Qwen3-Next-80B-A3B on 16 GB VRAM + 64 GB RAM

Qwen3-Next-80B-A3B on 16 GB VRAM + 64 GB RAM
80Bweights in total
3Bused for each token
512experts in each MoE layer
10 + 1experts picked per token, plus one shared

A mixture-of-experts layer holds many small networks, the experts. A router reads each token and picks a few of them. The rest do nothing for that token, so they do not need to be in the fastest memory.

Three places to keep weights

GPU VRAM, 16 GB
896 GB/s on an RTX 5070 Ti. The GPU reads from here at full speed.
System RAM, 64 GB
Tens of GB/s. Holds what does not fit in VRAM.
NVMe SSD
A few GB/s. Holds the whole model file, so it always has every expert.

Try it: where do the 10 experts live?

One layer is drawn as 512 squares. Change the size of the model file and see how the squares split across the three tiers. Then send a token.

16 GB~48 GB: Q4 file of this model200 GB
in VRAM in RAM on SSD only picked by the router
This token: 1 from VRAM, 9 from RAM, 0 from SSD
25%of experts in VRAM
75%in RAM
0%on SSD only
28.9rough speed ceiling, tokens/s

Illustration, not a measurement. Assumptions: about 12 GB of VRAM and 56 GB of RAM are free for experts, experts are spread evenly, the router picks them at random, and the 3B active weights at 4-bit (about 1.8 GB) are read once per token at 896 GB/s from VRAM, 40 GB/s from RAM and 4 GB/s from the SSD. Real routing is uneven and real programs differ, so treat the speed as an upper bound for the idea, not a forecast.

What to remember

  • A 4-bit copy of this model is about 45 to 50 GB. That fits in 64 GB of RAM, so the SSD is rarely touched.
  • A model larger than RAM must read experts from the SSD again and again. Then the SSD speed sets the token speed.
  • Programs differ. Many, such as llama.cpp, put fixed groups of weights in VRAM and RAM and let the operating system load the rest from disk. Only some move single experts between tiers while running.

The GPU in this example: RTX 5070 Ti

SpecValueWhy it matters here
Video memory16 GB GDDR7The VRAM tier. Anything larger spills into RAM.
Memory bus256-bitTogether with the memory speed, sets the bandwidth.
Memory speed28 Gbps per pin256 bits x 28 Gbps / 8 = 896 GB/s.
Memory bandwidth896 GB/sThe VRAM speed used in the estimate above.
Host linkPCIe 5.0The path between the GPU and system RAM.

Source for the model figures: Qwen3-Next-80B-A3B model card (80B total, 3B active, 512 experts, 10 active, 1 shared).

A Small Throwback to Folding@home

Before crypto, before AI hype, before GPUs became financial instruments, there was a time when spare CPU cycles quietly went to science.

In the early 2000s, whenever I repaired a PC or laptop (which was 99.5% pro bono), I used to do one extra thing:
install Folding@home.

I’d explain to the owner that when their computer was idle, it would donate a tiny bit of electricity and processing power to medical research. Protein folding. Cancer. Alzheimer’s. Real problems. Real science. Electricity was cheap, machines were slow, and nobody expected anything in return – except the quiet satisfaction of contributing.

It wasn’t flashy. There were no tokens, no dashboards promising ROI, no buzzwords. Just a screensaver showing abstract molecules twisting in silence, doing something meaningful while the machine waited.

Fast forward to today. GPUs cost a fortune, watts are measured like fuel, and almost every spare cycle is expected to justify itself financially. And yet… Folding@home is still there. Smaller, quieter, a bit forgotten – but still running. Still doing the work.

I recently stumbled upon my old stats and noticed something funny: I was ranked #1 in my team by points, despite having fewer completed work units. Turns out the system always rewarded impact, not volume. Faster machines, harder simulations, better use of time. That feels oddly consistent with the whole philosophy back then.

No hype. No shortcuts. Just doing the job well.

This isn’t really about idle computing any more.
It’s about a broader shift in mindset.

Back then, the idea that many people could contribute small, invisible pieces toward a shared goal felt natural. There was no expectation of reward, no personal branding, no monetization layer attached to every action. The value was in the collective outcome, not the individual gain.

Today, even our spare resources – time, attention, compute – are framed almost exclusively through profit. If something doesn’t pay, scale, or signal status, it’s often dismissed as naïve or inefficient. In that environment, projects like Folding@home don’t disappear because they stopped mattering, but because the idea of a common, unpaid effort has quietly fallen out of fashion.

Maybe that’s the real loss.
Not idle cycles, but the assumption that doing something together, for no personal return, was once enough.

“As Tolkien wrote, it is often the small, unseen acts of ordinary people that keep the darkness at bay — and perhaps they still can.”

Ode to the 3D Printer in the Sunlit Workshop

My new Creality K2 Plus has been working overtime lately, and between the scent of PLA and the soft sound of the extruder, I found myself inspired 😛 This little poem is a love letter to that moment — where technology meets sunlight and imagination.

In the hum of gears and quiet clicks,
the printer sings — a hymn of molten dreams.
Filament threads through polished brass,
a river of corn and code intertwined.

Outside, the sun spills gold through glass,
balcony air dances with faint sweet resin.
The scent — half nature, half invention —
a whisper of fields reborn as form.

Here, man and machine conspire in light,
to summon shapes from nothing but heat and will.
Each layer, a heartbeat; each pause, a breath —
proof that creation still hums in mortal hands.

And when the print is done,
and the room falls quiet but alive,
it smells a little of corn,
and a lot like the future.

Happy printing to you all 🙂

When Windows Backup (7) Insists on Including the Wrong Drive: A Deep Dive

I recently spent far too many hours wrestling with an odd behavior in Windows 11:
when creating a System Image Backup with the built-in Backup and Restore (Windows 7) tool, Windows insisted on including my D: drive as part of the “system image.”

The problem?

  • My OS lives on C: (NVMe).

  • EFI and Recovery partitions are correct.

  • D: is just a data/games disk.
    Yet the backup wizard grayed it out as “System” and demanded to include it.

Chasing Ghosts

Like any good sysadmin, I went through the usual suspects:

  • BCD (Boot Configuration Data):
    bcdedit /enum all → no references to D:.

  • VSS (Volume Shadow Copy):
    vssadmin list shadowstorage → nothing allocated on D:.

  • Mount points and recovery:
    mountvol showed C:, EFI, and Recovery only.
    WinRE was on the right recovery partition.

Still, the backup wizard stubbornly listed D:.

The Smoking Gun: CBS Registry Entries

After some deeper digging, I hit the jackpot:
the Component Based Servicing (CBS) registry hive contained hundreds of stale entries like:

HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Component Based Servicing\Packages\...
    InstallLocation    REG_SZ    \\?\D:\wd\compilerTemp\...

At some point, Windows Update had used D: as a scratch/temp directory.
CBS dutifully recorded those paths as part of the “system.”
So when Backup scanned the servicing database, it decided D: must be critical.

The Fix

The key was forcing CBS to rebuild itself:

DISM /Online /Cleanup-Image /StartComponentCleanup
DISM /Online /Cleanup-Image /RestoreHealth
sfc /scannow

Patience paid off: it eventually finished, cleaned up the junk, and a reboot later…

The backup wizard no longer forced D: as “System.”

Running:

DISM /Online /Cleanup-Image /CheckHealth

Great success!

Lessons Learned

  1. Windows Backup (Windows 7) is ancient.
    It keys off CBS, BCD, and a variety of heuristics. If anything in your system state points to a drive, it might get flagged as “System.”
  2. It’s not always just boot data.
    Services installed on another drive, drivers, scheduled tasks, even leftover registry entries can make a disk appear “critical.”
  3. The only truth in Windows is the Registry.
    On Linux, you’d search /etc/.
    On Windows, if you want the real answer to “why does the OS think this drive matters?” →
    grep the damn registry.
reg query HKLM /f "D:\" /s
  1. That’s how I found the culprit.

Should You Bother?

Honestly, Microsoft hasn’t touched this tool in years.
For imaging, you’re usually better off with Veeam Agent which has a free full blown community edition! , https://archive.org/details/Macrium_Reflect_Free_Latest, or some other recent tool. They let you pick exactly which partitions matter.

But if you’re stubborn (like me) or want the satisfaction of making Windows behave:

  • clean up CBS,
  • check the registry,
  • and remember: if a disk shows up as “System,” don’t just trust Disk Management — chase the references until you kill them.

Final Word

This bug wasn’t about bootloaders or UEFI quirks. It was about Windows Update leaving crumbs in CBS.
Once those registry ghosts were gone, Backup behaved.

So, next time you’re fighting Windows Backup…
Search the registry first. Always.

That’s All Folks 😀