated like a Superflower Leadex 1200W.

Yeah I'd shoot for 1200 or even a 1600W, especially if you wanted to run something like a 666W BIOS. In terms of what to get, not sure, haven't done any research on such PSUs.

Wait, and you were using the 666W Galax BIOS? That's not a quality of PSU problem, that's literally shoving way more power through a daisy chained cable designed for like 300W. Definitely lucky you did not burn the house down.

Well glad to hear the 4090 didn't die. Just curious since you said daisy chaining. Did you have separate 8-pin (3 or 4?) coming from the PSU and plugging into the adapter to the 16-pin 12VHPWR? Or were you using one of those daisy chained cables that go from PSU to two 8-pins?

Two packs of these and 4 chrome fittings came in today.

You do get a nice IPC bump and clock bump going to a 5700X. If it's not outside your budget, honestly that'd be my choice over just buying another Zen 2 part. 5800X should be around $200 now, but you should be able to find a 5700X for around $180ish. Sell the 3600 and I think you've done decent for $120.

Ah ok, that does look a bit different than mine, but ultimately the same thing. Yeah that will just add to the potential upper limit that your maximum boost clocks can reach. 5600X can boost to 4.6GHz stock, so a +200 override should allow for theoretically boosting up to 4.8GHz for bursty single threaded stuff. Curve Optimizer comes in and by undervolting gives you additional headroom to potentially boost even higher is really the best way to explain it. Stock PBO with +200, max I get on the boost with 5900X is 4.95GHz. Add in CO on auto and I can get up to 5.05GHz. Optimize CO manually and I've gotten it up to 5.15GHz boosts.

I don't know. I wouldn't call 2GB a "big" difference. But it is a difference.

I could be wrong, but I am pretty sure you still need VRAM with DLSS. While I am sure it does reduce its usage a bit by nature of rendering a lower texture resolution, I have not heard of DLSS necessarily being a big VRAM saver. Maybe a question for @J7SC_Orion. I can tell you I definitely still ran out of VRAM on my 12GB card on Witcher 3 with DLSS. I really had to crank down some settings and push DLSS to balanced to make it work without getting slideshow mode a little while into the playthrough.

Positive, as in a positive Curve Optimizer? Your BIOS maybe looks different then mine. I am not following the ask in terms of Auto + positive.

Yeah keep in kind though the way a console handles direct access to storage and how a PC does it while running Windows is different. You are definitely going to need more VRAM on a Windows machine than you will on a console. I could see a 40-series refresh. Unless TSMC has sorted out 3nm yields yet, then 50-series is a way off.

8GB just never made sense. Hell neither did 10 for the original 3080. The fact is we had 8GB cards in 2015 from AMD and 2016 from Nvidia. Why are we still on 8GB as a standard? Moreover, VRAM used to double ever generation at every segment. Tesla to Fermi doubled. Fermi to Kepler doubled. Kepler to Maxwell doubled. And Maxwell to Pascal doubled. Then it stopped and the only thing that doubled was the 3090 which finally doubled from the 1080 Ti which was artificially cut down to 11GB (same as the 2080 Ti) just to differentiate it from the Titans. 3 gens of practically the same RAM in all mainstream SKUs.

Probably depends on what you want to use it for. If it's for gaming, see if your motherboard has a BIOS that supports dropping in a 5800X3D. Otherwise if you have productivity use cases can look at a 5700X or 5800X too as those will boost higher. Any of these would be a great upgrade to the 3600 if your motherboard supports them. Preliminary research looks like they added Ryzen 5000 support with this BIOS from 2021.

This 100%. Curve Optimizer definitely make idle/light load the most prone to stability issues if too aggressive. Actually what makes it one of the harder things to stability test.

I submitted a claim. Might as well see what I get out of it. Unless you opt-out of the settlement, you are part of it and either get money or not.

Code didn't work for these, but they are half off what they are on PPCs. Picked these up just to have on hand.

Taiphoon Burner is not always accurate and sometimes can't even read what the memory ICs are. While you may be able to figure out like Micron, Samsung, Hynix, if you are struggling to figure out like B-die, A-die, CJR, DJR, rev E, etc (depending on manufacturer), you may look here and this might help you figure it out: MemTestHelper/DDR4 OC Guide.md at oc-guide · integralfx/MemTestHelper GITHUB.COM C# WPF to automate HCI MemTest. Contribute to integralfx/MemTestHelper development by creating an account on GitHub.

Yeah I got those too yesterday.

I mean the 3080 was $699. Why would them suddenly raising the price for it's successor by $500 lead to good sales? There should be no surprise at all that the 4080 isn't selling well. When your price increase is proportional to your gen over gen performance increase that means its literally zero performance per dollar increase gen over gen for the 4080. 4070 is clearly a misbranded 4060 Ti in reality for $599, again a rip off. And the 4070 Ti has all the performance characteristics of a real "4070" as 70-class products usually matched previous gen flagship products and the 4070 Ti does this. Yet again, $799, so grossly overpriced, and again, misbranded - just maybe not as poorly as the "4080 12GB", but still misbranded. The 4090 is the only good selling 40-series product. And my opinions about 90 cards being a rebrand and successor to the 2080 Ti in the last 2 gens with a massive price hike again aside, it's at least a great performer that really is unmatched and remains the only card in the lineup that they didn't actually try to massively raise price on from last gen. Yet it is still a cutdown card and I don't think they have any reason to release a 4090 Ti given AMD's current showing. All I can say to Nvidia having poor sales is: GOOD, hope they learn something, and couldn't happen to a more deserving greedy company that is now more interested in being absolutely predatory with misbranding and pricing. I am an enthusiast of graphics, doesn't mean I have to be ok with how they are trying to price everyone out of it and ultimately give the average joe who doesn't buy the top end halo product for the price of what one not very long ago could build a whole high end/enthusiast grade PC for, every gen ultimately less for more. So Nvidia poor sales? HA!

Or Nvidia will just reallocate wafers to server/data center/AI products and keep prices as is. Retailers on the other hand might drop prices. At this stage, I am seeing too many situations where high supply and low demand is not actually resulting in lower prices. See: housing markets, graphics cards.

+200 just means you don't have as aggressive a curve optimizer. It's all in what you are looking to get out of it. Optimized CO can sometimes overcome that gap. For example I was boosting higher with per core CO and +125 than with a straight +200. That said, still had some instability on low load/idle, so needs some more work. The one downside to CO, really a PITA to tweak. Though should be easier with a 5600X than a 5900X. Can confirm, for non 3d v-cache Zen chips, specifically tuned memory makes a big difference. Throwing in a generic 3200 CL16 kit and then going to a tuned C14 3600 kit OC'd to max your max FCLK really makes a difference. For Zen 3, it's pretty rare to have an FCLK that hits 1900 or above stable, so usually I'd say a 3600 kit is the way to go. Looks like you have that already though.

I found this guy's charts in the first part of the video pretty good for understanding this.

You got that right.

Core Cycler for figuring out each individual core per curve optimizer I have heard. I've yet to try the tool though. I don't know, I've heard of so many having bad effects from OCing baseclock, but it seems hit or miss. Might turn out ok.

So they are regressing in CUDA core count now. And I am sure it will cost more too. What a joke....again.