Contribute
Register

[Success] GIGABYTE Z370 Gaming 7 + Intel Core i7-8700K + RX 580 + (2x) Dell P2715Q 4k @ 60Hz

This is great info. Ok, I downloaded BruceX and ran the benchmarks in Final Cut Pro. I definitely have slower numbers than your results:

Pro Res 422 (default) - 16.09 secs
H264 - 19.10 secs

Also, as a test, I rebooted into the BIOS and went into the BIOS Advanced Memory Settings and changed Extreme Memory Profile (X.M.P.) off, and geekbench numbers obviously went down as well.

With X.M.P. Geekbench is 25,125
Without X.M.P. Geekbench is 23,797

(These are both with the CPU overclock)

I just wanted to see the performance number hit when turning off X.M.P. and it's fairly significant.

That tells me that a speed bump from 2666MHz RAM to 3600Mhz RAM could be the exact reasons for not only the slower Geekbench numbers, but it actually is slower real-world performance based on the BruceX results as well.

I may do a 3600Mhz RAM swap out to see if that gets similar ~30,000 geekbench numbers, without changing system identifiers, etc. Even thought the aircooler I bought makes updating RAM a pain, it might be worth it for +~5000 in performance in geekbench.

P.S. @jb007, if you can post the exact brand of 3600 RAM you purchased, maybe even with an Amazon link, I'd love to check it out in case I test out a RAM swap. Thanks!

No problems @HackaShaq.

You are right, RAM speed through put speed makes a real difference as us both have demonstrated.

RAM
My RAM is 2 x 16GB Kits (4 x 8GB = 32GB): DDR4 3600MHz G Skill Intl F4-3600C16D-8GVK here at AMAZON. I purchased it locally here in Oz.

Note: This is Cas Latency 16 memory, G.Skill has others that are slower Cas Latency > 16 but with the same 3600MHz speed just to confuse everyone. Their is other models in the range with RGB etc which should fit. I took these as there was a shortage (high price) of RAM in Australia when I bought these. Also the Amazon link might take you to Australian pricing, as Amazon redirects all of here in Oz to local pricing as we have to pay GST (Goods & Service Tax @10%) on all our internet purchases now. Just search on the part number F4-3600C16D-8GVK if this is the case.

This motherboard RAM support list can be found here at Gigabyte. Note any RAM over the INTEL spec of 2666MHz is considered to be overclocked. You will notice as the speed of the RAM is increased, the latency rating is also increased... this will then mean that it's not a linear progression of speed as the MHz goes up, or that's my understanding. You will also notice the RAM I'm using is in the recommend list and is a lower latency part.

Also do you have the latest BIOS revision F6? Versions F4a & F5h according to Gigabyte had enhanced XMP performance.

I ran a test from a free Apple Store app - Novabench... my RAM speed is 33,813MB/sec according to it. From memory the Memtest app that you boot from a USB stick would be a better test.

Novabench - jb007.jpeg
 
Finished my nearly identical build to this today. Thank you, HachaShaq for posting your build and to all who have added their experience. This was the easiest build I've done by far. The only changes I made were to increase the 500Gb M.2 to 1Tb, add a 2Tb Micron SSD as a Final Cut Pro X scratch/work disk and omit the Firewire and WiFi cards. Can anyone address the pros and cons of overclocking this build? I'm most interested in durability but extra speed without temperature drama would be nice with editing very large (1-2 hour) videos that take a lot of processing time.

I'm no expert on OC'ing this board, but CPU temp is very important. Luckily I've fitted a AIO Corsair H110i 240mm water cooling and it keeps the CPU very cool. I thought I said in a previous post that I have not seen above 80C (176F), but last night at non-OC on my i7-8700K running the prime95 'torture test' I didn't see much over 60C (140F) and my water temp stabilised at around 35C (95F) after around 20 minutes of all cores running 100%. I might try OCing later and posting my results etc.
But as @jimmakos01 said, 4.9GHz is very achievable for most and stable, but test well over a reasonable period of time using something like prime95 if you can.
 
In bios settings for the overclock, avxoffset and uncore ratio is at auto? If no, could you tick the auto settings and run the geekbench again? You will not have any problem to boot.
Hey, jimmakos01.

avxoffset was already set to Auto, and uncore ratio is at 40 but there's no way to change that individually (option to change is likely dependent on another setting.)

Here's what my current CPU settings look like with thew overclock:

IMG_3809_preview.jpeg

I'll be happy to test any other settings, but performance seems to definitely be tied to my memory speed.
 

Attachments

  • image.png
    image.png
    154.5 KB · Views: 131
No problems @HackaShaq.

You are right, RAM speed through put speed makes a real difference as us both have demonstrated.

RAM
My RAM is 2 x 16GB Kits (4 x 8GB = 32GB): DDR4 3600MHz G Skill Intl F4-3600C16D-8GVK here at AMAZON. I purchased it locally here in Oz.

Note: This is Cas Latency 16 memory, G.Skill has others that are slower Cas Latency > 16 but with the same 3600MHz speed just to confuse everyone. Their is other models in the range with RGB etc which should fit. I took these as there was a shortage (high price) of RAM in Australia when I bought these. Also the Amazon link might take you to Australian pricing, as Amazon redirects all of here in Oz to local pricing as we have to pay GST (Goods & Service Tax @10%) on all our internet purchases now. Just search on the part number F4-3600C16D-8GVK if this is the case.

This motherboard RAM support list can be found here at Gigabyte. Note any RAM over the INTEL spec of 2666MHz is considered to be overclocked. You will notice as the speed of the RAM is increased, the latency rating is also increased... this will then mean that it's not a linear progression of speed as the MHz goes up, or that's my understanding. You will also notice the RAM I'm using is in the recommend list and is a lower latency part.

Also do you have the latest BIOS revision F6? Versions F4a & F5h according to Gigabyte had enhanced XMP performance.

I ran a test from a free Apple Store app - Novabench... my RAM speed is 33,813MB/sec according to it. From memory the Memtest app that you boot from a USB stick would be a better test.

View attachment 339956
This is all such fantastic info, @jb007, thank you!

I was able to run a Novabench score, and sure enough, I'm getting dramatically slower RAM speeds than your test.
Granted, I didn't run this from a USB stick, but this should at least give a rough baseline.

It also seems to be accurate with my CPU score since I have the 5.0 overclock, and I'm guessing you maybe have no overclock.


Screen Shot 2018-07-08 at 11.41.10 PM.png

I was also thinking about updating my BIOS as well. I'm currently running F4. I was a bit skittish about updating the BIOS only because everything is working so smoothly, but I think it's worth doing, especially because as you noted there is increased XMP DDR compatibility (and possibly performance) listed in the BIOS updates.

And thank you for your Amazon link and specific info about your memory. That makes total sense now. I found that Gaming 7 Memory chart before, and your post clarified some things. I can see what you mean that by the chart, although there are faster RAM Frequency speeds out there, only yours has the lower Cas Latency 16 rating @3600Mhz. Most others are in the 19, 18, and 17 for latency at that frequency.

My current memory is also at a Cas Latency 16 rating, but the frequency is way way down on the list at 2666Mhz, so that explains the lower performance overall in real world tasks and geekbench. Great info, and this clarified the memory settings. Thanks so much. I may do a memory swap with this exact RAM to get the max frequency and lowest latency.

I'm going to update the BIOS and re-run Geekbench and report back.
 
Cool, it will be interesting to see if you have any performance increase after BIOS update... From what I have viewed on the youTube and also in tech articles, INTEL CPU's don't respond as well to faster memory as do Ryzen.

Also have a look at this article on memory speed and latency et al by Crucial.

I also couldn't help myself, I OC to 5GHz just using the Gigabyte CPU selected to i7-8700K @ 5GHz and then ran GB 4 and got 6724 & 32534, then ran Prime95 torture test whilst also running a Blender benchmark that uses OpenCL on both of my RX 580's, it got to half way (around 4 minutes) then the system re-booted... temp didn't go over 80C as far as I could tell.

I have a Corsair 850 power supply that is on the Corsair Link USB bus, but there is no macOS app for it, but I have downloaded the SDK from Corsair, and will try to write a macOS version. I also will look into the Gigabyte RGB side of things... hopefully!
 
Also do you have the latest BIOS revision F6? Versions F4a & F5h according to Gigabyte had enhanced XMP performance.
Hey @jb007, just flashed the BIOS to the latest version (F6). I wish I could give your post multiple "likes" because your post both educated me and helped me try something new.

With my slower 2666Mhz RAM, I'm still seeing slower geekbench numbers that your machine with faster RAM. This has to be the reason I'm not seeing 30k geekbench numbers. In theory, changing the system identifier might make a slight difference as well, but using the data points that others have included where changing the system identifier didn't result in a speed bump, it makes total sense that it just has to be the RAM speed (Cas Latency 16).

However, after updating the BIOS from F4 to F6 I started over with the BIOS from scratch.

1. With the new F6 BIOS, I reverted the BIOS to the default settings.
2. I then re-entered my BIOS settings needed from my guide here which were needed for the hackintosh.

This was now a stock hackintosh BIOS with my parts (no overclock) and I ran geekbench.
I got the excepted Geekbench result of ~23,000 (no overclock)

Exactly the same as before.

3. Then I tried a new way to overclock it. Instead of following my own guide steps for the manual overclock, I simply used the dummy approach and selected Gigabyte's super easy "Processor Upgrade" as discussed here.

proc_upgrade.jpeg

I then ran geekbench again with, this time with the simple "CPU Upgrade" style overclock applied.

I got the following Geekbench number of 25,671.

geekbench-7-9-18.png

This is still a slower number than your faster RAM machine, but it was a really good data point for me. It showed that my more advanced manual settings entered in to overlock the CPU actually ran slower than the easy "CPU Upgrade" option.

In the past, my geekbench numbers were always 25,200, with a +/- of about 150. The lowest I saw with the overclock was about 25,050 and the highest was about 25,275.

I was expecting this new easy "CPU Upgrade" option to yield slightly lower Geekbench numbers, but it was the opposite. It actually gave me a slightly higher score of 25,671 and by far the highest geekbench numbers I've seen with these parts.

So I think going forward, I'll update the guide and simply tell people who want to overlock, that this easy "CPU Upgrade" option is the way to go.
 

Attachments

  • geekbench-7-9-18.png
    geekbench-7-9-18.png
    311.7 KB · Views: 84
Hey @jb007, just flashed the BIOS to the latest version (F6). I wish I could give your post multiple "likes" because your post both educated me and helped me try something new.

With my slower 2666Mhz RAM, I'm still seeing slower geekbench numbers that your machine with faster RAM. This has to be the reason I'm not seeing 30k geekbench numbers. In theory, changing the system identifier might make a slight difference as well, but using the data points that others have included where changing the system identifier didn't result in a speed bump, it makes total sense that it just has to be the RAM speed (Cas Latency 16).

However, after updating the BIOS from F4 to F6 I started over with the BIOS from scratch.

1. With the new F6 BIOS, I reverted the BIOS to the default settings.
2. I then re-entered my BIOS settings needed from my guide here which were needed for the hackintosh.

This was now a stock hackintosh BIOS with my parts (no overclock) and I ran geekbench.
I got the excepted Geekbench result of ~23,000 (no overclock)

Exactly the same as before.

3. Then I tried a new way to overclock it. Instead of following my own guide steps for the manual overclock, I simply used the dummy approach and selected Gigabyte's super easy "Processor Upgrade" as discussed here.

View attachment 339971

I then ran geekbench again with, this time with the simple "CPU Upgrade" style overclock applied.

I got the following Geekbench number of 25,671.

View attachment 339972

This is still a slower number than your faster RAM machine, but it was a really good data point for me. It showed that my more advanced manual settings entered in to overlock the CPU actually ran slower than the easy "CPU Upgrade" option.

In the past, my geekbench numbers were always 25,200, with a +/- of about 150. The lowest I saw with the overclock was about 25,050 and the highest was about 25,275.

I was expecting this new easy "CPU Upgrade" option to yield slightly lower Geekbench numbers, but it was the opposite. It actually gave me a slightly higher score of 25,671 and by far the highest geekbench numbers I've seen with these parts.

So I think going forward, I'll update the guide and simply tell people who want to overlock, that this easy "CPU Upgrade" option is the way to go.

Thanks for the feedback. I also used the 'dummy approach and selected Gigabyte's super easy "Processor Upgrade"' but it did crash out when I ran Blender and prime95 simultaneously @5GHz. I remember reading somewhere that these settings Gigabytes use tend to be a bit over the top with the voltage settings, and hence the CPU runs hotter. I might look into this a bit further down the track including de-lidding and using liquid metal for further cooling.

I personally would only mention this OC method with the proviso of the users doing proper testing for stability of the system, after all, a system that runs 5% faster and crashes in a heap is not faster at all! Also when ever I have a unplanned reboot, I always run diskutil and first aid on all drives... but I'm a bit of a cautious nelly! Also if you have a cloned drive, you could do the testing for stability on that etc.

I do have one problem since or around the time I updated to 10.13.5 is that sleep reboots after a few seconds. If I run this command in terminal wake reason it tells me that USB was the reason (can't recall exact text, but it is USB related). Can you test your system also?

Cheers.
 
One more data point for anyone who wants to squeak out the maximum performance from my exact parts list, I also just did a one other simple change in my BIOS settings to increase the performance of the memory.

In BIOS:

M.I.T > Advanced Memory Options

Memory Enhancement Settings > Enhanced Performance

IMG_3811_preview.jpeg

This resulted in my highest geekbench numbers yet: 25,991.

geekbench25991.png

Again, not nearly as great as your faster RAM, but that just helps me think these are the optimal BIOS settings, even over my original manual overlock settings, etc.

Novabench also had similar results where the CPU is fast, and @jb007 I'm seeing exact matching numbers to your GPU and Disk Score, along with a CPU that reflects the overclock. The only thing lacking again is that RAM Speed, which makes perfect sense.

So the next step is, do I need to upgrade my RAM to achieve 32,534 geekbench numbers similar to yours? Well, no. But am I? Of course. :D

Using the somewhat simplistic formula of (CL/(freq-in-MHZ) * 1000) = timing in ns then according to the supported RAM list of the Gaming 7 board, some my supported RAM vs. your supported RAM is the following:

My RAM - Corsair Vengeance LPX 32GB (2x16GB) DDR4 DRAM 2666MHz (CMK32GX4M2A2666C16)
Delay = 6.0ns

Your RAM - 16GB G.Skill DDR4 PC4-28800 3600MHz Ripjaws V (F4-3600C16D-16GVK)
Delay = 4.4ns

Your RAM is way faster and although more expensive than mine, it's still an optimal value.

For example, another RAM option is let's say the 32GB G.Skill DDR4 Trident Z 4000Mhz PC4-32000 CL18 (F4-4000C18Q-32GTZ) which is on the supported list, but quite a bit more expensive at $644. Well using that formula, the speed is:
Delay = 4.5ns

In other words, yours is faster, and less expensive.

There is one other on the list of supported RAM from the Gaming 7 that jumps out at me:

G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200
F4-3000C14Q2-128GVKD (14-14-14-34)

It's "only" 3200 speed, but the CL 14 makes the delay 4.37ns. That might actually be the best performer on the list, but it looks like that's only available in a 128GB kit. Ah well.

There are other supported contenders based on that formula.
 

Attachments

  • novabench2.png
    novabench2.png
    271.1 KB · Views: 108
Last edited:
Thanks for the feedback. I also used the 'dummy approach and selected Gigabyte's super easy "Processor Upgrade"' but it did crash out when I ran Blender and prime95 simultaneously @5GHz. I remember reading somewhere that these settings Gigabytes use tend to be a bit over the top with the voltage settings, and hence the CPU runs hotter. I might look into this a bit further down the track including de-lidding and using liquid metal for further cooling.

I personally would only mention this OC method with the proviso of the users doing proper testing for stability of the system, after all, a system that runs 5% faster and crashes in a heap is not faster at all! Also when ever I have a unplanned reboot, I always run diskutil and first aid on all drives... but I'm a bit of a cautious nelly! Also if you have a cloned drive, you could do the testing for stability on that etc.

I do have one problem since or around the time I updated to 10.13.5 is that sleep reboots after a few seconds. If I run this command in terminal wake reason it tells me that USB was the reason (can't recall exact text, but it is USB related). Can you test your system also?

Cheers.
Good info, and you're exactly right about stability. If the machine isn't stable, then it's pointless. But what I like about this "dummy" approach is that it's so easy to turn off and/or go back to "normal", or dial down the CPU overclock without having to remember all the manual settings from my original overlock step-by-step.

I just did a few sleep/wakes, and my system sleeps and it properly stays asleep. So I'm not experiencing any sleep/wake issues at all. Then again, I just installed the new BIOS today, so I'm going to keep a close eye on things. But I haven't had any sleep issues (or any OS issues at all really) on 10.13.4 or 10.13.5. Hope that info helps.
 
One more data point for anyone who wants to squeak out the maximum performance from my exact parts list, I also just did a one other simple change in my BIOS settings to increase the performance of the memory.

In BIOS:

M.I.T > Advanced Memory Options

Memory Enhancement Settings > Enhanced Performance

View attachment 339987

This resulted in my highest geekbench numbers yet: 25,991.

Again, not nearly as great as your faster RAM, but that just helps me think these are the optimal BIOS settings, even over my original manual overlock settings, etc.

Novabench also had similar results where the CPU is fast, and @jb007 I'm seeing exact matching numbers to your GPU and Disk Score, along with a CPU that reflects the overclock. The only thing lacking again is that RAM Speed, which makes perfect sense.

So the next step is, do I need to upgrade my RAM to achieve 32,534 geekbench numbers similar to yours? Well, no. But am I? Of course. :D

Using the somewhat simplistic formula of (CL/(freq-in-MHZ) * 1000) = timing in ns then according to the supported RAM list of the Gaming 7 board, some my supported RAM vs. your supported RAM is the following:

My RAM - Corsair Vengeance LPX 32GB (2x16GB) DDR4 DRAM 2666MHz (CMK32GX4M2A2666C16)
Delay = 6.0ns

Your RAM - 16GB G.Skill DDR4 PC4-28800 3600MHz Ripjaws V (F4-3600C16D-16GVK)
Delay = 4.4ns

Your RAM is way faster and although more expensive than mine, it's still an optimal value.

For example, another RAM option is let's say the 32GB G.Skill DDR4 Trident Z 4000Mhz PC4-32000 CL18 (F4-4000C18Q-32GTZ) which is on the supported list, but quite a bit more expensive at $644. Well using that formula, the speed is:
Delay = 4.5ns

In other words, yours is faster, and less expensive.

There is one other on the list of supported RAM from the Gaming 7 that jumps out at me:

G.SKILL Ripjaws V Series 16GB (2 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200
F4-3000C14Q2-128GVKD (14-14-14-34)

It's "only" 3200 speed, but the CL 14 makes the delay 4.37ns. That might actually be the best performer on the list.
There might be other supported contenders, but that's an option as well.

I remember it took me a long time to work out what RAM to get as supplies were low at all 'best price' local stores, so it might have been more of what was available than 'smarts', if you know what I mean. Looks like I was lucky though, even though it cost a arm and leg!
 
Back
Top