If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Rate Thread | Display Modes |
#1
|
|||
|
|||
Windows 10 Ultimate performance power plan
On 2019-06-10 2:23 p.m., Paul wrote:
Rene Lamontagne wrote: I set my system to this plan today for a test and really didn't see much difference to the High performance plan, Under prime 95 the CPU ramped toÂ* about 4.34 GHz compared to about 4.28 GHz in the high performance plan, temperature stayed the same at about 66 deg C. System was drawing 128 watts full throttle. At idle it was drawing about 44 watts no load and about 39 deg C. Anyone here tried it or will it really benefit only heavy usage programs Â*or newer high end games. Rene Intel lists this on the Ark for i7-8700 Â*Â* 6C 12T Â*Â* Processor Base FrequencyÂ* 3.20 GHz Â*Â* Max Turbo FrequencyÂ*Â*Â*Â*Â*Â* 4.60 GHz This site has the turbo table. https://en.wikichip.org/wiki/intel/core_i7/i7-8700 ModeÂ*Â*Â* BaseÂ*Â*Â*Â*Â*Â*Â* Turbo Frequency/Active Cores Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* 1Â*Â*Â*Â*Â*Â*Â*Â*Â* 2Â*Â*Â*Â*Â*Â*Â*Â*Â* 3Â*Â*Â*Â*Â*Â*Â*Â*Â* 4 5Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* 6 NormalÂ* 3,200 MHzÂ*Â* 4,600 MHzÂ* 4,500 MHzÂ* 4,400 MHzÂ* 4,300 MHzÂ* 4,300 MHzÂ*Â* 4,300 MHz So the first question I've got it, "why turbo on all six, yet have a base of 3.2?". I thought the concept of Turbo, was to increase clock when fewer cores were running. The TDP of 65W is probably met when running at base. There might also be a BIOS option to "lock the cores together", but I don't know what benefit that has in the real world ("user can't notice any difference"). There's some issue about maintaining the running OS clock when the Cores run at different frequencies. There's some sort of handoff in the OS which makes corrections or something. Locking the clocks on the cores together, means that code doesn't have to work too hard. On my motherboard, I had to select operating conditions that prevented VCore heatsink temps from going sky-high. It used to be, that VCore prevention-of-thermal-runaway was done via passive design. My current motherboard "has an acre of copper" that keeps Vcore at about 38C all day long. That philosophy stopped some time ago, where power limiters were fitted to VCore, presumably with the idea that a MOSFET couldn't go into thermal runaway. And that's how they can justify running the teeny tiny Vcore heatsink at a temperature that burns flesh. (Mine went over 65C. Checked with digital thermometer.) So rather than doing what you're doing, I had to be content with running the machine at conditions that didn't leave VCore at high temperatures. ******* If you test with Cinebench, you might notice the processors with ring connections, kinda "starve". This means, if you find a way to run them at high speed, there might not be enough memory bandwidth to get a benefit. My 6 core processor, benches at "5 cores", which means the ring bus is slow enough to eliminate the effective performance of one core. Other processors Intel makes, are "grid connected", with horizontal and vertical busses. Which means in terms of raw bandwidth, there's more capability between cores. There is still the possibility though, that a puny dual channel memory bus, can't keep up. This means, for core-bound computing, where the operands live in local cache, you're probably getting a "win" out of the setting. But under more "real-world" conditions, you have to wonder whether the uncore and memory stuff can keep up. As an example, a sad example, consider my Athlon XP3200. Paul overclocks it a bit, because all the kids are overclocking theirs, and Paul wants to see what all the fuss is about. Paul runs stock 2200MHz, tries 2600MHz, benches and... "no difference". In other words, the infrastructure available back in those days (Northbridge and "thin bus" to CPU), means there's absolute no point in overclocking! The chipset is gutless. Your machine no longer has a Northbridge. All of that is pulled inside the CPU. This means, your CPU has a lot more cores, and is better able to feed them. But at the edges, when everything is cranked, the architecture might still not be "over-generous". And if you find a good benching tool, it might tell you whether all this fuss is really worth it. There are two patterns for game performance. Most games follow the "boss thread" principle. One core is railed and runs all the other threads (synchronizes them). The other cores run at 30% or less. The hardware drivers have the interrupt load spread over all the cores like peanut butter (IOAPIC?). This pattern is a good fit for the Turbo Table. The other pattern, might be strategic games, where the graphics animation isn't all that flashy, and most of the computing goes into AI threads. Perhaps stuff like Fritz Chess would fall into this category, of loading the threads evenly. And then the problem just might be the "starved for memory" one, depending on how often the game is forced to consult something in main memory. ******* Your job as an experimenter then, is to tease out these details with benchmarks, and like my lousy AthlonXP experiment, figure out what's really worthwhile :-) After discovering that 2600MHz was useless, I turned it back down again. I like a free lunch as much as the next guy, but not if it's "rice cakes". I've long since lost track of the ideal conditions for the modern CPUs. I'm just not running into as many "experimental results" web pages as I used to see. Â*Â* Paul Thanks Paul, Boy that Wikichip site sure gives a tremendous amount of info on that i7 8700 CPU, a lot to digest. I did some more testing and ran various benches on the different power plans and came to the conclusion that there really is not much advantage to running the Ultimate and High performance plans, at least not in my case. The best it would do in Ultimate was 4.34 GHz and the temps stayed at 66 Deg C even after a half hour run of Prime95. I did learn something though, by watching Task Manager while playing a game of hearts, When playing steadily it would turbo up to 4.28 GHz and when at rest between plays the speed would drop down as low as 1.02 GHz, which supersized me, I never thought it would drop that low. My old i7 950 would only ever drop to 1.56 GHz So seeing all this I will leave it at Balanced plan, I am glad to see the temp staying so stable and reasonable under full load.. Rene |
Ads |
Thread Tools | |
Display Modes | Rate This Thread |
|
|