So, the W3690 (equivalent to an i7-990X) processor's TDP is 130 Watts. If I leave the TDP Limit to 130 Watts any overclock increase in multiplier over stock will throttle after awhile. Even x27 6-core Turbo, which is within Intel Turbo spec, will throttle after 5 minutes. But at the stock TDP Limit of 130 the motherboard's VRM should be ok and not overheat, even for 24/7 hours-long full load runs. See chart below, which is from testing my system while running Prime95 on default "Blend" setting on all threads.
On the other hand, to eliminate throttling for overclock multipliers the TDP Limit must be increased and a lot more power will go through the 435T CPU voltage regulator. (I'm not worried about the CPU - it's fine at 70C with an Intel DBX-B CPU cooler.) My W3690 needs a TDP Limit of 175W to run at 4 GHz without throttling (I'm not stable at 4 GHz, but can run for 3 to 50 minutes before something crashes). Even my stable 3.86 GHz overclock needs a TDP Limit of 165W to run without throttling.
There doesn't seem to be any VRM temperature sensors on the 435T, so I have no idea if it's ok or overheating. I'm thinking the 435T motherboard can't 24/7 deliver 165W to the CPU without shortening it's life...
Thank you for the detailed posts, as always.
Your postings in the past about possible upgrades for the xps9000 / 435T has been a valued resource, and it is appreciated.
With the work around for the turbo boost using throttlestop, does this mean that even increasing the base clock mulitplier +1, will cause an issue even if you do not increase the wattage above the stock 130TDP ?
Based on your post above that indicates the stock MOBO may have issues with some components, ie the VRM if one were to try and adjust the mulitplier, and power settings. I too think that this may cause issues with the service life of the machine, as the boards weren't intended for this on these setups.
If it is safe to boost the base clock +1, to at lease obtain the added horsepower that the CPU was designed for (even though it is marginal), like the turbo boost feature, I would consider that option.
Do you still think the 3.59ghz (or slightly higher) option is possible, for long term use without prematurely shortening the lifespan of the board ?
The Turbo workaround just enables standard Turbo Boost. Since the system was designed to Turbo with 130W TDP CPUs, I don't see an issue as long as the default 130W Turbo Power Limit is left in place.
If I bring up ThrottleStop on a PC with working Turbo, like the 435MT with a W3580 (similar to an i7-975), it picks up the existing settings which already have the Multiplier set to 1+ with a T next to it, "26 T" for the W3580 which has a stock Multiplier of 25. The workaround just does what the BIOS should have done to enable Turbo.
What I didn't realize was how Turbo interacts with TDP. I assumed the CPU would never use more power than the TDP if not overclocking. This doesn't appear to be the case. It appears that while TDP wattage isn't exceeded in a sustained power-usage case, it can and is exceeded (apparently up to ~145W for a non-overclocked CPU) for short durations of several minutes during normal all-core Turbo.
This has several implications. The electronics of the CPU voltage regulator (VRM) on a normal motherboard should be designed to supply at least ~145W. This means the VRM electronics should have some robustness over the 130W TDP, which is a good thing. However, the VRM heat dissipation (heat sink) only has to dissipate 130W on a sustained basis.
I think Intel engineers came up with Turbo when they realized that CPU temperature lags power usage. Meaning the CPU and it's surround mass (heat sink, etc.) take time to heat up when power usage goes up. While this is happening additional power/frequency could temporarily be taken advantage of and exceed the TDP until temperature rise caught up. Thus, the idea of Turbo Boost was born, I believe.
To me this partially answers the question of why have a TDP Power Limit and associated throttling in the CPU at all? Why not just throttle based on temperature when the CPU is too hot? I personally believe Turbo inside the CPU is quite sophisticated and actually collects run-time statistics to calculate and use the heat-related mass of your CPU heat sink, it's ability to dissipate heat and your systems steady-state temperature for a given power usage on your system. Simply stated, Turbo knows your specific systems CPU temperature at a steady-state power usage (130W for a non-overclocked CPU) and allows exceeding TDP until temperature gets there. That's what Turbo is from an heat engineering perspective, I think.
I posted the wattage/throttling chart partially because it shows wattage at various frequencies at least on my system. Note this is while running Prime95 (Blend setting), which is likely more intensive than other usage.
While my testing shows my system's motherboard VRM can apparently supply 175W to the CPU, I don't feel comfortable running continuously at that level of power and associated VRM heat. On the other hand, I could probably place a fan on the VRM to cool it, but without a temperature sensor, I really wouldn't know if the cooling is sufficient.
I haven't posted how I did the overclocking yet because I wanted a better understanding of motherboard heat and longevity issues before doing so. My personal feeling right now is that if overclocking the Turbo Power Limit needs to be left in place, more or less as is, at 130W for the 435MT (black Mini Tower) due to it's limited heat dissipation (limited air-flow and smaller case size). For the 435T/9000, my feeling is the Turbo Power Limit can be loosened up a bit to 140W or so (this is ~7% over 130W), but still needs to be left in place so throttling occurs to limit heat in the motherboard's VRM.
I did want to mention that Throttling only seems to occur down to stock multiplier and is a mix of available cores. Usually, when throttling starts it's only one or two cores partially throttling. Heavier throttling shows all cores partially throttling somewhere between stock multiplier and max multiplier. I've never seen throttling occur so heavily that all cores are at stock multiplier, even at 4 GHz. Additionally, if using all cores at a fixed frequency, but at less than 100% utilization, CPU will take longer to throttle and may not throttle at all, even with extended 4 GHz usage.
This raises a quandary for specific system testing to determine it's maximum stable overclock, though. Does one need to risk motherboard VRM potentially shortened lifespan by temporarily removing the Turbo Power Limit so throttling doesn't occur to determine stable overclock frequency (which usually involves running something intensive like Prime95 for several hours)?...