Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mojave]

I'm building a new HTPC using a Haswell-E i7-5820K processor. All parts have arrived and are listed below.

Case:  Silvertonetek GD07 with rack ears
CPU:  Intel I7-5820K
Motherboard:  ASUS X99-Deluxe
Thunderbolt card:  ASUS ThunderboltEX II Dual
Memory:  GSkill Ripjaw DDR4 2400 (4x4GB)
PSU:  Seasonic Platinum SS-860XP2
SSD:  Samsung XP941 256MB 4X M.2
CPU Cooler:  Corsair Hydro H75
Blu-ray drive:  ASUS 12x Blu-ray internal drive
OS: Windows 8.1 64-bit

I will be using my GTX 550 Ti video card temporarily. I may get the new GTX 970 STRIX.

The Samsung XP941 M.2 SSD is very small. It came in an envelope from newegg via USPS.

My son wants to help put this together so I'll have to wait until tonight. The case has plenty of space for more internal drives so I may add some for storage. I currently use a NAS (very old and slow) and some external USB 3.0 3TB drives.

[Hendrik]

That is one serious HTPC. Are you sure you're not mislabeling it slightly?
I'm totally excited about my development system rebuild, but I'm still trying to keep myself from ordering now, and instead waiting for things to settle, NVIDIA to announce, etc. Also hoping for some new SSD miracles. I want a 1TB SSD, and unless a M.2 PCIe shows up real soon, I'll go with a 850 PRO instead.

[mojave]

All I care about is the JRMark benchmark score. 

I keep thinking I'll find time to play some video games at home, but with 5 kids and a wife there isn't much free time. So, it will be primarily used for audio playback, football recording/viewing, and the weekly Blu-ray movie. Don't tell anyone  , but I watched Mary Poppins last night.

I'll probably keep it for 4 years at least. My 4 year old work computer is an i7-920 OC'd to 3780 Mhz with 12Gb of RAM and a GTX 660. It is on 24/7 and is rock solid. It still feels extremely fast.

[6233638]

I will be using my GTX 550 Ti video card temporarily. I may get the new GTX 970 STRIX.

For what it's worth, some concerns have recently been raised about the STRIX cards.
While the card monitors the GPU temperature to control the fan speed, it seems that the rest of the board is not monitored, and some components (e.g. VRMs) are running dangerously hot when checked via thermal imaging cameras - reaching temperatures over 100℃
 
Since this is a recent development, and the 900 series is due soon, I suspect there will not be enough time to address this.
Of course the 970 should consume a lot less power and run cooler, so perhaps it would not be an issue.
 
Edit: ASUS 780 STRIX
MSI 780 Gaming OC (essentially the same noise level under load; 1dB quieter)

[Sparks67]

My son wants to help put this together so I'll have to wait until tonight. The case has plenty of space for more internal drives so I may add some for storage. I currently use a NAS (very old and slow) and some external USB 3.0 3TB drives.

I recommend that you don't put the drives in the case, because of heat and possible vibration in the case. I had the issue with my second HTPC build, but it the drives would vibrate in the case.  Yeah, all the latest ways to keep the drives quiet.   If you don't want to do a NAS, then another option is a sas expander.   http://www.sasexpanders.com/ Depends on your storage needs. 

[6233638]

I recommend that you don't put the drives in the case, because of heat and possible vibration in the case. I had the issue with my second HTPC build, but it the drives would vibrate in the case.  Yeah, all the latest ways to keep the drives quiet.   If you don't want to do a NAS, then another option is a sas expander.   http://www.sasexpanders.com/ Depends on your storage needs.

I really like the idea of moving all of my storage except the SSD outside of the main PC case.
Expansion is becoming a problem for me, and keeping drives cool without making a lot of noise is increasingly difficult.
But I don't think it's so big a problem that I can justify spending $1400+ for an enclosure, plus the cost of a SAS card (another $800) and be limited to a 10m range.
 
Hopefully Thunderbolt will solve these problems. It would be cheaper to build a complete PC with Thunderbolt, than it would be to buy one of these enclosures and a SAS card, and there are now optical thunderbolt cables which go far beyond 10m.

[Hilton]

I have a couple external USB3 and Esata RAID enclosures that are cheap and reliable.
Also a 4 bay backup dock for 4TB greens that I use for backups with Acronis Backup 11.5 backing up all my machines over the network to the HTPC. (HTPC backs up locally)

4 bay which has been running 3TB WD greens in Raid 10 for almost 3 years now. $189 for enclosure
http://www.mediasonicstore.com/servlet/the-14/Mediasonic-PRORAID-4-Bay/Detail

8 bay which also has 3TB WD greens in Raid 10 for 2 years. $350 for enclosure
http://www.mediasonicstore.com/servlet/the-24/Mediasonic-ProRaid-8-Bay/Detail

4 bay backup Dock - using with 4TB greens
http://www.mediasonicstore.com/servlet/the-43/Mediasonic-ProBox-4-Bay/Detail

Reliable and FAST. No problem serving up multiple 1080P streams with a read speed of 200>300Mb/s and writes of 150>250Mb/s.

If you spend a bit more on the 4TB/6TB Reds you could have a shed load of storage in RAID 5 or RAID 50.

[Hilton]

That is one serious HTPC. Are you sure you're not mislabeling it slightly?
I'm totally excited about my development system rebuild, but I'm still trying to keep myself from ordering now, and instead waiting for things to settle, NVIDIA to announce, etc. Also hoping for some new SSD miracles. I want a 1TB SSD, and unless a M.2 PCIe shows up real soon, I'll go with a 850 PRO instead.

I just put a 1TB SSD in my Id. The Samsung was about $450 for the EVO and the Pro 1TB is $750.
The EVO is plenty fast enough. I'll be upgrading my 6 x 120GB Sandisk SSDs to 512GB or 1TB soon as im out of system capacity.

PS. nice "HTPC" Mojave..
PPS. You could always put an Id (or similar) with one of the 8 bay enclosures in a cupboard or other room and have 36TB usable if noise is really an issue.

[mojave]

I got the HTPC put together, but didn't have time to boot it up yet.

Here are some comments:
Case - Silvertonetek GD07: My current HTPC is in a Silverstonetek case that I've owned for about 9 years (2005). I really like their cases and recently did a build for a friend in a GD08 case. It is easy to work in and supports four 120mm fans and two 80 mm fans for lots of airflow options. It looks good and I like the door that hides the drive and USB ports. The case has a removable rack for all 5 1/2", 3 1/2", and 2 1/2" drives.
CPU - Intel I7-5820K:  Largest CPU size-wise I've installed, doesn't include a cooler
Motherboard - ASUS X99-Deluxe:  Easy to install, backplate had sound/vibration damping on it, layout seems good, 6 fan headers (plus more on an included add-in card), easy to read memory slot labels, M.2 slot for vertical installation of SSD M.2 drive, extra PCI-e M.2 expansion card included that supports 4x, Thunderbolt header
Memory - GSkill Ripjaw DDR4 2400 (4x4GB): These have a slightly curved bottom instead of flat. Very light.
PSU - Seasonic Platinum SS-860XP2: Nice packaging! Completely modular. Plenty of cables with various options. For example, some GPU cables are dual 6/8 pin. There was also a single 6/8 pin cable that I used. There were long and short SATA power cables with different quantities of SATA connectors. I only needed one short cable for the Blu-ray drive. There is a switch on the back for hybrid or normal operation. Hybrid has three stages:  fanless, silent, and cooling mode.
SSD - Samsung XP941 256MB 4X M.2:  Installed vertically into a special ASUS mount for the motherboard. Very small and easy to install.
CPU Cooler - Corsair Hydro H75:  I had less space between the case and motherboard than I expected. I ended up mounting the radiator to the inside of the case and putting one fan on the outside of the case blowing in. I'll have to see how it works. I'm not sure it will really work or fit once I try it in the final rack. I may need to switch to air cooling and rear 80 mm Noctua fans. It was very easy to install and only required screwing in studs to the motherboard, attaching the pump, and screwing down thumbscrews over the pump. No rear plate is necessary for an LGA 2011-3 socket.
Thunderbolt card - ASUS ThunderboltEX II:  This looks nice and installs easy. However, it comes with a 9 pin header cable for connecting to the motherboard. The manual says it only works with a certain ASUS X77 motherboard. The X99-Deluxe has a 5 pin Thunderbolt header. I found online that ASUS makes a 9 pin to 5 pin cable, but it is hard/impossible to get. Someone also said ASUS has recalled all current ThunderboltEX II cards so they can include the necessary cable to work with the new X99 motherboards. I'll have to look into it more today.

So far both the cooler and ThunderboltEX card have presented problems. I also got all done with the build and my son asked about removing the feet from the case. I forgot to remove them!  For the case to fit in my rack, I'll have to remove all components, unscrew the feet from inside the case, and reinstall all components. Maybe I can find an extra 1U space and put a spacer below the HTPC.

I pulled my GTX 550 Ti card with Artic Cooling GPU cooler out of my current HTPC (since it has also has video on the CPU) and installed in the new HTPC. Not that it really matters, but it matches the color scheme very nice.

I just checked online and the latest bios was released on September 4. I'm also downloading any recent driver updates. I'll be updating the bios next.

Empty case:

[mojave]

Motherboard and PSU installed:

[mojave]

CPU Cooler, memory, SSD M.2 drive and ThunderboltEX II card installed:

[mojave]

GTX 550ti with Artic Cooling GPU cooler installed:

[mojave]

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 3.122 seconds
    Single-threaded floating point math... 1.918 seconds
    Multi-threaded integer math... 0.595 seconds
    Multi-threaded mixed math... 0.398 seconds
Score: 3150

Running 'Image' benchmark...
    Image creation / destruction... 0.226 seconds
    Flood filling... 0.258 seconds
    Direct copying... 0.237 seconds
    Small renders... 0.877 seconds
    Bilinear rendering... 0.645 seconds
    Bicubic rendering... 0.349 seconds
Score: 8489

Running 'Database' benchmark...
    Create database... 0.141 seconds
    Populate database... 0.801 seconds
    Save database... 0.210 seconds
    Reload database... 0.039 seconds
    Search database... 0.640 seconds
    Sort database... 0.565 seconds
    Group database... 0.559 seconds
Score: 7275

JRMark (version 20.0.14): 6305

[6233638]

Interesting to see that the score is so close to the 4790K result posted recently. Seems like you need to start bumping up the clockspeed.

[kensn]

Awesome score !!  Nice rig..

[Sparks67]

I really like the idea of moving all of my storage except the SSD outside of the main PC case.
Expansion is becoming a problem for me, and keeping drives cool without making a lot of noise is increasingly difficult.
But I don't think it's so big a problem that I can justify spending $1400+ for an enclosure, plus the cost of a SAS card (another $800) and be limited to a 10m range.
 
Hopefully Thunderbolt will solve these problems. It would be cheaper to build a complete PC with Thunderbolt, than it would be to buy one of these enclosures and a SAS card, and there are now optical thunderbolt cables which go far beyond 10m.

Actually, I built an HTPC server, because of the cost was cheaper.     The motherboard is an ASUS P8Z77-V PRO/THUNDERBOLT.  Areca ARC-1882ix-16 http://www.areca.us/products/1882.htm   The case that I wanted wasn't available.  http://www.akiwa.com/product_detail.php?pid=2&id=713&img=pimages%2Frackmount%2FGHS-2000%2FGHS-2000-1.jpg  or Lian Li http://www.frozencpu.com/products/5244/cst-472/Lian-Li_PC-343B_Modular_Aluminum_Cube_Case_-_Black.html      I had a Chenbro 10769 case.  http://usa.chenbro.com/corporatesite/products_detail.php?sku=104, it holds up 11 or 13 drives.   Depends on the options. You can also do 2 hotswap SSD drives in the floppy slot.   Built like a tank, but it is quiet after you swap the fans.  I have 2 of them.  

The only issues with the Areaca Thunderbolt box is you are limited to 8 drives, and cost is $1800.  You can go to a farther distance with a SAS Expander.  Just need to look at their Thunderbolt option.  It is more expensive box, but distance is not an issue.  http://www.pc-pitstop.com/thunderbolt/  

Nice HTPC case.  http://www.hfx.at/ There is write about JRiver on one of their products.       

[mojave]

Last night I booted up the HTPC for the first time. I went into the bios so I could monitor temperatures. I let it run for about 30 minutes. The highest temperature during idle was 30C.

I then updated the bios. It was very easy and just required the bios file to be on a USB flash drive. Now that the OS is installed, future bios updates involve putting the USB flash drive in a certain port on the rear of the motherboard and pushing a button also located on the back of the motherboard. Anybody need a USB 3.5" floppy drive. 

I initially had trouble getting the OS to load on the M.2 drive. This article gave some help, but it still didn't work.  I needed to reboot/install about 4 times before getting the bios settings right that allowed for booting from the M.2 drive. I have no idea why Microsoft made the license code for Windows 8.1 so small and hard to read.

Once I got the OS loaded, I installed drivers and the ASUS utilities. I used the ASUS 5-Way Optimization to overclock.

The combination of automated overclocking, fan controls, power efficiency, digital power delivery and per-app customization have returned in full force with upgrades on the X99-Deluxe.

There are many settings in the Optimization and I went basic for the first time. The board overclocks by 100Hz and runs some tests and then repeats until it reaches a max. It does an analysis of all the fans and evaluates their minimum and maximum RPM and cooling capability and then comes up with fan profiles. It also creates several power management profiles. I'm looking forward to working with it some more.

The first automatic overclock came in at 4700Mhz - a 42% increase over stock! I saved this profile. ASUS allows you to assign clocking profiles to software programs. I assigned the overclock profile to JRiver and then ran the benchmark. I was watching CPU-Z during the benchmark and the CPU would drop back to idle speed several times. I think the benchmark could be higher if I set the clocking in the bios and turned off any C-States. I'll have to play with it some more.

The maximum temperature during 5-way Optimization was 74C. This is with only one fan on the Corsair H75 water cooling radiator. It dropped back down to the low 30's almost immediately after the CPU returning to idle. Overall, the fan noise was very quiet. However, it isn't completely silent at idle and I think it is from the three 120mm Silverstonetek fans. It could also be that I need to slow the fans on the Arctic Cooling system on the GPU. I'll be looking into it more in the next few days.

You can see from the benchmark how much faster the multi-threading is with the Haswell-e i7-5820k and extra 2 cores:

Intel i7 4790k @ 4487Mhz on Air (Johnny_Friendly)
Running 'Math' benchmark...
    Single-threaded integer math... 3.289 seconds
    Single-threaded floating point math... 2.015 seconds
    Multi-threaded integer math... 0.979 seconds
    Multi-threaded mixed math... 0.619 seconds
Score: 2753

Intel i7-5820K @ 4700Mhz (Mojave)
Running 'Math' benchmark...
    Single-threaded integer math... 3.122 seconds
    Single-threaded floating point math... 1.918 seconds
    Multi-threaded integer math... 0.595 seconds
    Multi-threaded mixed math... 0.398 seconds
Score: 3150

[mojave]

Overclockers.com has a review of the ASUS X99-Deluxe with lots of pictures. It is one of the most comprehensive looks at the motherboard that I've seen.

http://www.overclockers.com/asus-x99-deluxe-motherboard-review

[6233638]

The first automatic overclock came in at 4700Mhz - a 42% increase over stock! I saved this profile. ASUS allows you to assign clocking profiles to software programs. I assigned the overclock profile to JRiver and then ran the benchmark. I was watching CPU-Z during the benchmark and the CPU would drop back to idle speed several times. I think the benchmark could be higher if I set the clocking in the bios and turned off any C-States. I'll have to play with it some more.

Nice!
Adjusting C-States, disabling turbo (forcing it to stay at the max OC speed) and so on boosts the JRmark score from about 4500 to 5000 on my 2500K.
 
I generally do allow it to dynamically adjust the clock and so on though, unless I'm trying to troubleshoot a problem.
 
I wonder what your score would be like when set to the same clockspeed as the 4790K.
Theoretically with them both being Haswell chips, and with Haswell-E using DDR4, it should be the same or higher for single-threaded tasks?
 
I've seen it suggested that a 4790K is the better chip if your workload includes single-threaded tasks, but I wonder if that's just because they can clock higher due to having fewer cores. If you get the same performance clock-for-clock, that pushes me back in the direction of Haswell-E.
 
 
I'm curious about what the CPU usage is like with multichannel SACD playback, any chance you could post results from this track with Memory Playback enabled/disabled?
Especially with one core per channel, I'd be surprised if there was any difficulty with playback though. (like I'm seeing on my current system)

[mojave]

The 5-Way Optimization is starting to scare me. It is now up to 7250Mhz.

[Hendrik]

I used that tool on my 4770K once, and it set it to  4.7 GHz (IIRC),  and claimed everything was stable. Too bad the system sporadically crashed afterwards anyway, so much for stable!

[mojave]

Still going at 10000 Mhz. They need more digits for the readout. Max temp has now reached 80C.

[Hendrik]

I think something is wrong with your system, its impossible that its actually running at full load on that.
Also, you attached the same screenshot on both images, and it claims 6625 MHz in both  ?

[mojave]

I have attached the correct screenshot. I stopped it at about 11000 Mhz. It then rebooted and started increasing the BCLK up from 125. At 127 the computer crashed and wouldn't reboot. I reset the CMOS and rebooted. I tried overclocking again with a limit of 5000. It went to 5000 and rebooted for the next stage of overclocking, but then I got a Windows 8 error. I'm fixing things now.

[6233638]

It seems like something is probably not reporting values correctly if you're getting clockspeeds that high.
Is it advisable to raise the base clock on these CPUs? I know it was fine on older LGA2011 platforms, but haven't paid much attention to these yet. I thought it would all just be via multipliers like 115X.

[Hendrik]

You can select a clock "strap" of either 100, 125 or 167  MHz, which the multiplier works against, that part is perfectly save. From what I've heard, X99 uses 125 MHz strap automatically if you use "OC" RAM, so it can reach the clocks required there.
The straps are designed to have a clear divisor to get clean 100 MHz back for the PCIe and DMI bus, which is why they work, since you're not overclocking the bus when changing them.

You can also try to muck with the clock directly, ie. raise it slightly for the last 1% of speed, but that is usually not possible beyond 3% clock speed increase, and may become unstable quick, generally not advised.

[mojave]

This is again at 4700 Mhz on the CPU and 3104 Mhz on the memory with C-States turned off. The CPU stayed at 4700 Mhz the whole time.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 3.121 seconds
    Single-threaded floating point math... 1.917 seconds
    Multi-threaded integer math... 0.612 seconds
    Multi-threaded mixed math... 0.434 seconds
Score: 3123

Running 'Image' benchmark...
    Image creation / destruction... 0.231 seconds
    Flood filling... 0.258 seconds
    Direct copying... 0.236 seconds
    Small renders... 0.866 seconds
    Bilinear rendering... 0.614 seconds
    Bicubic rendering... 0.347 seconds
Score: 8622

Running 'Database' benchmark...
    Create database... 0.145 seconds
    Populate database... 0.803 seconds
    Save database... 0.210 seconds
    Reload database... 0.042 seconds
    Search database... 0.642 seconds
    Sort database... 0.565 seconds
    Group database... 0.560 seconds
Score: 7249

JRMark (version 20.0.14): 6331

[6233638]

For what it's worth, I must have been mistaken before, or the results have changed since, because I did some testing and tried pushing my overclock further today, and was not able to hit ~5K just by disabling all the power management features. I probably misremembered things.
It runs a little bit faster, but it's not a significant change and it looks like Haswell does an even better job than Sandy Bridge in that regard, showing an even smaller difference between having the power-saving options enabled and disabled.
 
I had to bump the overclock on my 2500K up to 4.7GHz to reach 4948 on JRmark.
However, to go beyond 4.5GHz requires setting all the overclocking-related options to "extreme" and putting more volts into the CPU than I'd like. The BIOS setting is in the "safe" range (1.35V) but LLC pushes that to ~1.375 under load. Apparently these chips have no problem up to 1.45V, but I'd rather stay within what my BIOS classifies as the safe range.
Unlike 4.7GHz, the overclock from the stock 3.3GHz to 4.5GHz is stable on this chip with all the power-saving options enabled, rather than requiring the use of the overclocking ones.
 
I could probably have broken 5K if I ran the JRmark test at 4.8GHz but it wasn't stable in the more demanding tests I had run without pushing the voltage beyond the safe levels, and even if the benchmark runs, I don't consider results on an unstable system to be worthwhile.
 
 
To be honest, I'm thinking that I may just stick with my system for a bit longer now, and maybe I will just keep it at 4.7GHz since most people consider anything below 1.45V safe.
I'm not sure that it's worth all this expense for ~22% increase in performance.
I was hoping for something closer to 50% (or 100%) considering that it's a jump from 4 cores to 6 (or 8) and there are two generations of IPC improvements on top of that.
 
Considering that both our results were at 4.7GHz, and it's 4 cores vs 6, it indicates that the benefits of moving from 4 to 6 cores are potentially a lot smaller than I expected - at least in many applications.
Even if IPC had remained the same as Sandy Bridge, linear scaling should have meant a score of 7422. Perhaps that's something Hendrik will be able to optimize once his system arrives.

[Hendrik]

Not every workload is magically 50% faster if you give it two more cores. :p
Very often you just have very small datasets where adding multi-threading is more overhead than worth the trouble.

[6233638]

Not every workload is magically 50% faster if you give it two more cores. :p
Very often you just have very small datasets where adding multi-threading is more overhead than worth the trouble.

Yes, I know. But seeing this and other people's "real-world" results, rather than specific tasks like rendering, compiling code etc. seem to show similar results.
 
Actually, I would be very interested to see what the results would be like if you set Media Center to only use 4 cores and ran the benchmark.
 
If I could get the same or better performance than I have now and have two cores spare for other tasks, that would be nice - that's actually one of the things which appeals to me about moving to 6 cores, since SACD playback is such a hog in MC19/20.

[mojave]

Here is the benchmark with 4 cores:

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 3.123 seconds
    Single-threaded floating point math... 1.918 seconds
    Multi-threaded integer math... 0.851 seconds
    Multi-threaded mixed math... 0.627 seconds
Score: 2914

Running 'Image' benchmark...
    Image creation / destruction... 0.222 seconds
    Flood filling... 0.256 seconds
    Direct copying... 0.239 seconds
    Small renders... 0.867 seconds
    Bilinear rendering... 0.737 seconds
    Bicubic rendering... 0.378 seconds
Score: 8151

Running 'Database' benchmark...
    Create database... 0.145 seconds
    Populate database... 0.807 seconds
    Save database... 0.211 seconds
    Reload database... 0.037 seconds
    Search database... 0.643 seconds
    Sort database... 0.565 seconds
    Group database... 0.567 seconds
Score: 7228

JRMark (version 20.0.14): 6098

[Hilton]

Intel i7 4790k @ 4487Mhz on Air (Johnny_Friendly)
Running 'Math' benchmark...
    Single-threaded integer math... 3.289 seconds
    Single-threaded floating point math... 2.015 seconds
    Multi-threaded integer math... 0.979 seconds
    Multi-threaded mixed math... 0.619 seconds
Score: 2753

Intel i7-5820K @ 4700Mhz (Mojave)
Running 'Math' benchmark...
    Single-threaded integer math... 3.122 seconds
    Single-threaded floating point math... 1.918 seconds
    Multi-threaded integer math... 0.595 seconds
    Multi-threaded mixed math... 0.398 seconds
Score: 3150

i7 4960X @ 4.7Ghz with X79 Asus Rampage IV Black Edition - Memory at a leisurely 1866mhz - GTX580 @ 1000/2200 1920x1080
=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 2.879 seconds
    Single-threaded floating point math... 1.917 seconds
    Multi-threaded integer math... 0.615 seconds
    Multi-threaded mixed math... 0.422 seconds
Score: 3258

Running 'Image' benchmark...
    Image creation / destruction... 0.323 seconds
    Flood filling... 0.380 seconds
    Direct copying... 0.245 seconds
    Small renders... 0.957 seconds
    Bilinear rendering... 0.654 seconds
    Bicubic rendering... 0.385 seconds
Score: 7475

Running 'Database' benchmark...
    Create database... 0.172 seconds
    Populate database... 0.951 seconds
    Save database... 0.218 seconds
    Reload database... 0.035 seconds
    Search database... 0.695 seconds
    Sort database... 0.902 seconds
    Group database... 0.613 seconds
Score: 5995

JRMark (version 20.0.14): 5576

i7 4960X @ 4.8Ghz with X79 Asus Rampage IV Black Edition - Memory at a leisurely 1866mhz - GTX580 @ 1000/2200 1920x1080
=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 2.818 seconds
    Single-threaded floating point math... 1.877 seconds
    Multi-threaded integer math... 0.629 seconds
    Multi-threaded mixed math... 0.426 seconds
Score: 3305

Running 'Image' benchmark...
    Image creation / destruction... 0.338 seconds
    Flood filling... 0.375 seconds
    Direct copying... 0.246 seconds
    Small renders... 0.932 seconds
    Bilinear rendering... 0.644 seconds
    Bicubic rendering... 0.376 seconds
Score: 7558

Running 'Database' benchmark...
    Create database... 0.165 seconds
    Populate database... 0.905 seconds
    Save database... 0.213 seconds
    Reload database... 0.035 seconds
    Search database... 0.676 seconds
    Sort database... 0.883 seconds
    Group database... 0.592 seconds
Score: 6198

JRMark (version 20.0.14): 5687

The 5820K is certainly good bang for your buck.
I wonder why you get a much better image benchmark?

edit: Looks like memory speed has an influence on the image score.
Changing my memory from 1866 9-10-9-27 to 2133 9-11-11-31  gained a few CPU and image points.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 2.820 seconds
    Single-threaded floating point math... 1.883 seconds
    Multi-threaded integer math... 0.544 seconds
    Multi-threaded mixed math... 0.408 seconds
Score: 3360

Running 'Image' benchmark...
    Image creation / destruction... 0.326 seconds
    Flood filling... 0.374 seconds
    Direct copying... 0.249 seconds
    Small renders... 0.929 seconds
    Bilinear rendering... 0.642 seconds
    Bicubic rendering... 0.376 seconds
Score: 7595

Running 'Database' benchmark...
    Create database... 0.173 seconds
    Populate database... 0.921 seconds
    Save database... 0.211 seconds
    Reload database... 0.035 seconds
    Search database... 0.673 seconds
    Sort database... 0.884 seconds
    Group database... 0.569 seconds
Score: 6204

JRMark (version 20.0.14): 5720

And a little bit more with memory @ 2400 10-12-12-31

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 2.817 seconds
    Single-threaded floating point math... 1.878 seconds
    Multi-threaded integer math... 0.576 seconds
    Multi-threaded mixed math... 0.393 seconds
Score: 3354

Running 'Image' benchmark...
    Image creation / destruction... 0.297 seconds
    Flood filling... 0.372 seconds
    Direct copying... 0.209 seconds
    Small renders... 0.933 seconds
    Bilinear rendering... 0.642 seconds
    Bicubic rendering... 0.376 seconds
Score: 7779

Running 'Database' benchmark...
    Create database... 0.167 seconds
    Populate database... 0.912 seconds
    Save database... 0.210 seconds
    Reload database... 0.034 seconds
    Search database... 0.675 seconds
    Sort database... 0.880 seconds
    Group database... 0.607 seconds
Score: 6170

JRMark (version 20.0.14): 5768

And slower with Memory @ 1600 7-7-7-20 - proves memory bandwidth is more important than latency for image benchmark and CPU benchmark.

=== Running Benchmarks (please do not interrupt) ===

Running 'Math' benchmark...
    Single-threaded integer math... 2.821 seconds
    Single-threaded floating point math... 1.882 seconds
    Multi-threaded integer math... 0.653 seconds
    Multi-threaded mixed math... 0.396 seconds
Score: 3303

Running 'Image' benchmark...
    Image creation / destruction... 0.383 seconds
    Flood filling... 0.379 seconds
    Direct copying... 0.382 seconds
    Small renders... 0.941 seconds
    Bilinear rendering... 0.645 seconds
    Bicubic rendering... 0.383 seconds
Score: 7069

Running 'Database' benchmark...
    Create database... 0.162 seconds
    Populate database... 0.919 seconds
    Save database... 0.212 seconds
    Reload database... 0.037 seconds
    Search database... 0.682 seconds
    Sort database... 0.882 seconds
    Group database... 0.578 seconds
Score: 6191

JRMark (version 20.0.14): 5521

[6233638]

Here is the benchmark with 4 cores:
[…]
JRMark (version 20.0.14): 6098

Thanks for the test - so it is as I suspected and the test is barely scaling beyond 4 cores at all then.
 
Testing DSD playback would be useful to see whether it can take advantage of 6 cores, but it looks like you would be able to set MC to 4 cores and have 2 free for other applications regardless of what the result is.
 
FYI, you can limit an application using the "Set Affinity" command in the task manager (Details tab) rather than having to disable cores in the UEFI - that's how I was able to confirm that DSD playback is smooth when MC is limited to a single CPU core on my 2500K, and constantly interrupts the system when allowed access to all four.

[Hendrik]

You can configure the number of threads used for SACD DST decompression in an upcoming MC build, and I also tweaked the default to be smarter about the number of threads it uses.

[mojave]

6233638, I'll be downloading and testing your multi-channel DSD file. I just haven't gotten around to it yet.

[6233638]

6233638, I'll be downloading and testing your multi-channel DSD file. I just haven't gotten around to it yet.

No problem. Might be worth waiting to see what the new changes are like.

[mojave]

I did a little modification to the case. The GD07 case is designed so that three 120mm fans can be installed in the front pulling in air from the bottom. However, due to the drive cage you can only use two of those locations. I decided I wanted the Corsair H75 water cooler completely inside the case. I removed the drive cage and moved the radiator and fans to the front left. I was able to use both radiator fans in a push/pull configuration.

[mojave]

I don't really need the drive rack spaces for 3 1/2" and 2 1/2" drives so I removed them. The 2 1/2" section was screwed on and was easy to remove. The 3 1/2" section was riveted. I drilled out the rivets and then used my right angle grinder to cut the part off that I didn't want. I left a little as a support for the Blu-ray bay.

[mojave]

Now the drive cage fits back in the case! There is also good airflow over the motherboard. I will have two 120mm fans on the right side of the case.

Yesterday I ordered 6 Noctua fans:
Four NF-S12A PWM for the front and side of the case
Two NF-F12 PWM for the radiator

[Hendrik]

The S12A are my goto 120mm case fans as well, plus their 140mm cousins when required. Mighty nice fans indeed, some people are just annoyed by the color scheme, but i actually kinda like it, its something unique.

[mojave]

The Noctua fans just arrived so I can install them this weekend.

My case says it supports a GPU length of 13.6" long and 5.25" high. This morning I purchased the Gigabyte GTX 970 G1 Gaming which is 12.28" long and 5.08" high.

I just looked at the picture I posted earlier of the Corsair H75 cooler mounted in the front of the case. It looks like the radiator intake and hoses will interfere with the card.  I may need to go with an air cooler instead of water cooling. Hmm, I may have an idea that will still make it all work.

[6233638]

Isn't there space for a 240mm rad on the right?

[mojave]

Isn't there space for a 240mm rad on the right?

The two 120mm fan mounts aren't the right distance apart for a 240mm radiator. I did have my 120mm radiator mounted on the inside right location in your picture with a fan on the outside. It was plenty cool with idle temps at 30C. Here is how the temps changed during overclocking at 100% CPU with only 1 push fan on the radiator.

4100 - 53
4200 - 59
4300 - 62
4400 - 67
4500 - 68
4600 - 71
4700 - 72
4800 - 75

Since the HTPC will be in an enclosed rack, I couldn't really have a fan hanging off the outside of the case. Also, I wanted to pull air from the front or below the HTPC into the radiator.

Here is a picture of the radiator in the first position. The motherboard heatsinks/shrouds are in the way of mounting an inside 120mm fan. I have seen some use a 90mm fan screwed in to the top right of the radiator.

[mojave]

With the Gigabyte GTX 970, the case will have 10 fans including the PSU fan! My goal is for it to be inaudible from a couple feet away.

[Hendrik]

10 seems excessive, even if 3 of those are on the GPU.
Like, 5 case fans now?

My new build will have 2x120 in the front and 1x140 in the rear, and that needs to do it!

[6233638]

Generally, if there's a spot to mount a fan, I use it now.
The more fans you are using, the quieter the system will be for the same amount of air intake, and the fewer spots of "dead air" there are likely to be in the case.
 
This is especially true with the PWM Noctua fans, since they can go all the way down to 25% speed. (typically 300 RPM unless you're using the new industrial fans)

[mojave]

With the case I'm using, one can also reverse the power supply so that its intake fan is on the inside. Since the PSU is at the back left, this would pull hot air from the GPU through the power supply and exhaust it out the back. On my current(former?) HTPC, the PSU was on the back right of the case and was also used to exhaust air out the case. This is one reason I didn't have any exhaust fans for so long. The other reason was that most 60mm fans are too loud. The Noctua ones finally were silent and have sold me on Noctua fans.

You can see the PSU here in relation to the GPU:

[Hendrik]

I always reverse my power supply if possible, in a standing tower it would pull air from under the case otherwise, and me having a carpet i would imagine it would collect quite a bit of dust from that area. Might as well pull air out of the case instead.

[6233638]

I always reverse my power supply if possible, in a standing tower it would pull air from under the case otherwise, and me having a carpet i would imagine it would collect quite a bit of dust from that area. Might as well pull air out of the case instead.

Yeah, while I understand the reasoning behind pulling cool air from outside the case over the PSU rather than "hot" air from inside the case, it really doesn't matter when power supplies and PC components are so much more efficient now.
I'm not actually sure that the fan has ever turned on with my AX850 since I ended up not using SLI.
 
If the fan does come on, it might as well just help exhaust air that's inside the case and avoid pulling dust into the PSU. If nothing else, it's one less filter to clean.
 

The other reason was that most 60mm fans are too loud. The Noctua ones finally were silent and have sold me on Noctua fans.

Yes, I don't think most people fully appreciate how much better Noctua fans are than anything else until they actually use them.
Even with fans rated at or similar to the noise levels of Noctua fans, Noctuas are focused on removing distracting noises from the fans rather than SPL alone.

[bulldogger]

Yeah, while I understand the reasoning behind pulling cool air from outside the case over the PSU rather than "hot" air from inside the case, it really doesn't matter when power supplies and PC components are so much more efficient now.
I'm not actually sure that the fan has ever turned on with my AX850 since I ended up not using SLI.
 
If the fan does come on, it might as well just help exhaust air that's inside the case and avoid pulling dust into the PSU. If nothing else, it's one less filter to clean.
 Yes, I don't think most people fully appreciate how much better Noctua fans are than anything else until they actually use them.
Even with fans rated at or similar to the noise levels of Noctua fans, Noctuas are focused on removing distracting noises from the fans rather than SPL alone.

I am running an Asus R9 290x Directcu II, GPU with Seasonic 1200 watt platinum supply. With madVR at very high settings, the GPU can produce a lot of heat. Takes awhile, about 30 mins when watching movies but things get  warm inside the case. I am using a passive CPU cooler which is contributing as well, Nofan CR95  and one Noiseblocker M12s1 case fan and two Noiseblocker NB-BlackSilentPro PE-P 92mm intake fans, running at 300rpm under the bottom of a Siverstone CW02 case. I7 4790k CPU temps never go over 49 degrees and get JR mark of 5700 with  with no overclocking. Ram is 2400mhz Kingston 16gb.

However, the Seasonic power supply fans can get very noticeable when temps on the GPU start to rise. My next move is to mount a 280mm rad on the outside of the case to cool the GPU and see what happens. With the case top off, the fan on the Seasonic never comes on and the whole rig is silent. The noise of the Seasonic supply is very noticeable when the fan goes on high speed. It's a platinum supply but the fan comes on at high at certain temp and it is awful.