Ghetto Intel Core i7 from CHINA for $100!!

[Bitter]

That's a nifty board. Would make a hell of a pfsense box, I can't seem to track down an affordable socket G2 board for my i3 370m with dual lan to build one of those again, the local cache at gigabit speed was neat.

 

So it looks like a 4870HQ is available and in stock as well as a 4950HQ, is the 41 vs 42 multiplier or the 100mhz difference in clock speeds (stock for stock) a bonus or a detriment between the two? They seem like they're basically the same processor just the 100mhz and one tick of multi.

https://en.wikichip.org/wiki/intel/core_i7/i7-4870hq

https://en.wikichip.org/wiki/intel/core_i7/i7-4950hq

The 4950 is $134 and the 4870 is $128, seems like it's just splitting hairs but I don't want to split the wrong one.

[X_X]

The refreshes are binned a little higher but whether that's due to them actually being better idk. Might make a difference at stock clocks but maybe not so much when OC'd.

 

I was hoping to do something with all those serial ports on the mini ITX but just didn't happen.

[Bitter]

So I'll go for the one with the higher multiplier I suppose. 4950 purchased.

 

Serial ports are fun, but rarely do you need more than one unless you're controlling multiple POS or a complex CNC lathe.

[X_X]

Remotely operated vehicles use a lot of sensors many of which communicate by serial.

[Bitter]

I could certainly see that board with a mobile Haswell being used for something like medical imaging, running a CT or MRI where you need some CPU horsepower to composite the data into imagines our human brains can comprehend.

[Bitter]

So how poorly will a stock Intel copper core 1150 cooler do on the 4950HQ until I get a little better cooler going on? Just running stock speeds, it'll be OK right? I've seen how tiny the cooler is in my partners MBP with a 4980HQ.

[X_X]

On 3/3/2019 at 10:07 AM, Bitter said:

So how poorly will a stock Intel copper core 1150 cooler do on the 4950HQ until I get a little better cooler going on? Just running stock speeds, it'll be OK right? I've seen how tiny the cooler is in my partners MBP with a 4980HQ.

Might be an issue installing the cooler with the extra height of the CPU. If set to stock TDP of 47W and you can get it to fit then shouldn't be a problem.

[Bitter]

5 hours ago, X_X said:

Might be an issue installing the cooler with the extra height of the CPU. If set to stock TDP of 47W and you can get it to fit then shouldn't be a problem.

Oh yeah the push clips will be pretty tight. I'll get a different low profile cooler, you're probably right the stock one may not snap down. I'll be swapping boards at the same time so it'll be OK. Might run the tall 84W HP one on it since it uses a M3 screw size back plate that's part of the CPU retention mechanism. That's also the same screw size a lot of other coolers use. I'll make it work, thanks for reminding me about the push clips.

[Bitter]

@X_X

It's running my test bench at 4Ghz across all 4 cores, about 80-90W TDP reported by XTU, I bumped the core voltage up a bit to make sure it's stable if it needed more volts but it's only at .6ish currently but hasn't been running long enough to judge stability. Temps aren't stellar but I am punishing it with Prime95 and seeing about 70-76C across the 4 cores with a Hyper 212 on top of it running a better fan that's at about 90% speed. The temp of the air coming out of the heat sink doesn't seem to really jive with the CPU temps I'm seeing though so I may need to see what's going on under that IHS. The seller claims they install liquid metal under there but who knows, also who knows how well it's making contact with the dies. Happy though, even at stock clocks it's plenty of CPU for my gaming/HTPC uses.

[X_X]

Are you having problems at stock voltage? Should be able to undervolt a bit somewhere around 100mV on core, typically.

 

"0.6ish" sounds to me that your actually running 1.2V IOW you have not used any microcode patches that would fix the incorrect VID reading.

[Bitter]

I'm going from what XTU is reporting and in BIOS it's showing it set to .5ish and I added a .1 positive offset, I BSOD'd at 4.0 without the added voltage, I want to see if it's stable and then work back down.

 

What do you suggest I should do?

 

 

Edit: Whoops, it was a +0.150 offset, lowered it to +.100 in XTU. Showing 0.573 in XTU currently under Prime95 abuse with a TDP around 85W (package, always talking package, should I be stating core TDP? That's 81W) and a temp around 70C. I'll check if this is stable and then try dropping it back some more.

 

Edit 2: I forgot I had a small cache voltage offset, +0.030 from the 4590S I had in here with it's OC. I bumped that back down to nothing and dropped to the vcore offset to 0.070V, now at 75-80W package TDP and 65C temp with XTU reporting 0.559V core., but I'm running a more aggressive fan curve than before for testing purposes since it's getting warmer than it has been. Does changing the processor cache ratio also affect the L4 cache and was there any benefit from bumping that past the original 3.6Ghz speeds?

 

Edit 3: Turned t down to zero offset in steps of 0.020, as I neared zero a prime95 thread crashed then I BSOD'd at zero offset. I set it to a +0.080 offset in BIOS and I'll more slowly work down from there tomorrow, going to let it run prime95 work for the next 24ish hours and come back to it.

[X_X]

2 hours ago, Bitter said:

I'm going from what XTU is reporting and in BIOS it's showing it set to .5ish and I added a .1 positive offset, I BSOD'd at 4.0 without the added voltage, I want to see if it's stable and then work back down.

 

What do you suggest I should do?

Get a microcode patch installed. Lack of an update may or may not be causing instability issues as well as the CPU reporting half the actual core voltage. A temporary method could be to use the VMware software to update during Windows booting which will give the correct value in Windows but for BIOS reading microcode needs to be integrated with BIOS (UBU perhaps?) Try the VMWare app first (google vmware microcode). Note also that the processor is unlocked until a microcode patch is applied but together with default clocks and bclk strap support that shouldn't be a problem.

 

L4 and on-board processor cache use separate clocks.

 

Edit: Oh wait a minute, I was thinking you had the 4980 but you changed it to a 4950. The default VID flatlines at 3.6 on the 4950 and 4.0 on the 4980 so you wouldn't get that -100mV offset because your already at 4 bins over default. My bad, carry on with what you are doing with voltage offset.

 

I do have some tablets for Alzheimer's but not sure they are working so well as I keep forgetting to take them lol.

[Bitter]

Where would I find the microcode patch for BIOS?

https://labs.vmware.com/flings/vmware-cpu-microcode-update-driver Is that the vmware thing you're talking about?

If I know it's reporting the voltage at 1/2, is it really that big of a deal? Is it doing damage to the CPU by applying too much voltage?

Would I lose the ability to overclock by upping the turbo ratio with the microcode patch applied? I haven't tried to OC via base clock with this CPU but I've never been able to get any of the other CPU to post past 104 base clock in this Z87 board. I don't know if I'm doing something wrong (I probably am) or if base clock overclocking just isn't a viable option for this particular board.

[X_X]

For BIOS should find the microcode comes with UBU, I will check for you as I've never used it myself but believe that's what most people have success with that don't have the knowledge to do it themselves or ask for someone to BIOS mod.

 

That's the one. Perhaps https://www.youtube.com/watch?v=nEYqw5tvZHs might help, haven't watched the video but the comments seem to contain everything.

 

It's just the reading that's incorrect, as long as we know 0.5V means 1.0V it's not going to hurt but maybe some of the others fixes might be causing stability issues.

 

With microcode applied and default clocks you still get an extra 6 bins you can use so 3.6 can go to 4.2 and IINM all core at 3.4 can go to 4.0.

 

You'll get a few MHz either side of 100MHz, to go higher the bclk strap needs changing. The first one will allow 125MHz bclk plus a few MHz either side while keeping PCIe and DMI clocks around their normal values. DRAM clock however will also increase by 25% so RAM settings need to be set correctly and of course CPU, Cache and iGD ratio's plus voltages adjusted accordingly. 

[X_X]

@Bitter You can checkout UBU here but doesn't seem to have CrystalWell microcode of which ver 0x19 can be found here.

 

I'm current using ver 0x17 due to ease of modding into BIOS mostly, if you would like it to try let me know.

 

FWIW default VID for my 4890 at 40x ratio is 1.2V so -100mV gives 1.1V operation at 4GHz.

[Bitter]

I'll have to figure out what Asus is calling base clock strap on this board, I don't recall seeing anything quite like that. It sounds similar to back in the day with Pentium 4 and Core2 with the PCI/AGP and then PCIe/NB ratios.

 

I think I'll see how it's doing with that little extra vcore at 4.0 across all the cores, I'm not terribly worried about just an incorrect reading so long as I understand what's going on. Sorry for needing the ELI5 version of this, it's above my pay grade and the google results are mostly forum posts which seem to build on the assumption of knowledge that I'm lacking. I'm pretty happy with how it's doing at the current speed and temps are reasonable under an unreasonable load, I need to swap in the cooler I'll be using in the HTPC and see how temps are with that before swapping the board over to the HTPC later on. I don't mind taking a little lower clocks if needed.

[X_X]

What board / BIOS?

[Bitter]

Just now, X_X said:

What board / BIOS?

Asus Z87I Deluxe, I'll have to check the BIOS revision tomorrow I don't recall off the top of my head. I want to say it's the most current version but I'm not confident, I may be thinking of my H87 board.

[X_X]

Should be under Ai Tweaker
 

' CPU Strap' - 'Select a strap close to your target BCLK or leave it at Auto for the BIOS to decide.'
'Auto'
'100MHz'
'125MHz'
'167MHz'
'250MHz'

Sounds like if Auto and you dial in 125MHz for BCLK it will adjust for you. See if it gives you the manual options first. TBH I never liked Asus BIOS decisions from a few years back but haven't tried them since.

[Bitter]

I wasn't my first choice but it was the right price in the used market.

[Bitter]

Weird, that option wasn't there before I installed this CPU...I don't think. I'll swap in a different CPU at some point once I'm satisfied this is stable (I'll do a few weeks of prime95 work and some more 24hr runs of torture test) and see if that option was there as well. Well it works at 4Ghz across all 4 cores without getting too terribly hot so far, I'll keep hacking away at vcore and mess with some other options to see if I can get it running any cooler under unreasonable load. Running Prime95 work it's hanging around 60-65W with OK temps

[Bitter]

@X_X

 

I have a fairly nice Fluke DVOM with some pin leads on the cables, I'm assuming there's some points I could measure voltage on the system board while it's running to check actual voltage to verify it's reporting 1/2. I noticed that the voltage is slightly different in BIOS when I boot up and check it, sometimes it's 0.5xx and sometimes it's 0.6xx. I'd like to let it run and lower/raise itself as needed for obvious cooling and power consumption reasons, but that seems weird that it's different each time. I'm not confident about doing the microcode update just yet, I need to learn some more about it on my own when I have the time (been busy with some issues, spent 12 hours in the ER with my partner) and I need to experiment with that blck stuff to see if I can get it running at the same 4ghz speeds with that before I go forward.

[X_X]

AFAIK the analogue core voltage is taken from an undocumented ball so if it's not wired through then it's not possible to measure with a multi-meter. Usually there would be a test point on the mainboard if the manufacturer included it otherwise for desktop it is usually wired to the Super I/O which converts the analogue reading into a digital one which various softwares can display. For instance CPU-Z usually shows the analogue vcore so if it's connected you should see a reading there. If not connected then it will be nonsense, usually 0V. To display VID instead in CPU-Z the "cpuz.ini" file can be edited to change Sensor=1 to Sensor=0.

 

There was a microcode version 0x1A but seems to have been removed and replaced with version 0x19 in the Intel link provided earlier so I have provided 0x19 ("microcode.dat" file). VMWare includes "install.bat" for installing microcode at Windows startup (reboot may be necessary) and "uninstall.bat" for removing it. Right click and admin privilege required. You could download the files from VMWare to be the same as the attached and add the microcode.dat file.

 

ucode.zip

[Bitter]

51 minutes ago, X_X said:

AFAIK the analogue core voltage is taken from an undocumented ball so if it's not wired through then it's not possible to measure with a multi-meter. Usually there would be a test point on the mainboard if the manufacturer included it otherwise for desktop it is usually wired to the Super I/O which converts the analogue reading into a digital one which various softwares can display. For instance CPU-Z usually shows the analogue vcore so if it's connected you should see a reading there. If not connected then it will be nonsense, usually 0V. To display VID instead in CPU-Z the "cpuz.ini" file can be edited to change Sensor=1 to Sensor=0.

 

There was a microcode version 0x1A but seems to have been removed and replaced with version 0x19 in the Intel link provided earlier so I have provided 0x19 ("microcode.dat" file). VMWare includes "install.bat" for installing microcode at Windows startup (reboot may be necessary) and "uninstall.bat" for removing it. Right click and admin privilege required. You could download the files from VMWare to be the same as the attached and add the microcode.dat file.

 

ucode.zip

Got the file, thank you. I dual boot, but use Ubuntu much less frequently than I do Windows for the HTPC. How's that going to affect things?

[X_X]

For Linux see here. Might find it already is updated. Another alternative could be to make a UEFI driver which would update before any OS is loaded. Ultimately you might want to integrate the microcode update into the BIOS (maybe ask at win-raid.com) but in the meantime you can see how it works for you, even change versions without having to flash. Might also consider whether spectre / meltdown are a concern.