In the Throttlestop overclocking thread at TPU Unclewebb, the developer of TS goes through the X58settings on around page 4. 4.15Ghz seems typical for Prime 95, and 4.3Ghz for softer gaming benchmarks.The board has a locked BIOS, so I believe I'm at the mercy of auto voltage. It would be interesting to see if there's an option buried in Throttlestop or an MSR tweak that can be done. But I don't think I'll running much over 4GHz anyway, the bus fluctuations make me concerned about the possibility of data loss (I use it as a file server).
Cooling isn't an issue at the moment. It ran roughly 55C with 3 fans at 1000RPM during Prime Blend. I do have the cover off the case while I wait for RAM and try to figure out what I want to do for fans for the (re)build. The Deltas I bought are junk. Not sure if they're knock offs or what. I'm debating whether or not to get new fan bearings or Yates or just use the assortment of fans that i have on hand.
Thank you, I'll check that out. I had stumbled upon his data about the W3680 power limit MSRs and MSR values on page 18, which has been invaluable (seems to be required info for overclocking on Linux).In the Throttlestop overclocking thread at TPU Unclewebb, the developer of TS goes through the X58settings on around page 4. 4.15Ghz seems typical for Prime 95, and 4.3Ghz for softer gaming benchmarks.
For RAM speed on locked boards you can try chasing lower latency. PC Wizard lets you read all the latency speeds at various clock speeds the RAM is programmed for.
On my LGA775 Optiplex 380 GSkill Ripjaws CAS7 1333 DDR3 runs at CAS6 on the 1066 bus speed. I've never seen these timings for lower than rated clock speeds published.
There may be some DDDR3 1600 or 1866 modules that will run faster at 1333 speed.
Most locked BIOS boards don't have any VRM cooling. You can glue some heatsinks on yourself for an improvement there. I would suggest separate heatsinks for each MOSFET. Expansion must be accounted for if you try to add 1 big heatsink, and the MOSFETs aren't always level with each other. Either Arctic 2 part thermal cement, or many people just use Superglue. Alphacool, and Enzotech both sell MOSFET heatsink kits. I've had this mod make more Voltage available to the CPU.
There is a TDP Wattage control, and a separate TDC current (Amperage)control. If you look at the equation Amps X Volts=Watts you will see that lower Amps will require higher Voltage for the same Watts output. Of course raisng Voltage by an indirect method like this is risky business since you don't know what Voltage you will actually end up with. What the Auto settings will actually allow is unknown so far.
There have been some reports of using a bootable SATA controller card to move the SATA controller to the PCIE bus so if the BCLK increase crashes the SATA clock it doesn't matter. You can also gain newer SATA speed and USB 3.0 capacity by doing that. This was on some 2 CPU Dell workstations but might be worth a try.Thank you, I'll check that out. I had stumbled upon his data about the W3680 power limit MSRs and MSR values on page 18, which has been invaluable (seems to be required info for overclocking on Linux).
Nice, I'll check out PC Wizard as well. CAS 6 is very low latency indeed, so far I've only been able to achieve the default of CAS 7. By all indications, the 0x501h it's also locked at 1066MHz. There was some mention that 1333 was available on a previous BIOS but downgrading is risky enough business that I'll pass on finding out, it would be a shame to brick the board.
Hopefully I'll have some time to revisit the chips capabilities tonight. I'm going to try the 1333 from my "workstation" in the X58 and see what happens. I'm not sure if it's worth playing around with bus/BCLK/FSB to increase RAM speed or not, or even if it's possible, but I'll probably give that a go too. I think my previous concern over bus speed fluctuations may be unfounded; where the i7 920 would crash with all of +5 on the FSB, it seems these Xeons are much more capable.
Luckily this board seems to have a pretty robust heatsink, and I replaced the thermal pads a couple years back. Added a small fan blowing air at the VRM as well. But I appreciate the info regardless.
I don't own that game but I have been playing games with 5660 for a few years now and never had any issue like the one you describe.Hey guys, just registered to chime in. Finally swapped my 930 for a 5660 and have overclocked to 3.8GHz. She posts and survives Intel Burn Test at 4.2GHz but windows...acts weird. I'm honestly happy with 3.8GHz.
Anyhow, I have a GTX1070 non-OC. I'm trying to play Assassin's Creed Origins and...my framerates are completely unstable and routinely dip below 30.
Other 1070 users report staying above 60. I've followed their guides and tweaked in-game settings along with the nVidia 3D settings panel to mixed results. These exact settings have other users jumping for joy as their FPS spiked up by 20-30fps!
I've read that people in this position are typically CPU constrained, but watching task manager as I play the CPU never peaks above 90% (HT on).
Is what I'm seeing a result of the lack of AVX? I can't come up with another explanation...
Anyone here who's played the game?
Thank you for mentioning that. I noticed PCIE SATA being used to overcome FSB limits in this T7500 thread, but for some reason never thought about using them to upgrade, great idea.There have been some reports of using a bootable SATA controller card to move the SATA controller to the PCIE bus so if the BCLK increase crashes the SATA clock it doesn't matter. You can also gain newer SATA speed and USB 3.0 capacity by doing that. This was on some 2 CPU Dell workstations but might be worth a try.
The meory controller is on the CPU. The i7 only had 1066 speed, the W3680 should have 1333 available. May be you need to reset the CMOS to clear out some old settings?
Well, I wasn't going to even broach it, but maybe a quick look at my settings will see my error(s)? I mean, it benches fine for 12+ hours, but windows has...quirks. Explorer sometimes takes forever to show the contents of a folder--this doesn't happen at 3.8Ghz. I have a gigabyte UD3R rev2. Added an NVMe but ran into corruptions whenever I added an SSD. Found exactly 1 mention of it online. From all my reading I wish I'd gone for the sabertooth as it doesn't have this (random) problem.I don't own that game but I have been playing games with 5660 for a few years now and never had any issue like the one you describe.
I don't think the issue is from only running at 3.8 but there is something not right if you cannot get 4.2GHz tbh. You should really look into that as it may or may not be the issue but worth looking into. I can run 4.6GHz but I do not bother to daily like that as I have no need for it and I have been running 4.2GHz for a few years now.
SetFSB PLL support ended just about when X58 came out. Hit or miss for which PLL are supported. The website has a virus warning. I just boot a Linux Live DVD with no drives connected and go there anyway.Thank you for mentioning that. I noticed PCIE SATA being used to overcome FSB limits in this T7500 thread, but for some reason never thought about using them to upgrade, great idea.
I connected a junk HDD to get some SATA activity (Windows is installed on a flash drive) and tried SetFSB. Windows crashed at ~140 FSB according to HWinfo (I took this guys word for the PLL and used ICS9LPRS113AKLF, too lazy to disconnect everything again and look for it), with and without the HDD IIRC. I'll have to try it again at some point. 4150MHz 190w crashed prime blend with 80% RAM; It never downclocked but after an hour 1 thread randomly decided it wouldn't go past 20% utilization and crashed. Not sure if 200-210w will help or just cook the VRM from going so far out of spec. I've got a fans blowing on the VRM from the top and bottom of the board, but still.
I pulled the original DIMMs and installed 3x2GB 1333MHz CL8 DIMMs from my "workstation" and 3x8GB 1600MHz CL11 ECC UDIMMs. PCWizard and HWinfo indicate they can do 1333MHz CL9 but it runs them at 1066MHz CL7. I believe latency would be worse with 1333MHz C9 anyway. All my DDR3 is M die 2Rx8 and can run at 1066 CL7 with nearly identical subtimings, so I'm not sure why the board won't run the ECC UDIMMs solo. I had cleared CMOS when I tried running the 8GB DIMMs solo, I'll try it again now that I"ve swapped DIMMs, another great idea. +rep for all the help and suggestions, I appreciate it.
Trying to overclock in Linux is proving to be time consuming; after several weeks, many hours, and lots of swearing, cpupower refuses to work. For most people with this problem, it looks like the board/BIOS somehow won't allow the OS to set a CPU governor. It would seem at this point I'm best off asking unclewebb what MSR Throttlestop uses to raise the multi or maybe use RWEverything (never tried it before), then try to replicate it with msr-tools in Linux. Edit: He already posted the info in the TPU thread here. Now to figure out how to do it.
That makes sense, maybe the PLL isn't supported. I noticed the FSB did go up, but it seems strange that it hard locked with only ~+7 with or without an HDD attached. I'm glad the OS was on a flashdrive, I had no idea there was a virus warning. Thanks for pointing that out (+rep).SetFSB PLL support ended just about when X58 came out. Hit or miss for which PLL are supported. The website has a virus warning. I just boot a Linux Live DVD with no drives connected and go there anyway.
If you can disable ECC in the BIOS you may gain something from the reduced overhead. But maybe the JEDEC timings won't change.
The notebook modders use RW Eveything maybe you cna get some ideas here.
Thanks for providing your settings! Seems I'll have to adjust something as windows wouldn't boot (I left my DRAM settings as is, as we have different RAM).I have a X5660 @4.2Ghz for a few years no issue with a Gigabyte board and I set everything to auto except Vcore: 1.3000 and QPI/VTT 1.235 and DDR voltage or what ever it's called set to 1.64 but I don't know if you need to set yours as high as that as you have a lower speed RAM than me. In the spec sheets my RAM need 1.65V at 1600MHz. Oh yea and Load line calibration set to standard.
Oh I forget to mention that is just the voltages obviously I have turbo off and all other settings in the turbo page turned on all power saves and BCLK set to 200 and 21 multi and QPI 36, Uncore 16, and spd 8 but I need them like that as I have 1600MHz RAM.
Also I have the correct timings for my RAM set got them from the manufacturers website. I also do not have XMP set either.
I'm actually running a spectre patched version of FH.They are discussions overclocking on locked motherboard.
I just remembered are you on the latest beta bios from the Gigabyte website? As I had issues a few years ago by not being on the latest bios.
The +7 overclock is just about what I get with setFSB on my Optiplex 380 X5470. It's such a barebone office computer there's no extra PCIe slot to try an SATA work around. While you're at the Notebook site you might check out the TME unlock method. My programming skills =0 so my understanding is very limited. But the PLL can have a hex bitunclewebb helped me make sense of overclocking with MSRs and I was able to make a script for it, which is posted on TPU here: https://www.techpowerup.com/forums/threads/throttlestop-overclocking-desktop-pcs.235975/page-27. MSR documentation for (all?) CPUs is here: https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf . Finally got the CPU overclocked in Linux and doing a 24hr stress test with 80% RAM.
That makes sense, maybe the PLL isn't supported. I noticed the FSB did go up, but it seems strange that it hard locked with only ~+7 with or without an HDD attached. I'm glad the OS was on a flashdrive, I had no idea there was a virus warning. Thanks for pointing that out (+rep).
No ECC option unfortunately. I'll have a look at RW Everything and the link you provided, thank you.
Thank you for all the detailed info. Evidently a lot of laptop users are dedicated to overriding BIOS limitations. I can see why, it's fascinating stuff.The +7 overclock is just about what I get with setFSB on my Optiplex 380 X5470. It's such a barebone office computer there's no extra PCIe slot to try an SATA work around. While you're at the Notebook site you might check out the TME unlock method. My programming skills =0 so my understanding is very limited. But the PLL can have a hex bit
set at boot that puts a hard limit on FSB speed. Since X58 CPUs all run the same BCLK there might be a default setting that's been applied. To avoid spoofing with a resistor it then outputs a 33Mhz PCI frequency. I think the edit tool in SetFSB might be able to access this and change it. BIOSmods.com does do some work with Dell BIOS. I haven't heard any mention of TME unlock there though. iIf you can get the datasheet for your PLL you can locate the pin involved. That's far as I got with it. The BIOS modders might be able to do this if you could tell them what you want done.
This was used by Dell in LGA775 to lock the 65nm to 266fsb, 45 nm to 333fsb, and only the high end dual Xeon workstations got 400fsb CPU support. This defeated the pinmodding overclocks back then. They may still be using it.
It might be interesting to lower the fsb and add another multiplier or 2 just to see how much control SetFSB actually has available.
I noticed that as well, but I thank you for pointing it out all the same. His/her comments piqued my concern with running too much power through the VRM, though I do think the numbers mentioned seem (extremely) conservative. I should pull the heatsinks one of these days and see what components were used for power delivery.Here are some OC settings someone else got for that system.
I couldn't tell if the 145 FSB reading was just a glitch or not. Looking back through screenshots, perhaps not. FSB/DRAM starts to fluctuate higher along with the uncore/QPI while the effective clock speed remains the same (only happens at or above x30 ~4GHz). Then again, HWiNFO threw me several garbage readings (at one point x37 4.9GHz single core). I'll have to try it again with other software to get a second opinion.If you're getting 145fsb then you don't have a TME lock.
Thank you for sharing that Dell rebranded MSI for X58 boards (+rep). I couldn't quite tell if they were Foxconn or MSI or something else. I'm surprised the Alienware ALX and 435MT can accept the same BIOS. From the images I'm seeing on ebay, they have a different number of DIMMs and SATA, which is very interesting, I never knew that was possible. I could've sworn I saw an MSI similar to the 0x501h, but couldn't find it again last night. I will have another look at some point.It looks like the Alienware Area 51 ALX, and Dell XPS 435MT can swap BIOS, but not the XPS 435T/XPS9000. Alienware went with an ASUS P6T for the bigger tower version.
The XPS730x with H2C peltier had some factory overclock settings. It's another MSI MB. Maybe the BIOS modders can splice a BIOS together for you?
MS 7543 is that MB.
It looks like The Studio XPS 435MT (Micro ATX) got an overclocking BIOS through Alienware as the Aurora ALX.
The Full Tower XPS 730X got overclocking when ordered with a Hybrid Peltier cooler H2C, This may also be an Alienware Area 51-ALX I7 with a water loop.
The XPS435T/ XPS 9000 got left out due to Alienware using an ASUS MB for the midtower systems. I think it's an MSI MB like the other 2 Dell X58 XPS systems.