Underperforming i7 6700K ~20% after i5 upgrade

[Peter_160]

I've just upgraded my mini ITX PC from an i5-6600 to an i7-6700K alongside an R9 Nano. In Cinebench, PassMark, The Witcher 3 (Novigrad) and in AC: Unity the new CPU has been fairly substantially underperforming. With all graphical detail turned down, fps can drop to the 40s in Novigrad and the same in crowded Paris. Cinebench's Multicore score is down by about 100 points from average for 6700K at 4.0GHz, and PassMark down by 2000.

Temperatures remain around 75-80C during benchmark testing, and there is no fluctuation of clock speed nor FSB (just under 100MHz x 40.00) at full belt.

The computer is using a Shuttle 500W PSU on an ASUS H110i PLUS D3 motherboard with an R9 Nano with a SanDisk Ultra Plus SSD as a Boot drive. The CPU is cooled with a Noctua NH-L9i.

Things that I've tried thus far are:

- Reinstalling Windows 10
- Factory Resetting BIOS[
- Setting PC to High Performance Mode
- Using more fans around the CPU to see whether temperature was affecting things.

What seems to be interesting is that performance has been quite similar to that which I experienced with the i5, especially in the two games.

 

[Obakasama]

It depends on your expectations. While the i7 is a more powerful CPU, the power really comes from having more threads. Unfortunately, not all software takes advantage this. If you mostly play games on the PC, then you won't see a huge performance increase unless the game itself is coded to take advantage of all available threads (and even some professional software don't manage that apparently). For thread heavy tasks (eg. rendering, video editing) the i7 should prove its worth.

Some benchmarks suggest negligible improvements for some games:
https://tech4gamers.com/intel-core-i7-6700k-vs-i5-6600k-vs-i7-4790k-vs-i7-3770k-vs-core-i7-2600k/
http://www.gamersnexus.net/hwreviews/2035-intel-skylake-i7-6700k-gaming-benchmarks-and-review/Page-2

EDIT: for info.

 

[Samer1970]

I've just upgraded my mini ITX PC from an i5-6600 to an i7-6700K alongside an R9 Nano. In Cinebench, PassMark, The Witcher 3 (Novigrad) and in AC: Unity the new CPU has been fairly substantially underperforming. With all graphical detail turned down, fps can drop to the 40s in Novigrad and the same in crowded Paris. Cinebench's Multicore score is down by about 100 points from average for 6700K at 4.0GHz, and PassMark down by 2000.

Temperatures remain around 75-80C during benchmark testing, and there is no fluctuation of clock speed nor FSB (just under 100MHz x 40.00) at full belt.

The computer is using a Shuttle 500W PSU on an ASUS H110i PLUS D3 motherboard with an R9 Nano with a SanDisk Ultra Plus SSD as a Boot drive. The CPU is cooled with a Noctua NH-L9i.

Things that I've tried thus far are:

- Reinstalling Windows 10
- Factory Resetting BIOS[
- Setting PC to High Performance Mode
- Using more fans around the CPU to see whether temperature was affecting things.

What seems to be interesting is that performance has been quite similar to that which I experienced with the i5, especially in the two games.

try disabling HT and check those two games again ? I need to see ...

 

[Peter_160]

I will show you what effect HT has tomorrow, but what's striking me is that the average results on these benchmarks without overclock is considerably more than what I can achieve. Even on the single core in Cinebench does the average score beat what I'm getting (185 vs 165). Looking at real-time benchmarks of The Witcher 3, the CPU should be throttling around 70-80fps around Hierarch square and by the river market.

 

[scuzzycard]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

 

[Samer1970]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

wait wait What single channel ? The memory controller is inside the CPU , how come the Chipset has anything to do with it ? did I miss something ? can you explain please this is new info.

Edit : I just checked the Manual of H110 motherboard and it says it supports dual Channel memory.

 

[Supernova1138]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

wait wait What single channel ? The memory controller is inside the CPU , how come the Chipset has anything to do with it ? did I miss something ? can you explain please this is new info.

Edit : I just checked the Manual of H110 motherboard and it says it supports dual Channel memory.

One of the limitations of the H110 chipset is it only allows on DIMM per channel, and is limited to two memory slots, each on its own separate channel. The memory controller might support dual channel, but the memory slots on the board are limited to run in single channel mode. It's one of the many reasons why I usually would recommend splashing out the extra 10-20 dollars for a B150 or H170 chipset board if you really needed to save money, H110 is just too badly cut down for my tastes.

 

[Samer1970]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

wait wait What single channel ? The memory controller is inside the CPU , how come the Chipset has anything to do with it ? did I miss something ? can you explain please this is new info.

Edit : I just checked the Manual of H110 motherboard and it says it supports dual Channel memory.

One of the limitations of the H110 chipset is it only allows on DIMM per channel, and is limited to two memory slots, each on its own separate channel. The memory controller might support dual channel, but the memory slots on the board are limited to run in single channel mode. It's one of the many reasons why I usually would recommend splashing out the extra 10-20 dollars for a B150 or H170 chipset board if you really needed to save money, H110 is just too badly cut down for my tastes.

can you confirm this ?

here is the manual

http://dlcdnet.asus.com/pub/ASUS/mb/LGA1151/H110I-PLUS_D3/E10783_H110I-PLUS_D3_UM_web.pdf


says :

page 6 : Š Dual channel memory architecture

I guess one dim per channel means this motherboard cant have more than 2 dimms in dual channels .. no 4 dimms , but Dual Channels are still there , but cant have Dual Channel 4 DIMMs only 2 !!!

Meaning : The CPU has TWO CHANNELS , each Channel can take 1 or 2 DIMMS

the H110 allows 1 DIM per Channel /

other chipsets Allow 2 dimms per channel .

GOT it ?

Dual Channel is in BOTH , but the difference is in how many DIMMS you can use in Dual Channel mode.

 

[scuzzycard]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

wait wait What single channel ? The memory controller is inside the CPU , how come the Chipset has anything to do with it ? did I miss something ? can you explain please this is new info.

Edit : I just checked the Manual of H110 motherboard and it says it supports dual Channel memory.

One of the limitations of the H110 chipset is it only allows on DIMM per channel, and is limited to two memory slots, each on its own separate channel. The memory controller might support dual channel, but the memory slots on the board are limited to run in single channel mode. It's one of the many reasons why I usually would recommend splashing out the extra 10-20 dollars for a B150 or H170 chipset board if you really needed to save money, H110 is just too badly cut down for my tastes.

can you confirm this ?

here is the manual

http://dlcdnet.asus.com/pub/ASUS/mb/LGA1151/H110I-PLUS_D3/E10783_H110I-PLUS_D3_UM_web.pdf


says :

page 6 : Š Dual channel memory architecture

I guess one dim per channel means this motherboard cant have more than 2 dimms in dual channels .. no 4 dimms , but Dual Channels are still there , but cant have Dual Channel 4 DIMMs only 2 !!!

I thought 1 DIMM per channel was the same thing as single-channel. So does it have the same memory bandwidth but half the slots as H170, or is it half the bandwidth and half the slots?

 

[Supernova1138]

Yeah, it looks like it can do dual channel. In any case I think part of your problem is the fact you have a DDR3 board, and you're using DDR3 RAM. Low voltage DDR3 caps out at 1600MHz and that really is going to hurt your benchmark scores vs. most other Skylake systems that are running DDR4 RAM up to 3200MHz or so. Even with DDR4's higher latencies, the DDR3L is going to have trouble keeping up.

Another thing you may want to check is to make sure your CPU is staying at its maximum clockspeed while gaming and you're not encountering any throttling, either from the CPU overheating, or the motherboard components overheating. VRM throttling isn't usually an issue on Intel's CPUs, but you put the most power hungry CPU on the mainstream socket onto one of the cheapest motherboards, so it might be a possibility.

 

[Karadjgne]

4 dimm boards run AABB slots. 2 dimm boards run AB. Its still dual channel memory, just using slots 1-2 instead of 1-3 or 2-4. So discounting the H110 boards for that reason is bunk. There are other reasons why the B150 or H170 are better off but mostly its in the gimmicks like extra sata, usb, fan headers, better lan and audio, heatsinking etc.

 

[Samer1970]

The i7 is not going to significantly outperform the i5 in games - that's not it's strength. If that's the only reason you traded up, then you wasted your money. There's one other problem - you are using a 6700K in an H110 motherboard. The H110's single-channel memory is hurting it in the synthetic tests, and it's unable to be overclocked because of the chipset. It's never a good idea to hamper a top-of-the-line CPU with a bargain-basement motherboard.

wait wait What single channel ? The memory controller is inside the CPU , how come the Chipset has anything to do with it ? did I miss something ? can you explain please this is new info.

Edit : I just checked the Manual of H110 motherboard and it says it supports dual Channel memory.

One of the limitations of the H110 chipset is it only allows on DIMM per channel, and is limited to two memory slots, each on its own separate channel. The memory controller might support dual channel, but the memory slots on the board are limited to run in single channel mode. It's one of the many reasons why I usually would recommend splashing out the extra 10-20 dollars for a B150 or H170 chipset board if you really needed to save money, H110 is just too badly cut down for my tastes.

can you confirm this ?

here is the manual

http://dlcdnet.asus.com/pub/ASUS/mb/LGA1151/H110I-PLUS_D3/E10783_H110I-PLUS_D3_UM_web.pdf


says :

page 6 : Š Dual channel memory architecture

I guess one dim per channel means this motherboard cant have more than 2 dimms in dual channels .. no 4 dimms , but Dual Channels are still there , but cant have Dual Channel 4 DIMMs only 2 !!!

I thought 1 DIMM per channel was the same thing as single-channel. So does it have the same memory bandwidth but half the slots as H170, or is it half the bandwidth and half the slots?

It is the same bandwidth , but instead of 4 DIMMS in dual channels , you are allowed only 2 ...

 

[Samer1970]

Yeah, it looks like it can do dual channel. In any case I think part of your problem is the fact you have a DDR3 board, and you're using DDR3 RAM. Low voltage DDR3 caps out at 1600MHz and that really is going to hurt your benchmark scores vs. most other Skylake systems that are running DDR4 RAM up to 3200MHz or so. Even with DDR4's higher latencies, the DDR3L is going to have trouble keeping up.

Another thing you may want to check is to make sure your CPU is staying at its maximum clockspeed while gaming and you're not encountering any throttling, either from the CPU overheating, or the motherboard components overheating. VRM throttling isn't usually an issue on Intel's CPUs, but you put the most power hungry CPU on the mainstream socket onto one of the cheapest motherboards, so it might be a possibility.

He was using the same memory on his i5 6600 .. The memory speed is not the reason .

I think if he disables the HT will see better results .. This is game coding related issue and not hardware issue ..

 

[Rogue Leader]

Guys,

His comparison numbers are to averages he sees for the 6700k. He did say he was expecting more performance over the i5-6600 but he did not provide numbers related to that.

Peter_160, do not disable Hyperthreading, you should not need to do that. Now as Obakasama posted, gaming wise there is little difference between the two processors, so unless its running SLOWER, everything is fine. I would like to turn back to your memory however, being that you are running DDR3 what frequency is it at? Most likely you are comparing to folks who have memory at least at 2133mhz if not faster. In a benchmark like Cinebench this can make a significant difference.

Note Witcher 3 at 48fps minimum with the R9 Nano here:

http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-4.html

If you have slow memory, or you have Hairworks on or too much AA on, I can see dips to 40.

I'm pretty sure everything is working as it should, your system is handicapped by your DDR3 memory.

 

[RedJaron]

Samer and Kar are correct, the H110 is a dual-channel board. That it's limited to only two DIMM slots doesn't matter on an ITX board since even Z170 ITX only has two DIMM slots.

Witcher 3 generally doesn't see any performance improvement above an i5 ( see here ), and even an 8C/16T 5960X won't boost the performance. AC: Unity is also notorious for being a very poorly optimized game and even an extremely powerful system won't make a huge difference in it.

Disabling HT will lower CPU temperatures, so if you're hitting thermal throttling, it might help keep the clock up. Those four extra threads won't make a big difference in your games. However, if heat is your problem, looking at your case and airflow would be a better place to start.

What resolution are you playing at?

 

[scuzzycard]

Guys,

His comparison numbers are to averages he sees for the 6700k. He did say he was expecting more performance over the i5-6600 but he did not provide numbers related to that.

Peter_160, do not disable Hyperthreading, you should not need to do that. Now as Obakasama posted, gaming wise there is little difference between the two processors, so unless its running SLOWER, everything is fine. I would like to turn back to your memory however, being that you are running DDR3 what frequency is it at? Most likely you are comparing to folks who have memory at least at 2133mhz if not faster. In a benchmark like Cinebench this can make a significant difference.

I'm pretty sure everything is working as it should, your system is handicapped by your DDR3 memory.

I agree, it has to be the memory - not for the reason I thought (unless he's running one DIMM), but because of the DDR3. And to disable hyperthreading would be to turn his i7 back into the i5 he just upgraded from.

 

[James Mason]

I was gonna say, it could be his PSU, an R9 Nano and an I7-6700k together can use up a decent amount of power, and he has some POS "here's a PSU with your case" PSU.

It might have been fine for his old i5, but not with an unlocked i7 in there.

Even if it isn't his PSU, it will likely be the problem later on anyways.

 

[Peter_160]

Guys,

Note Witcher 3 at 48fps minimum with the R9 Nano here:

http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-4.html

If you have slow memory, or you have Hairworks on or too much AA on, I can see dips to 40.

I'm pretty sure everything is working as it should, your system is handicapped by your DDR3 memory.

That's an average across the whole game at full settings - I'm running it Ultra without hairworks and background
characters and foliage down to high. It runs around 50-60fps at 1080p by the river and in Hierarch square dips into the 40s regardless of the graphical settings. If it were the memory, then this https://www.youtube.com/watch?v=Rutk9ErhKG4 around 2:00 would show that DDR3 for the 4790K would keep it at framerates similar to what I'm seeing, but it keeps up around 75-80fps.

I'm sorry if this seems like I'm being defensive about the machine, but a bit of research seems to show that an i7 and an R9 Nano at the right settings at 1080p should be churning out 60-80fps in the centre of Novigrad. This isn't happening right now, and it doesn't seem to be either the memory (if a 4790K can do it), the processing power of the graphics card (I've tried it at low settings), nor the temperature - I've done tests with low fan speeds and there has been no throttling.

Following from that, I've noticed an interesting thing about the R9: when the CPU seems to the one throttling, the GPU will not settle at a steady load, like 70-80% for example, but will bounce from 0-100% when it is seemingly required.

The only thing I can see left is the PSU, which I thought was a decent 500W supply I bought from Amazon that cost me £100. Is there any way to test a power supply in the rig?

 

[Samer1970]

Are you using Shuttle XPC case ?

 

[scuzzycard]

Guys,

Note Witcher 3 at 48fps minimum with the R9 Nano here:

http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-4.html

If you have slow memory, or you have Hairworks on or too much AA on, I can see dips to 40.

I'm pretty sure everything is working as it should, your system is handicapped by your DDR3 memory.

That's an average across the whole game at full settings - I'm running it Ultra without hairworks and background
characters and foliage down to high. It runs around 50-60fps at 1080p by the river and in Hierarch square dips into the 40s regardless of the graphical settings. If it were the memory, then this https://www.youtube.com/watch?v=Rutk9ErhKG4 around 2:00 would show that DDR3 for the 4790K would keep it at framerates similar to what I'm seeing, but it keeps up around 75-80fps.

I'm sorry if this seems like I'm being defensive about the machine, but a bit of research seems to show that an i7 and an R9 Nano at the right settings at 1080p should be churning out 60-80fps in the centre of Novigrad. This isn't happening right now, and it doesn't seem to be either the memory (if a 4790K can do it), the processing power of the graphics card (I've tried it at low settings), nor the temperature - I've done tests with low fan speeds and there has been no throttling.

Following from that, I've noticed an interesting thing about the R9: when the CPU seems to the one throttling, the GPU will not settle at a steady load, like 70-80% for example, but will bounce from 0-100% when it is seemingly required.

The only thing I can see left is the PSU, which I thought was a decent 500W supply I bought from Amazon that cost me £100. Is there any way to test a power supply in the rig?

Was your in-game performance the same before the upgrade, or did it get worse? If it's the same, it could be driver related. Even if you have the latest drivers, AMD and NVidia both screw up sometimes, requiring a driver roll-back for a certain game to run at its best. Again, assuming you had the same in-game performance before, I would run a google search for your exact driver version, video card, and the game in question. That way you can see if other people experienced slow-downs with whatever driver version you're running, even if it's the latest.

 

[James Mason]

That's an average across the whole game at full settings - I'm running it Ultra without hairworks and background
characters and foliage down to high. It runs around 50-60fps at 1080p by the river and in Hierarch square dips into the 40s regardless of the graphical settings. If it were the memory, then this https://www.youtube.com/watch?v=Rutk9ErhKG4 around 2:00 would show that DDR3 for the 4790K would keep it at framerates similar to what I'm seeing, but it keeps up around 75-80fps.

I'm sorry if this seems like I'm being defensive about the machine, but a bit of research seems to show that an i7 and an R9 Nano at the right settings at 1080p should be churning out 60-80fps in the centre of Novigrad. This isn't happening right now, and it doesn't seem to be either the memory (if a 4790K can do it), the processing power of the graphics card (I've tried it at low settings), nor the temperature - I've done tests with low fan speeds and there has been no throttling.

Following from that, I've noticed an interesting thing about the R9: when the CPU seems to the one throttling, the GPU will not settle at a steady load, like 70-80% for example, but will bounce from 0-100% when it is seemingly required.

The only thing I can see left is the PSU, which I thought was a decent 500W supply I bought from Amazon that cost me £100. Is there any way to test a power supply in the rig?

For quality? not without spending 10k on testing equipment.
There are ways to test whether it's alive or dead, but that's not the issue.

GPU usage won't be steady because the game doesn't give a steady demand.

 

[Rogue Leader]

Thanks for the additional information, based on the original info you gave the memory made sense. Now I am of course not so sure.

That PSU is definitely quite weak, if you are getting CPU throttling without thermal issues then it definitely could be the culprit. a GOOD 500w psu will just about do your system.

That said depending on your case you may not be able to fit much else. Maybe an SFX PSU like Corsair or Silverstone offers. Go for a 650w one.

 

[Samer1970]

Shuttle power supplies in XPC are good , they are 1u server grade , how much is the 12v output on it ?

 

[Rogue Leader]

Shuttle power supplies in XPS are good , they are 1u server grade , how much is the 12v output on it ?

Its a bronze rated 500w PSU built to a small form factor Shuttle PC63J. They are not 1u server grade servers use much higher efficiency PSUs.

Corsair or Silverstones SFX PSU's would be far better in this situation.

 

[Samer1970]

Shuttle power supplies in XPS are good , they are 1u server grade , how much is the 12v output on it ?

Its a bronze rated 500w PSU built to a small form factor Shuttle PC63J. They are not 1u server grade servers use much higher efficiency PSUs.

Corsair or Silverstones SFX PSU's would be far better in this situation.

not in old times , If you remember in old times were 80 plus Only , later they became more efficient ..

I have good experience with Shuttle power supplies in XPC , 10 years working full load . they are good .

how much is the 12v ?