I built a 7700X system as an upgrade to my previous i9-9900K system. Performance and efficiency wise it's fantastic, but I've had TONS of problems. Memory unable to run at its rated speed, GPU not initializing at boot, glacially slow boot times even with MCR enabled, janky BIOS releases...

My old-old system is a Haswell Xeon E3 in a Dell and that sucker can boot from power button to the Windows 11 login screen before the monitor even wakes up on the Ryzen.

I really, really hope Intel can release something competitive with AMD in the efficiency space. I feel guilty using more electricity than absolutely necessary, but I also don't want to spend my evenings troubleshooting.

I have read something on Reddit that seems to make MCR actually work nicely on Asus boards (I wouldn't recommend Asus at this point, but I have one). You have to enable MCR in BOTH places in the BIOS. Under DRAM timings under the "A.I. Tweaker" tab (also enable "DRAM Power Down Mode" there, happened automatically for me when enabling MCR) and under... I forgot... something like CPU features, MCU, MCU features, Memory Context Restore. Boot time is pretty fast for me now, I think it's better than on the previous Asus AM4 board, which was also not bad.

New socket and new DDR standard teething, mostly, probably. Granted Intel seems to be doing a much better job with DDR5, so there's no excuse really for AMD.

For best results on AM5 prefer two-DIMM (1DPC) configurations, DDR5-6000 kits with EXPO, and avoid the Samsung B-die stuff (e.g. G.SKILL FlareX 36-36-36). If that's too complicated or whatever I would suggest waiting for Ryzen 9000 and the next wave of AM5 boards where this should all be ironed out, or go with Intel LGA1700 this generation.

Personally I am waiting for 64GB DDR5 UDIMMs and stable four-DIMM (2DPC) configurations before I can even contemplate upgrading from AM4. And since I'm already waiting I also want to see a faster chipset link (B650/X670 are still just PCIe Gen4 x4) and TB5 on my next board.

I unfortunately have the exact kit of ram you mentioned to avoid, it came with the Microcenter bundle.

I live ~2.5 hours from Microcenter, so I took a gamble-- reviews were very mixed but the price was insane. If I had bought this setup from my local BestBuy or online, I'd have returned it.

What's your plan if you can get that 256GB RAM config? Run some full-fat LLM's on your iGPU?

> What's your plan if you can get that 256GB RAM config? Run some full-fat LLM's on your iGPU?

No, mostly running lots of VMs and databases, which tend to chew up a ton of RAM. And I dabble in a few cryptocurrency projects that can be a bit RAM-heavy and/or sometimes benefit from a fast RAM disk.

I actually downgraded, going from 256GB on X99 to 128GB on AM4, so it'll be nice to get back to 256GB with modern features and CPU performance (without having to spend thousands of dollars on Threadripper or Xeon W).

I only came to the comments to see if my instinct (which matches your first sentence) was corroborated. I no longer do anything that particularly requires a cutting edge CPU, I've an M1 Pro MBP and an M2 Mac Studio and they're so far beyond fast enough for me. I've been toying with building a linux box for home but if I do it'll definitely be ARM - fast enough, low power, low heat - perfect.

Apple CPU's are well-implemented, wide and on a cutting edge process. That's why they're fast & efficient. In 2024 the architecture has approximately nothing to with it.

I'm an idiot when it comes to chips, but the one thing I thought I knew was that ARM CPUs run considerably cooler and more efficient in terms of performance per watt, especially compared to Intel. Isn't that why they're in every mobile device?

It's historical. In the 80's when an extra 10,000 transistors for decode could be 20% of the chip, ARM had an efficiency advantage. So it got used in certain markets. Which meant chips were designed for that market. So because there were chips specifically designed for the market, they were used. Feedback loop, essentially.

Your explanation is great except for the beginning of the story: in the 80s and early 90s, CISC cores didn't have a CISC to RISC decoder frontend - they implemented complex instructions directly (though with the help of microcode). The size penalty of that was more like 300%, and it was slower, too(!). RISC was a really big deal at the time. It was even mentioned in the, ahem, documentary "Hackers" from 1995.

A "CISC to RISC decoder" isn't properly a decoder, it's a translator. What they had in the 80's were proper decoders: instruction's in, enable and clock lines out.

I don't want to argue about that. My point is just that the size (transistor count) and performance differences were massive at the time, like 2-3x the performance at 1/4th or so of the size.

Thanks for the explanation!

> Almost nobody in their right mind (except gamers) would buy an Intel CPU

I recently was at a computer shop (looking for keyboard repair) and someone brought in their PC because it kept freezing, ended up they are getting a new PC and even when the salesperson recommended an AMD based PC the guy said that he didn't want to because he didn't have good experiences with AMD. Mind you that PC had an Intel 6000 series CPU in it, so best case that experience likely is 10+ years ago.

I kinda died inside while standing there.

I wanted to buy an AMD laptop for my work, but as soon as I read about AMD CPUs are running hotter than Intel, I decided to stay with Intel. Because I run on Intel since 20 years and I know what I get when I buy one. With AMD I can't be sure and I don't have the time to find out. Just give me a working laptop, I don't care about benchmarks.

Dunno. With quality laptops, the fiddliness of the platform is the laptop manufacturer's problem. My AMD Thinkpad is just as dumb (basically no performance-related options, but the preset is fine) and reliable as my previous Intel Thinkpads. And there's no reason why a CPU would run hotter at the same power consumption and cooling solution. Probably different biases in the thermal sensors or whatever.

AMD desktop chiplet CPUs draw almost 3x power compared to Intel at idle. It makes a huge difference in overall power consumption. Of course they are much more efficient in the top end, but my PC is not running full load most of the time, so I don't see that efficiency.

Of course their monolithic CPUs are much better (just look at the APU in the Steam Deck), but you can't buy them directly. They are either an APU or only available in laptops.

So, I stick to Intel CPU and Nvidia GPU for the power efficiency. I can't even switch to AMD GPUs because they are bad even at top end.

In Windows or Linux? Or have you tested both?

Both, it is an inherent drawback of using chiplets.

{admittedly, my experience (since about 2000) with AMD has been dominantly server not workstation/laptop}

i have found AMD x86-64 CPUs to be overwhelmingly better than intel's poor mimicking of AMD's innovation!

If you want to live in x64-64 Land ... AMD is the winner. Hands down. And has been for a quarter century.

The world is full people who say things like "I used an AMD video card or CPU once in 1967 and it had driver problems, never again."

I'm one of those people. The thing is, I try again every decade or so, and every time without fail there are issues. Intel is the 1990s Apple of the CPU world. "It just works", and I'll pay a fairly high premium for that.

a gamer would wind up with worse power efficiency from an intel CPU because of how much wattage they draw under load, there's a reason i'm on ryzen and not core atm

Meh, I recently built 7700X and 13700K towers. Both with premium ASUS motherboards. The Intel one was faster, the CPU was cheaper when I bought it, it booted insanely fast compared to the AMD system and it accepted the overclocked RAM without fuss or tinkering. On top of that it ran cooler (even though it uses more wattage at load) and had true Thunderbolt (rather than USB4) out of the box. It’s also well known that AMD systems use much more power at idle.