2 hours ago, dominikv95 said:

Based on your and the other opinions in this thread, I will certainly be overclocking it with the company who builds it - it's only about £150 more expensive to have them overclock it. 

As for the H115 pro, are you referring to the cooling system? Is cooling really important for the CPU? What does it actually help with: performance, longer life or some other factors?

You're yet another person (I use other forums too) to advise on having 2x16GB vs 4x8GB RAM - why's that?

Oh and what do you mean, you'd suggest 'the better warranty'? Do they offer different ones?

EDIT: and finally, how do you know what company I am talking about? I don't think I mentioned a company name on here, have I? 😄 I didn't want to break the rules by advertising etc. Just curious to see how you got to that, and also whether it's actually the same company haha 😄 

Thank you for you help!

the H115i pro is for cooling yes. Its important to keep the CPU as cool as possible, especially if its overclocked. Heat is your enemy. It will help with both performance and longevity.

2x16 is better than 4x8, as its less stress on the memory controller i believe.

When i bought mine, the were 3 warranty options in the drop down menu where you build the system. 1 year return / collect rather than 3 months.

...the clue...

_{"PCS Ultra Quiet fans"}

😉

 

2 hours ago, w6kd said:

The most significant difference between the 9700K and the 9900K with HT off is 33% more level-3 cache on the i9...16MB vs 12MB.  I am in the camp that believes memory performance is generally and too-casually overlooked when trying to max the program.

The price delta between the 9700K and the 9900K on the day I picked up my 9900K (for my portable rig) was $60...for me, that extra cache was something I really wanted to have paired up with my 8.33 ns RAM.  It's "nice to have" for most folks, I think...probably not worth $100 more for the average bear.

Regards

i agree , the 9900k is a better binned cpu in general , if you disable HT it do 200-300mhz higer with same vcore then the 9700k, and the extra cache dont hurt.

2 x16 vs 4 x 8gb its more easy to get 4x8 to work then the 2x16 sticks ( dual rank) they can be tricky to get over 3600mhz , i run 2x8gb 4266 c16 1.45v 7.5ns give a nice boost in P3D

Edited July 26, 20196 yr by westman

20 hours ago, w6kd said:

P3D depends on fast single-core performance because the main thread, which drives the train for all the others, runs on a single core, and the faster that core runs, the better your sim performance.  When a stock-configured CPU downclocks due to turbo boost scheduling or power constraints, it downclocks *all* of the cores together.  P3D makes use of all the available cores, which in a stock 9900K system will drive all of the cores down to the lowest (4.7GHz) 7/8-core turbo boost multiplier, including that all-important one on which the main thread runs.  And when the other cores are busy doing real (and AVX-enabled) work processing/loading textures and terrain, all of the CPU cores will downclock even further to keep the power levels within TDP limits.  So the difference between running stock and overclocked to 5 GHz isn't just a 300MHz bump from 4.7 to 5.0, it's from something significantly below 4.7 up to 5...from as low as the base clock of 3.6 GHz, in fact (which would be a 39% increase).

I understand all of that. The only thing which should stop the CPU from running at its designed turbo frequency, which should be 4.7GHz on all cores, is thermal throttling (or downclocking, as you refer to it), not specifically power constraints - TDP isn't the maximum power the chip is allowed to run at (or a limit), it's the average power it dissipates under load at base frequency. If thermal throttling is occurring without overclocking (not just turbo boost), then there's almost certainly something wrong with the CPU cooling solution. I certainly wouldn't accept a system which was incapable of running continuously at stock turbo speeds without exceeding the thermal limit. As the 9900k should be capable of a 4.7GHz turbo boost on all cores, I stand by my original comment that I can't see how you would get a 25-30% performance increase by overclocking it.

Edited July 27, 20196 yr by vortex681

Try overclocking a board without increasing the power limits and see what happens.  The chipset doesn't "throttle" the CPU, it downclocks it to keep it within the design power limits.  Why, otherwise, would they have power limit settings in the BIOS?

If I run AVX-enabled Prime95 on my board at stock settings while monitoring the temps and core clocks, the 9900K will downclock below the all-core turbo boost of 4.7 GHz with CPU temps in the 70s...still well below the 100 deg C thermal limit that would force the CPU to start throttling itself..

 

 

55 minutes ago, w6kd said:

Try overclocking a board without increasing the power limits and see what happens.  The chipset doesn't "throttle" the CPU, it downclocks it to keep it within the design power limits.  Why, otherwise, would they have power limit settings in the BIOS?

If I run AVX-enabled Prime95 on my board at stock settings while monitoring the temps and core clocks, the 9900K will downclock below the all-core turbo boost of 4.7 GHz with CPU temps in the 70s...still well below the 100 deg C thermal limit.

Then I can only think that your MB is limiting what the CPU is capable of. There are plenty of reviews of the 9900k where they seem to be able to run CPU benchmarks at 4.7GHz on all cores without the system reducing the clock. Some run quite hot (for example 85°C) but don't throttle or downclock. Intel uses a set of variables called Power Levels (PL) for the CPU. PL1 is the normal cooling limit (TDP), PL2 is the maximum sustainable power that the CPU can handle (turbo) until thermal issues occur and PL3 is the absolute power delivery limit. Intel has set the value of PL2 for the 9900k to 210W, which is significantly above the normal TDP of the chip. As board manufacturers can set their own levels, it's quite possible that ASUS has set them lower, which could give the downclocking you observe.

The other thing to consider is that, outside of benchmarks, the chip will rarely (if ever) run all cores at 100% load so it's highly unlikely that most people wouldn't be able to maintain 4.7GHz on all cores in P3D. Here's an example of gaming on a stock (not overclocked) 9900k with all cores running at 4.7GHz: https://www.youtube.com/watch?v=xJzxUNXYqYA

 

Edited July 27, 20196 yr by vortex681

I'd bet you that they are not running those benchmarks without overriding the power delivery limits in the BIOS.  But, it might be that some motherboards don't have the standard power delivery limits as default.

P3D doesn't present a steady 100% load, to be sure, but it does spike the CPU loads up pretty close to 100% across the cores at times, such as when it's loading a new scenery tile.  The last thing I'd want is to have my CPU clocks in a real-time process fluxing all over the map.  A lot of folks set AVX offsets as part of their overclocking drill, too, and then never take them back out, and unbeknownst to them, their CPU mults are fluttering back and forth during sim sessions because P3D uses AVX instructions that force the cores to downclock per the offset when those AVX instructions are executed.

21 hours ago, dominikv95 said:

 

Thanks, yeah I've already scrapped the idea of that 650W haha I will look at the monitor too, thanks!

 

 

 

650 Watt IS sufficient! Don't be fooled into thinking you need a very high wattage PSU. However... my preference would be 750 watt. It's not much more expensive and does nudge the system closer to the most efficient PSU range. It also gives you a little more margin for what the future might hold. Having said that, modern gold or platinum PSU's are very efficient, and maintain that high efficiency right up to 90% load. 

 

I have a 750 Watt EVGA 750 platinum and it handles my 8700K with a heavy overclock with no issue whatsoever. 

 

The Corsair 650 efficiency below...

 

Edited July 27, 20196 yr by martin-w

8 minutes ago, w6kd said:

I'd bet you that they are not running those benchmarks without overriding the power delivery limits in the BIOS.  But, it might be that some motherboards don't have the standard power delivery limits as default.

As these are supposed to be representative real-world tests, I'd be very surprised if they were overriding settings without saying so. What would they achieve by doing that? It's also possible that the ASUS board can't reliably deliver the 210W PL2 power required for maximum performance.

Edited July 27, 20196 yr by vortex681

6 hours ago, vortex681 said:

As these are supposed to be representative real-world tests, I'd be very surprised if they were overriding settings without saying so. What would they achieve by doing that? It's also possible that the ASUS board can't reliably deliver the 210W PL2 power required for maximum performance.

The ASUS board's power regulation does just fine, but it does require the default power limits to be overridden in the BIOS.

I'm not sure why a real-world test of the CPU would imply using default settings in the BIOS/chipset, any more than it would somehow be a foul to use a high-end CPU cooler for the test, or XMP memory profile rather than the default.

 

41 minutes ago, w6kd said:

I'm not sure why a real-world test of the CPU would imply using default settings in the BIOS/chipset, any more than it would somehow be a foul to use a high-end CPU cooler for the test, or XMP memory profile rather than the default.

Most reputable review sites tell you how they set up the system for testing and to over-ride the default BIOS settings would certainly be something they're likely to mention or comparisons with other stock CPU results wouldn't mean anything. Even if they're not using default settings, many still seem to manage 4.7GHz on all cores without the system clocking down. It just seems odd that you can't.

Edit: I may owe you an apology as I thought you had a 9900k! I see you have an 8086k and that's probably a whole different ball game. My comments about not achieving a 25-30% performance increase when overclocking were in reference only to the 9900k.

Edited July 27, 20196 yr by vortex681

No apology needed.  I just built a LAN box PC with a 9900K and an ASUS Maximus XI Hero for portable use while travelling...it's not my primary simming rig.  Right now I have FSX:SE loaded on it and am enjoying revisiting some old FSX favorites.

I'd be more concerned about tests run with unmentioned BIOS constraints in place that could affect the results.

 

On 7/26/2019 at 7:15 AM, vortex681 said:

Would anyone actually notice a speed difference of 0.25 Ns? Also, is the small gain you'd get in performance (very small, in some cases) actually worth up to twice the difference in price between 3200Mhz and 4000Mhz? If you really want those last few extra FPS and money is no object, then go for the fastest RAM possible. For the rest of us, 3200Mhz should be good enough. Faster RAM tends to be much more significant with AMD processors than with Intel.

Yes they would like a said there is plenty of Benchmarks to support it. I was just giving my opinion... and yes if you are on a budget then 3200Mhz will be the best option. 

On 7/25/2019 at 10:37 AM, dominikv95 said:

Hi everyone!

<snip>

Thank you

Dominik

Hi Dominik, looks like you've put together a pretty good list of specs for your new PC. 

First off, good choice on the CPU.  You mention "upgrading" to a 9920x - I can assure you this would not, in fact, be an upgrade for gaming (or simming).  The vast majority of games/sims cannot take advantage of more than 8 cores in any meaningful way.  The most you would gain from 12 vs. 8 cores in P3D is faster scenery loading.  Not only that, but I guarantee you that a stock 9900k is going to run games/sims at a higher framerate than a stock 9920x due to the difference in Turbo clock speeds offered by these CPUs.  On top of that, if you overclock you will almost certainly reach a higher clock speed with the 9900k.  Bottom line, games/sims today care more about clock speed than core count, so you should go with the CPU that offers higher clocks and in this case that's the 9900k.

Second, as others have mentioned, there is performance to be gained by using faster RAM.  There are two aspects of memory performance to consider, latency and bandwidth.  These are expressed through ratings of clockspeed and timings.  To put it simply, clockspeed should be high, and timings should be low (smaller numbers).  A good example is DDR4 in the 4000MHz range and with primary timing (CAS) of 16-18.  

Third, if you want to run at high resolution with high graphics settings and lots of 3rd party add-ons, you may want to consider an RTX 2080 Ti rather than a 2080/2080 Super.  The 2080 Ti has more VRAM onboard (11GB vs. 8GB).  It's possible to push past 8GB of VRAM usage in this scenario, and with a card that only has 8GB you may experience stutters as the graphics card has to fetch data from system RAM.

Fourth, the motherboard you have selected, while adequate for stock 9900k operation, is perhaps not going to be the best choice for overclocking.  I recently built a 9900k/Z390 system and while my last several PCs have used Asrock or Asus motherboards, I would suggest a Gigabyte or EVGA board for Z390 specifically due to the sheer inadequacy of VRM cooling on the majority of Z390 boards out there.  Gigabyte's Aorus line of Z390 motherboards has real VRM cooling, not just a solid block of aluminum like most other manufacturers, as do EVGA's Z390 FTW and Dark boards, though EVGA's boards are more oriented towards extreme overclockers and probably not the best choice for a beginner.  I went with the Z390 Aorus Master board myself and it's been great in the month or so I've had this PC.  The CPU runs at 5.0GHz on all cores with very low voltage and low temperatures.  Since I intend on not upgrading again for awhile I decided not to overclock to the limit which for this chip is around 5.2-5.3GHz with the cooling I have.  

Fifth, your cooling choice.  While the H100x is a fine CPU cooler, you have to remember that the 9900k is a very hot chip, especially when overclocked.  The more cooling the better.  I don't know what options your PC manufacturer offers, but a small increase in price can yield a significant increase in cooling capacity, which may ultimately allow you to overclock higher, and potentially affect the longevity of your system.  Also the "standard paste choice for sufficient cooling" line is slightly worrying.  I don't know what thermal paste the manufacturer intends to use, but if it is sub-par you're literally saving pennies only to severely restrict the cooling performance of your CPU cooling system.  Spending a tiny amount more money here can make a large difference in cooling system performance.  I recommend Thermal Grizzly Kryonaut for a thermal interface material to be used between a CPU IHS and water cooling coldplate.  It goes on easy and lasts a long time, while also conducting heat at a higher rate than most other thermal pastes on the market.       

Sixth and I would say finally, your power supply choice.  Somewhat against "conventional wisdom" here (more like urban legend in this case) you don't need a million gajiggawatts for a power supply in a high-end PC.  I have a 9900k and 2080 Ti in my sim PC and it is powered by a meager 750W power supply, only slightly larger than the one you picked out.  The key with power supplies is to make sure you buy a quality unit from a major brand name, with an 80 Plus rating of some sort (Silver or better).  You can overclock your hardware as high as it will go and honestly never see your power supply reach 100% capacity in even a worst case scenario.  All these online power supply calculators that lead to the rise in "conventional wisdom" are using inflated ratings to derive their calculations, and in all likelihood are nothing more than marketing tools to sell more expensive power supplies.  I've been building and upgrading PCs for over 20 years and have never needed to increase power supply capacity in any system I've ever built.  Just make sure you don't go with the bare minimum (i.e. if your graphics card requires at least a 500W power supply) and get something from a major manufacturer and you'll be fine.

Other than that, everything looks good.  We can debate about the merits of buying vs. building but that's a personal preference so I'll leave it up to you.  PC building can be as much of a hobby as anything, and if you just want to enjoy your sim without the potential headache and stress of building the machine then that's your call.  

12 hours ago, Alexandre6463 said:

Yes they would like a said there is plenty of Benchmarks to support it. I was just giving my opinion... and yes if you are on a budget then 3200Mhz will be the best option. 

While I'm all for faster than average memory in any gaming PC, I think this subject can lead to hyperbole.  When vortex said "Would anyone actually notice a speed difference of 0.25 Ns?" the answer is no.  The human brain cannot process quickly enough to allow anyone to distinguish a difference in .25ns latency.  You literally can't even define .25ns in human terms.  We don't even delve into the realm of perceptibility until we get into the millisecond range, several orders of magnitude larger than the difference between "fast" RAM and "very fast" RAM, which is honestly what you two are discussing.  As someone who was heavily engaged in benchmarking the effects of RAM performance on MS Flight Sim back in the days of FSXMark, I can state with confidence the following:

1) faster RAM is better - to a point

2) for modern Skylake-derived Intel CPUs, that point is somewhere in the 3600-4000MHz range, depending on the application, CPU clock, and mesh/uncore clocks.  

On 7/26/2019 at 9:31 PM, MaDDogz said:

the H115i pro is for cooling yes. Its important to keep the CPU as cool as possible, especially if its overclocked. Heat is your enemy. It will help with both performance and longevity.

2x16 is better than 4x8, as its less stress on the memory controller i believe.

When i bought mine, the were 3 warranty options in the drop down menu where you build the system. 1 year return / collect rather than 3 months.

...the clue...

_{"PCS Ultra Quiet fans"}

😉

 

Thanks. And yes, I can see the clue now 😄😄😄 

Edited July 28, 20196 yr by dominikv95