Sir Beregond

Nice! Galaxy? Did you get a HOF?

Merry Christmas / Happy Holidays everyone! Looking forward to seeing your shiny new 4090 you steal commandeer from "D'Jango", @The Pook.

Multiple rails? What PSU are you using?

Bigger issue is AMD was pretty transparent in the RDNA2 announcement and that included comparisons to the competition and for the most part their charts were fairly accurate. Fast forward 2 years and the RDNA3 announcement was quite different. Lot of "up to" performance claims that were one offs, not the average, and so they ended up in a situation that overhyped the 7900XTX and so when it released and average performance was down from what they tried to paint it as, well yeah they blew it. Then all the talks on gen on gen efficiency gains were flat out false. Yeah, they totally botched the launch and the goodwill they built up with gamers following RDNA2. I agree with you on the 7900XTX being a good value card relative to the competition, but it looses a lot to (power efficiency, RT), and frankly I started to see 4080's that sat on shelves at my local Micro Center start selling out after the RDNA3 launch. Also the 7900XT is not good value at all and is definitely overpriced. If we want to talk about how AMD handled expectations, marketing, branding, pricing, performance claims, they did absolutely blow it. Also the cards feel rushed. I get why they did it, but clearly they have issues. It's a botched launch in my mind.

Two? No you plug in three or four (depending on card model) 8-pin PCIe into the adapter which converts that to the single 16-pin 12VHPWR. Overclocking, you can literally push 600W+ with these things. If you don't have 4 dedicated 8-pin PCIe cables available from your PSU, may need to upgrade as I don't think running two daisy chained PCIe cables is a good idea (the ones that are two 8-pins on a single cable). Definitely also make sure it is plugged in all the way and try to avoid hard bends if possible.

I'm guessing we'll see an RDNA3+ first (think like Zen+ that ironed out some early issues with Zen 1), and then yeah makes sense that RDNA4 then should be hopefully a nice Zen 2 moment for RTG. Problem with AMD is its always the next thing that's supposed to be the big winner. Its too bad really, they won a lot of mindshare with RDNA2, but then blew it with RDNA3.

I believe the MSI card does only 3 (lol) 8-pins to 16-pin 12VHPWR while the others are 4. So from an enthusiast perspective probably better with the Gigabyte or Zotac if you are planning to overclock and play with power. @J7SC_Orion has that Gigabyte I believe. As far as warranty, not sure, I've not owned any of these brand GPUs before for an extended period of time. I've heard hit and miss things with each of them over the years, but I couldn't really tell you anything definitively.

Makes sense really. You have to think, both Nvidia and AMD have had decades at this point of developing drivers. I just hope Intel sticks to it.

Yep that trick with the prefetchers and the one to run realtime. Might have to try that performance bias setting.

I think you are right. 64 has Samsung memory and 56 has either Samsung or Hynix.

Not really sure, maybe @pioneerisloud would know. I think most were doing it for increased voltage/power limits and to get the clocks of a V64 on the V56. I'm guessing you won't unlock to V64 shader count as they likely were defective / fused off in the binning process that made the chip a 56 vs a 64.

Not shocking. 6600 XT has been the go-to 1080p card I think with the 3060 and 3050 being terrible value to performance by comparison. As for higher end, I think the 6800 XT, 6900 XT, 3080 still make the most sense. Anyway, is eBay the only factor in these graphs as far as the "now pricing", because I can get a 6950 XT right now for the same price as a 3080. (in the $700s).

Hadn't heard of this one. Thanks for the tip! Results with 4.8GHz / 4.675GHz @ 1.44v:

Thank you very much for the advice/tips @neurotix and @The Pook. I was able to do 4.8GHz/4.65GHz on the 5900X at 1.44v and crossed 24k multi-core score for the first time.

Very nice! I'm about 250 behind then. Were you also trying Dynamic OC mode?

Nice job @A Zayed!

Following the tips from @neurotix, I tried some per CCX all-core overclocks. I think I still have a lot of room to work with, hit max temp at about 75C. Here's my last CB R23 multi-core run with CCX0 at 4.75GHz and CCX1 at 4.6GHz under 1.35v. Looked like you hit 23,700ish, what did you end up getting your OC's up to @neurotix? EDIT: Tried again with 4.8GHz and 4.65GHz. Did a bit better.

I agree. There was going to always be the segment that buys no matter what. Might as well capitalize on it.

That, and have the drivers in a good launch state.

Really feels like they launched these too early.

I still see them in stock...the XT anyway. XTX is OOS though.

Right, you are right in the "multitude of factors", I am just saying $1200 4080's ain't in-line with a damn thing except greed. As for the 7900 XTX, I would imagine there's margin to play with on AMD's part given the chiplet approach. The GCD's are on functionally an old node at this point (TSMC 6nm being a derivative of TSMC 7nm) and the MCD's on newer 5nm are of a mid-range size given the separation of functionality into the other chips. This calculator is fun to play with: Die-Per-Wafer Estimator WWW.SILICON-EDGE.CO.UK When you get the stats for TSMC wafers, plug in the details for AD102/AD103, and then the GCD's and MCD's for AMD's Navi31, it's very interesting to see how much in a better position AMD is in on yields per wafer vs Nvidia, I feel the Moore's Law thing falls apart a bit as a factor in the AMD pricing for RDNA3. It was definitely a factor in the decision to move to multi-chip design, but now that they have, the actual results should be lower manufacturing costs vs Nvidia. Nvidia is definitely more screwed by the fact they have monolithic still, and I totally agree that's where you are running into the issues with Moore's Law. As for AMD, smaller chips, still mostly competing with these monolithic chips on functionally much cheaper to manufacture design, yet still $1k. My contention is that the 7900 XTX is only $1k because the 4080 is $1200. It's enough of a discount to look good (depending on who you ask), but in my mind still overpriced.

I guess, but I don't know. Profits are not in-line with simply inflation. Look at the bottom lines and at least for Nvidia they are making more money than they ever have in their history. With that knowledge I have to assume there is more going on here and wafer costs are not the same now as they were when 5nm/4nm was brand spanking new. Just like any new node, cost is high at first and then it goes down. Companies then negotiate deals for supply, etc. Given Nvidia's bottom line just seems to be growing (I have not looked at AMD), I have to assume the hit to them on manufacturing costs isn't as much as they are trying to make it out to be.

Beautiful car.

Its funny, at work I deal with releases for our apps on both Android and iOS platforms and 3 years ago, Android would take maybe at most 4 hours to approve a new build to the Prod App Store while Apple would take as much was 1-2 days. For at least the last year that flip flopped and now my iOS build get approved same day sometimes within a few hours, and Google takes like 1-2 days. Not sure what happened but thought that was interesting. Seems app review processes have changed at both places.