Question Identical laptops, but one has 3x slower CPU performance?

[jasonrohrer]

We buy out-dated "monster" gaming laptops used, and have good luck with them for running lower-end games, like LOL. We had particularly good luck with an ASUS ROG G74SX that we bought a few years ago, so we just got another identical one (ironically, the market price on this machine has gone up by $100 in that time).

Problem is, the second one is much slower, subjectively (running LOL side-by-side, for example, the difference is painful). Specs are identical. Same GPU (GTX 560M) and same CPU (Intel® Core™ i7 2630QM). Same RAM (8 GB). The faster one has an extra magnetic HD, but that shouldn't matter.

Ran a simple windows Novabench on both, and indeed, the second one is about 3x slower in CPU and RAM operations:

Good machine:
https://novabench.com/view/1672641

Bad machine:
https://novabench.com/view/1672642

Furthermore, Novabench is reporting the bad machine CPU at 789 MHz instead of 2 GHz.

They have identical GPU performance, and identical disk I/O performance, so that's good.

The slow one is running Windows 10, the fast one Windows 7, so perhaps that's the problem. I created a PuppyLinux LiveUSB, so I could boot identical OSs on both. Sure enough, inside Puppy, all 8 virtual cores are pegged at just under 800 MHz on the slow one (they vary on the fast one, depending on load, but many of them up around 2 GHz).

Running sysbench on Puppy, the single-threaded prime test takes 11.8 seconds on the slow one, and only 3.3 seconds on the fast one (3.5x faster).

/pro/cpuinfo is also showing each virtual core pegged at 799.26 MHz on the slow one. Every core has exactly the same value, whereas on the fast one, cpuinfo is showing most of them up above 2 GHz (including as high as 2.8 GHz), and a few slower than that, based on load (though I do see that two of the core on the fast one show 799.26 MHz).

The faster one is also running hotter, 55 deg C on the fast one vs 35 deg C on the slow one. The slow one hardly makes any fan noise, the fast one's fan is blowing a lot. This begs the question of what the fast one is doing to heat up running Puppy Linux (or why all 8 virtual cores are "spun up" all the time---not sure how it's supposed to work). But still, hot is better than 3x slow.

(For what it's worth, I have them both sitting on the BIOS screen right now, and the fan on the fast one is really blowing loud, while the fan on the slow one is barely running).



I checked the BIOS settings side by side, and they are identical. Exact same BIOS version number, and all the info displayed in the BIOS matches.

I even swapped the batteries (the slow one came with a battery that doesn't hold charge, but we're running both on wall power always anyway). I was thinking that there was some kind of power-management thing kicking in, but no change with good vs bad battery.


Opened them up, identical RAM chips inside (2 Samsung PC3 4GB chips, same model numbers). Noticed that the fast one has them in banks 1 and 3, while the slow one has them in banks 0 and 2. Not sure if that would make a difference, but for sanity, I tried moving the RAM to the other slots on the slow one---no change, CPU still slow.

In dmidecode in puppy linux, I'm seeing the same part number for the RAM chips. Both machines are showing 1333 MHz DDR3, Samsung, etc, internally.


I could try swapping RAM chips between them, but I'll hold off on doing that for now.


Has anyone ever experienced anything like this?

What could be keeping the CPU core speed ratcheted down like this?

 

[jasonrohrer]

Also, googling for "799.26 MHz" is returning nothing.

This seems to be some kind of magic number for this CPU (the slow one has all cores pegged there, and the fast one has a core or two pegged there when not much is running).

 

[jasonrohrer]

Running CPU-Z's Stress CPU feature on both, and watching both with CPUID HWMonitor, I'm seeing the core voltage of the slow one never getting above 0.77 V per core, while the fast one goes up to 1.21 under load.

CPU wattage grows, but never gets above 9 W for the slow one. Fast one climbs to 45 W for the CPU.

Slow one's temp never gets above 41. Fast one's temp will climb up to 88.

Fan speed remains at 2200 on the slow one (climbs to 3800 on the fast one).

On the slow one, multiplier remains stuck at 8.0 always (the lowest value). Fast one goes up to 26 under stress.

Also tried swapping power bricks, no change.

 

[jasonrohrer]

Okay, found that this is a widely-known issue with these (and many other) gaming laptops.

ThrottleStop fixes it, by turning off the BD PROCHOT flag. After that, benchmarks match the other (identical) laptop.

Now the question remains: Why has this BD PROCHOT issue never surfaced in the other, identical laptop after 2 years of heavy gaming use?

All of the other people online seem to be complaining that BD PROCHOT is triggered by heavy GPU and CPU activity at the same time (like gaming), but in this case, that flag seemed to be tripped all the time, and re-trips itself after every reboot, even without any gaming or otherwise heavy GPU or CPU activity.

And once I turn it off in ThrottleStop, it stays off, no matter how heavily I push things on the laptop.

But then after rebooting, it's back on again.

GPU and CPU temp are both low, down around 40 C.


So the question:

Is there something wrong with this laptop that is causing it to get throttled? Like some component on the motherboard that is sending BD PROCHOT for a good reason, and I'm disabling it at my peril?

I.e., should I return it?

Or keep it and ThrottleStop it every time I boot it up?

 

[jasonrohrer]

We have two G74SX laptops here.

One has CPU cores that speed up as needed under load.

The other has CPU cores that are stuck at 800 MHz, regardless of load. CPU and GPU are both showing low temp readings.

This behavior is the same regardless of operating system, swapping batteries, swapping power supplies, etc. Even booting both from an identical Linux LiveUSB stick, I can see that the slow one is throttled to 800 MHz all the time.

I've figured out that this is because of the BD PROCHOT signal, and if I turn it off with ThrottleStop, the slow one becomes fast again (and the CPU speeds rise and fall with load, as they should, instead of being stuck at 800 MHz all the time.)

After rebooting, the throttling comes back on right away, and I have to use ThrottleStop to disable it each time. (Yes, I could write a boot script to automate this).

This is not just during gaming. It's all the time, immediately after boot. Even if just running CPU tasks, it is slow, unless I use ThrottleStop.

Question: is this "normal"? If so, has anyone figured out why? And it apparently does not affect all machines of the same model.

Is it safe to just turn throttling off manually at every boot? What if BD PROCHOT is actually needed, for real, at some point?

Or might this be a sign of hardware failure of some kind?

Anyone experienced any ill effects of turning it off over the long haul?