TR Forums

I have an old core 2 Quad 9650 computer running Windows 7 64-bit.

The darn thing is stuck at 2 GHz (it is supposed to run at 3 GHz at load).

I noticed this when I was running the latest version of CPU-Z (1.89).

I tried using CPU-Z's bench CPU and stress CPU tests to get the processor to hit 3 GHz but was unable to.

I have my Windows 7 power profile set to high performance.

Can someone help me fix this?

Thank you.

How are the temperatures? My first suspicion is always throttling.

Reset your BIOS to factory.

Use HWiNFO64 to check temps, fan speeds, etc.

Get some compressed air and clean your tower out. And maybe the thermal paste is getting old? If your temps are high, try reinstalling the heatsink with new paste.

Try running a stress test in Linux to see if the issue is OS. If the CPU throttles in Linux, you know the issue is not Windows. Just use a bootable live USB of Ubuntu or Linux Mint.

Did you check your BIOS and make sure that multiplier is set to "9" not "6"? Also make sure that FSB clockspeed is set to 333Mhz. The motherboard's battery might be dying and it reverts to minimal multiplier and FSB clockspeed as factory defaults.

What are the bus speed and multiplier listed in CPU-Z at full load? (should be 333MHz and 9.0)

The multiplier is stuck at 6X even at load. Totally weird.

I ran HWINFO64 and the CPU status shows '-' for Clock and Ratio. Again, totally weird.

CPU temps for the 4 cores are: 44, 38, 42, 41 (degrees C).

When I run CPU-Z stress test, HWINFO64 does not indicate that the CPUs are throttling.

It seems like the multiplier is stuck at 6.0 regardless.

The BIOS indicates that the multiplier is set to 9.0.

Hmm, if the BIOS says the multiplier is 9, but it's stuck to 6 in Windows, then I wonder if the problem is in Windows.

As I mentioned, try Linux.

But also, check if the power settings in Windows are correct. I know you said it's set to "High Performance," but maybe something messed up the meaning of "High Performance."

So go to Power Options --> High performance --> Change plan settings --> Change advanced power settings --> scroll down to Processor power management --> check Minimum processor state and Maximum processor state.

Make sure that Maximum processor state is 100%, and maybe temporarily, set Minimum processor state to 100% too. See if that makes a difference.

Perhaps the Maximum processor state is set to something less than 100%. If so, then changing it to 100% should fix the problem.

Or perhaps Windows is throttling down to the minimum for some reason, in which case setting the Minimum to 100% should sort of fix the problem. (If you have to set the Minimum to 100%, then clearly, something else is wrong, because the processor shouldn't be stuck at its Minimum. But at least this might sort of fix the problem, kind of.)

I checked the processor state in Windows and it is set to 100% for minimum and maximum.

I am currently running with the box cover off and the fan seems to be spinning quite fast.

But there is a lot of dust on the fan and in the box in general.

I don't have handy access to a Linux drive so I can't try Linux.

If it's an ASUS motherboard, it's pretty normal to have to disable in the BIOS " CPU TM function" or the multiplier will drop to 6x under load and stay there until a reboot. Then it will go to 9x, until you put it under heavy load again.

I think it's not absolute temperature it sees and panics about, but a fast rate of change. In any case other brands do not seem to be so touchy-sensitive about it so this only applies to ASUS.

I'd check if your BIOS firmware is the latest/supports that CPU first.

Then try turning off Speedstep/EIST in BIOS to see if it starts with the multiplier locked at 9x. Then turn that back on and try disabling anything related to power saving or 'quiet'.

I don't have an Asus motherboard.

This is an HP machine with a Pegatraon Intel P35 chipset based motherboard.

The BIOS is an AMI production at version 5.21 dated 12/17/2008.

I did not see a Speedstep/EIST option in the BIOS setup.

Wait, did the 9650 ever run at 3ghz in the HP? I know HP does some limitations on their boards and you could have hit one of them if you were attempting an upgrade and the cpu isn't supported.

Ironically, I'm posting this from a q9550 powered gateway sx2803.

I suggest burning a Linux Live ISO or Hiren's Boot CD into a bootable USB drive with Rufus just to eliminate your current Windows install as a possible culprit.

@Samir
I mean @Mr. Camel has probably had the computer for a while so it must have been running at the expected clocks prior to this. Also if the board is what I think it is then yes, it should support Q9xxx.

If it's been installed for a while, definitely boot a linux live cd and let's see if that changes it. I've got a sneaky suspicion that some sort of windows update broke it.

Mr. Camel wrote:

I don't have an Asus motherboard.

Pegatron was spun off from ASUS in 2007 (in 2010 they bought Asrock which had been spun off from ASUS in 2002) which is why most IPIBL motherboards listed for sale are listed as ASUS.

As your board was designed before this, ASUS engineers wrote the original BIOS for it and customized it to HP's specifications, which undoubtedly hid most of the normally adjustable settings from view in order to minimize support calls and RMA due to misconfiguration. Unfortunately in this case it forces you to accept the default setting of "CPU TM function" enabled, so the BIOS performs its crude multiplier throttling before the internal processor throttling (which can drop a Core 2 to as low as 200MHz effective, by inserting stop-clock T-wait-states) can occur. The problem is, unlike the elegant internal CPU throttling mechanism, this keeps things throttled to 6x even after the load causing the overheat condition is removed, just one of the dumb "features" of ASUS' BIOS programming.

Best to clean the heatsink as best you can, and repaste. That Q9650 only draws ~72w at 3.0GHz despite its 95w TDP rating, so even a clean stock heatsink should be able to keep this from happening unless running something like Linpack.

Given the age of the system, it is probably appropriate to ask yourself how much effort you're willing to spend fixing this issue. If the motherboard has solid-state caps in the CPU VRMs it may still have a few good years left; but if it has old-school wet electrolytics I'd say you're on borrowed time already and this system ought to be retired as soon as you can afford to replace it.

I have a q6600 (supposed to be) running at 3 ghz and if I remember correctly to get it to go at that speed, speedstep had to to be turned on.

But this is a gigabyte board not an asus.

Also if you are running windows 10 definitely try the linux boot disk.
Windows 10 will limit alder processors from running at the right speed.
There was a file i had to deleted for my 2600k to run at the right over clocked speed or it was limited to 4.2 ghz max.

I have not checked the core2 quad lately to make sure it is running at the right speed since it is running windows 10.

just brew it! wrote:

Given the age of the system, it is probably appropriate to ask yourself how much effort you're willing to spend fixing this issue. If the motherboard has solid-state caps in the CPU VRMs it may still have a few good years left; but if it has old-school wet electrolytics I'd say you're on borrowed time already and this system ought to be retired as soon as you can afford to replace it.

Unless you're running legacy software that requires Windows 7, keep this in mind. You're talking about a fairly old CPU in an HP box.

For reference, Intel i3-9100F for $111.50 @ Newegg and AMD Ryzen 3 1200 for $64.99 @ Newegg. Those are the cheapest modern offerings from each team that match the quad core and 3.0+ GHz specs of your 9650.

bfg-9000 wrote:

Mr. Camel wrote:

I don't have an Asus motherboard.

Pegatron was spun off from ASUS in 2007 (in 2010 they bought Asrock which had been spun off from ASUS in 2002) which is why most IPIBL motherboards listed for sale are listed as ASUS.

As your board was designed before this, ASUS engineers wrote the original BIOS for it and customized it to HP's specifications, which undoubtedly hid most of the normally adjustable settings from view in order to minimize support calls and RMA due to misconfiguration. Unfortunately in this case it forces you to accept the default setting of "CPU TM function" enabled, so the BIOS performs its crude multiplier throttling before the internal processor throttling (which can drop a Core 2 to as low as 200MHz effective, by inserting stop-clock T-wait-states) can occur. The problem is, unlike the elegant internal CPU throttling mechanism, this keeps things throttled to 6x even after the load causing the overheat condition is removed, just one of the dumb "features" of ASUS' BIOS programming.

Best to clean the heatsink as best you can, and repaste. That Q9650 only draws ~72w at 3.0GHz despite its 95w TDP rating, so even a clean stock heatsink should be able to keep this from happening unless running something like Linpack.

Hmmm...so if this is a bios limitation, was there ever a hacked bios that could fully restore normal operation? I know the guy on the bios-mods forum are pretty good at this, especially when there is a full functional derivative of a crippled board.

superjawes wrote:

just brew it! wrote:

Given the age of the system, it is probably appropriate to ask yourself how much effort you're willing to spend fixing this issue. If the motherboard has solid-state caps in the CPU VRMs it may still have a few good years left; but if it has old-school wet electrolytics I'd say you're on borrowed time already and this system ought to be retired as soon as you can afford to replace it.

Unless you're running legacy software that requires Windows 7, keep this in mind. You're talking about a fairly old CPU in an HP box.

For reference, Intel i3-9100F for $111.50 @ Newegg and AMD Ryzen 3 1200 for $64.99 @ Newegg. Those are the cheapest modern offerings from each team that match the quad core and 3.0+ GHz specs of your 9650.

Those are the prices of just the processors. The Q9650 goes for half those at $35, and the motherboards and ram are cheaper as well. Either of those systems fully built will be $150+ while you can pick up used Q9650 systems for $50 if you look in the right places.

But what do you get for that cheapness? Well, here's the stats--about half the performance for less than half the price:
https://www.cpubenchmark.net/compare/In ... 3029vs3461

I think older systems have a value mark that's hard to beat if you can still use one for what you're trying to do.

I'm might regret asking, but (1) what exact model HP is it, (2) what is different to the shipping configuration (hardware/software), and (3) did it ever work at 3.0 GHz?

Edit: asking because that could well have been a Vista or XP machine...

Regardless, you're not going to find a BIOS update directly from the manufacturer of the board. Go to the HP website and punch in the S/N. I'm sure they have something newer than 2008.

HP have been systematically removing old PC drivers, manuals and BIOSes from their website for many years now (yet everything for 20+ year old printers still seems to be there). You'll have to browse the 3rd-party mirrors, and once you know the exact file name chances are good it's still on HP's FTP servers too--in case you are leery of downloading such a thing from a 3rd party.

Of course you have to extract the BIOS file from the softpaq, then it should be a simple matter for someone on one of those forums to toggle one default setting.

That's if it does run fine at 3GHz with a clean heatsink, at least until you put it under serious load. I have never had to change this setting below 3.6GHz for a 45nm quad, but then I have never used such crappy cooling as stock Core 2. Even the OEM Socket 775 cooling for 115w Prescott or 130w Pentium D was better than that.

As an aside, Win 10 can indeed misreport actual CPU speeds and in some cases even make CPU-Z report the wrong speed, but OP is running Windows 7-64. Note if a Core 2 is installed in a i865 or i875 or VIA PT880 chipset board, then it could even run Win 9x and dual-boot Win 10-64. Try that with your fancy i3.

Is it performing like a 2GHz CPU or a 3GHz CPU? I have an old i7-920 oc'd to 3.8 GHz and neither Task Manager nor HWMonitor (from the CPU-Z author) show a CPU speed bump while running the short CPU-Z benchmark, but the Min/Max CPU speeds DO change in HWMonitor after the test completes. I think the test just sucks up all CPU resources to the point utilities can't keep up / catch up until the test is over.

In any case, clean the dust however you must. USB thumbdrive and Linux LiveCD. Borrowed time, etc.

bfg-9000 wrote:

HP have been systematically removing old PC drivers, manuals and BIOSes from their website for many years now (yet everything for 20+ year old printers still seems to be there). You'll have to browse the 3rd-party mirrors, and once you know the exact file name chances are good it's still on HP's FTP servers too--in case you are leery of downloading such a thing from a 3rd party.

Of course you have to extract the BIOS file from the softpaq, then it should be a simple matter for someone on one of those forums to toggle one default setting.

That's if it does run fine at 3GHz with a clean heatsink, at least until you put it under serious load. I have never had to change this setting below 3.6GHz for a 45nm quad, but then I have never used such crappy cooling as stock Core 2. Even the OEM Socket 775 cooling for 115w Prescott or 130w Pentium D was better than that.

As an aside, Win 10 can indeed misreport actual CPU speeds and in some cases even make CPU-Z report the wrong speed, but OP is running Windows 7-64. Note if a Core 2 is installed in a i865 or i875 or VIA PT880 chipset board, then it could even run Win 9x and dual-boot Win 10-64. Try that with your fancy i3.

I've had good luck finding everything for all of our HP stuff that's at least 10yrs old like the dx2300, dl380 g5, 8760w and more.

Because I could care less about throttling fans, I just remove the pwm pin from the 4-pin headers so the fan stays floored and have used 65w heatsinks all day long on 95w 775 cpus, even under 100% load testing in a 80F environment. ymmv.

Crazy idea of booting win95 and win10-64, but why even stop there? win3.1 anyone?

Topinio wrote:

I'm might regret asking, but (1) what exact model HP is it, (2) what is different to the shipping configuration (hardware/software), and (3) did it ever work at 3.0 GHz?

Edit: asking because that could well have been a Vista or XP machine...

I too would like to know the model number.

As a matter of fact, someone did manage to get Windows For Workgroups 3.11 running on a Core 2 but obviously with unsupported drivers. The video driver was for a VM and he had to build his own sound card, but at least there are official Intel gigabit DOS drivers (I used to use those with a Ghost floppy!).

In contrast, the hardware mentioned in my last post is fully supported by both official Win9x and Windows 10 drivers.

bfg-9000 wrote:

As a matter of fact, someone did manage to get Windows For Workgroups 3.11 running on a Core 2 but obviously with unsupported drivers. The video driver was for a VM and he had to build his own sound card, but at least there are official Intel gigabit DOS drivers (I used to use those with a Ghost floppy!).

In contrast, the hardware mentioned in my last post is fully supported by both official Win9x and Windows 10 drivers.

Yeah, the only real driver you'd need would be a video card and if you can find a PCI card with drivers, that would seal the deal on that. Same with a PCI sound card with DOS drivers. Yep, I've seen those Intel DOS drivers too--kinda made me scratch my head when I saw them as I couldn't imagine the application.

The only problem with Win9x is having too much memory though, right? I haven't set up 95/98 on modern box yet, but in my research it was ram above Limit (256? 512?) that would cause some issues.

I'm even more curious now what the OP's HP machine model is...

Has someone mentioned removing the CMOS battery yet, metering it, and if necessary replacing it?

A failing CMOS battery can sometimes cause quirky behavior, and a CR-2032 is cheap.

After that, I'd ensure the BIOS is defaulted, the system is cleaned thoroughly to get rid of any dust, and I'd remove the heatsink/fan, clean it and the CPU, and re-apply thermal compound.

I'd also inspect the system board very thoroughly for bad or leaking capacitors.

LoneWolf15 wrote:

Has someone mentioned removing the CMOS battery yet, metering it, and if necessary replacing it?
A failing CMOS battery can sometimes cause quirky behavior, and a CR-2032 is cheap.
After that, I'd ensure the BIOS is defaulted, the system is cleaned thoroughly to get rid of any dust, and I'd remove the heatsink/fan, clean it and the CPU, and re-apply thermal compound.
I'd also inspect the system board very thoroughly for bad or leaking capacitors.

Generally a BIOS will throw up an error (Press F1 To Continue, etc) if it's been reset to default --- whether by battery, jumper, an update, or automatically due to error. Generally. Usually. You're correct that it's worth replacing the battery in a board of that age though.

My vote is to replace the board battery, update the BIOS to the latest from HP, and beyond that I'd probably donate it to Goodwill if it technically still functions. Probably not worth putting any more effort into our pondering either, at least until the OP posts back.