Experiments in dual-channel DDR4 - G4600 (KBL) / H110 chipset

[VirtualLarry]

So, I received my second G4600 CPU, which is a Kaby Lake Pentium dual-core with HyperThreading, 3.6Ghz, no AVX/AVX2. I plan to put it into my second DeskMini rig, which has an H110 chipset.

I also received my second 2x8GB DDR4-2400 kit for it.

Looking to benchmark the effects of dual-channel RAM on the platform and CPU. While this test isn't perfect (comparing 8GB single-channel to 16GB dual-channel), I'm hoping that it will at least be relevant.

So, what is the best way to go about this? Either I can benchmark my G4600 currently installed with the 8GB SC, and then install the 16GB DC kit, and benchmark again, or I can install the second G4600 into the other deskmini, and leave one with 8GB SC, and upgrade one with 16GB DC, and compare benchmarks. The second DeskMini currently has a G4400 (3.3Ghz, dual-core no HT, no AVX/AVX2).

What benchmarks are good? I would like some browser benchmarks, as well as iGPU benchmarks. Note that I don't use Chrome, and sorry, I'm not going to install it.

DDR4-2400 SC results:
PeaceKeeper (Waterfox 50.1.0 64-bit, Win10 Pro 1607 64-bit):
5538

http://www.3dmark.com/3dm/17608024
3DMark Time Spy:
350

For some reason, it shows "Generic VGA (x1)" when I uploaded my results online, but shows "Intel(R) HD630 (tm) graphics" on System Info on the actual program.

Graphics score
306
Graphics test 1
1.99 FPS
Graphics test 2
1.76 FPS
CPU score
2 009
CPU test
6.75 FPS

CPU-Z 1.78 benchmark results:
ST: 1793
MT: 4064

See the Handbrake 1.01 benchmarking thread here:
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...r-with-handbrake-1-01-and-x265.2497224/page-3
Kaby Lake G4600 3.6Ghz dual-core with HT, no AVX/AVX2, DDR4-2400 @ 2400 (single-channel), Windows 10 Pro 64-bit 1607
encoded 1497 frames in 580.38s (2.58 fps), 4036.02 kb/s, Avg QP:26.24

G4600, DDR4-2400 DC:
PeaceKeeper (Waterfox 50.1.0 64-bit, Win10 Pro 1607 64-bit):
5628

http://www.3dmark.com/3dm/17608315
3DMark Time Spy: 393
Graphics score
343
Graphics test 1
2.23 FPS
Graphics test 2
1.98 FPS
CPU score
2 266
CPU test
7.62 FPS

CPU-Z 1.78 benchmark:
ST: 1797
MT: 4076

Handbrake 1.01 benchmark:
encoded 1497 frames in 528.50s (2.83 fps), 4036.02 kb/s, Avg QP:26.24

WinRAR 5.40 64-bit:
4657 MB/sec.

So, some improvement with DC. (I calculate 7.75% improvement in the Time Spy CPU test scores.)
(CPU-Z scores are apparently not very memory-dependent, they seem roughly within the margin of error.)

Edit: Intel VGA drivers for Kaby Lake version 21.20.16.4534 for Windows 10 Pro 1607 64-bit was installed, automatically, when I upgraded to the KBL CPU in Win10 Pro.

Edit: 11.1% faster in Handbrake 1.01 benchmark.

 

[krumme]

I got a skl i5 thinkpad t460 or something like that with 8GB 1667 in a single slot.
I tried some igpu gaming and then installed a second 8GB 1667 module.
Just assessing subjectively i would guess the benefit was about 30%.
Clearly skl was here bandwith limited on 1667MHz igpu.
But igpu is where there will be a sizable difference. Pretty cool to experience. Your 2400 is perhaps to fast out the gate to notice it. We will know. I think its borderline even for gaming.
For the rest its just meh. A single channel is acceptable for 2c cpu but i would always go for dual just to sleep better i guess you will save it. Thats the smartest

 

[jana519]

Yup, that's what I would like to see benched with DC DDR4: the IGP. Run some games on it. How does it stack up vs a dGPU? I'd like to see how it does against low-end cards like the GT 730, GT 740, and R7 250.

 

[VirtualLarry]

Interestingly, in CPU-Z's memory tab, the NB (Northbridge, AKA UnCore) clocks between 800Mhz and 3300Mhz.

 

[Bouowmx]

File compression: 7-Zip or WinRAR.

 

[lopri]

Can you monitor the NB frequency?

Try running Stream and Linpack to test memory bandwidth. Linpack is not a memory bandwidth benchmark but it is rather responsive to memory performance.

 

[SPBHM]

I would be curious to see how it performs on 3dmark06, so it would be something to compare it with like old VGAs like a 8800GT.

 

[PingSpike]

Yeah, I'm always more curious to see how old benchmarks run. Typical reviews usually focus on telling us whether it gets 2 or 3 fps in BF1.

 

[hojnikb]

find an asrock z270 and OC it

 

[krumme]

Stuff like tf2...

 

[VirtualLarry]

Just a note, the DeskMini doesn't have a PCI-E x16 slot for a video card, and I'm not disassembling it to build a desktop rig with the chip. Although, in the coming months, I may build a desktop rig with a G4560 or G4600, too.

I'll think about some gaming benchmarks, I don't own very many modern games.

 

[DrMrLordX]

Play some f2p games like TF2, Dirty Bomb, um World of Tanks Blitz maybe?

 

[krumme]

Counterstrike
Bf3 at low res.
Tf2
Dota

All should be doable and show difference to dual channel. And is stuff people play.
Hell i might join in with my skl laptop.

I cant see 8 vs 16 GB should have an influence.

 

[VirtualLarry]

My subjective experience, was that the G3900 (2.8Ghz SKL dual-core) w/8GB DDR4-2400 @ 2133, when upgraded to the G4600 (3.6Ghz KBL dual-core w/HT), even while I still had the single 8GB DDR4-2400 @ 2400 SO-DIMM, was noticeably faster.

I upgraded my G4400 (SKL 3.3Ghz dual-core), to a G4600, and swapped in the additional 8GB SO-DIMM, giving me 16GB DDR4-2400 @ 2400, dual-channel.

And it didn't seem much faster, honestly. Maybe web browsing is limited by my internet connection, or I had gotten used to the speed of my other DeskMini with the G4600, or maybe it was just because performance (in Handbrake 1.01) of my G4400 @ 4.455Ghz in my Z170 Pro4S was similar to the G4600 at stock, and I was used to the browsing performance of that level already.

TBH, I had some browser lagginess too, that I felt that I shouldn't have.

Other than that, they're running well.

G4600 $90 from AntOnline on ebay
DeskMini 110W $144 from Newegg on ebay
GSkill 2x8GB DDR4-2400 CAS16 1.2V $103 from Newegg on ebay
PNY CS1111 240GB 2.5" SATA6G SSD $60 from BestBuy on ebay
Windows 10 Pro - from some Win7 Pro keys

$397 + OS cost

Plus a $30 Club3D DP-to-HDMI2.0 active adapter, to connect to my HDMI2.0 4K UHD TV @ 4K60.

 

[cbn]

find an asrock z270 and OC it

I thought the locked Kabylake chips couldn't be overclocked?

 

[cbn]

Stuff like tf2...

Yes, definitely TF2.

1080p and especially 4K......on the iGPU.

P.S. Here were my results using Athlon 5350 as a comparison point. (short story: 128sp with single channel DDR3 1600 was enough iGPU for 1080p low......with CPU as the weak link)

I'm done with the first phase of multiplayer testing for Athlon 5350 on Team Fortress 2.

(This phase involved changing resolution with the lowest detail settings being used as a constant.)

Some basic observations:

1. The game appears to spread the load out primarily to three cpu cores. (Activity does exist on core four, but it is extremely light)

2. Some player character classes appear to stress the game more than others. For example, I seem to get the lowest frame rates while playing the Scout.

3. According to FRAPS minimum frame rates have dropped to as low as 16 on multiple logs. I have tested this on resolutions as low as 800 x 600 with lowest settings and it still happens.

4. Determining average frame rate is extremely difficult because once my player dies (and I am waiting to respawn) the frame rate shoots up much higher than what it was in game. With that mentioned, average frame rate appears to be > 30 FPS up to 1920 x 1080 (lowest settings). Some variation in average FPS appears to be dependent on map and number of players.

Overall, I will give the Athlon 5350 a pass (for casual gamers) on Team Fortress 2 up to 1920 x 1080 (lowest settings). However, the minimum frame rates reported by FRAPs can be alarmingly low. As I have mentioned, lowering resolution does not help the reported minimums. Ideally I would like to see AM1 with a higher clock on the CPU in order to help the minimum FPS.

 

[VirtualLarry]

I thought the locked Kabylake chips couldn't be overclocked?

That was my understanding.

Something about the only reason that they could BCLK OC locked SKL CPUs, was that Intel let slip a "debug microcode" that shut off the PMU (power-management unit - it manages clocks, among other things), and that by using that debug microcode, the mobo makers could then increase the external bus clock, to overclock the SKL locked CPUs.

Unless they've got a hold of another "debug microcode" for KBL locked CPUs, I don't think that we'll see a repeat. Unless of course I'm wrong, and it doesn't depend on debug microcode. (Which, then, why is the PMU shut off to BCLK OC SKL CPUs?)

 

[cbn]

Deleted.

 

[cbn]

My subjective experience, was that the G3900 (2.8Ghz SKL dual-core) w/8GB DDR4-2400 @ 2133, when upgraded to the G4600 (3.6Ghz KBL dual-core w/HT), even while I still had the single 8GB DDR4-2400 @ 2400 SO-DIMM, was noticeably faster.

I upgraded my G4400 (SKL 3.3Ghz dual-core), to a G4600, and swapped in the additional 8GB SO-DIMM, giving me 16GB DDR4-2400 @ 2400, dual-channel.

And it didn't seem much faster, honestly. Maybe web browsing is limited by my internet connection, or I had gotten used to the speed of my other DeskMini with the G4600, or maybe it was just because performance (in Handbrake 1.01) of my G4400 @ 4.455Ghz in my Z170 Pro4S was similar to the G4600 at stock, and I was used to the browsing performance of that level already.

TBH, I had some browser lagginess too, that I felt that I shouldn't have.

How was the webpage scrolling on your 4K TV once the page was fully loaded?

 

[Conroe]

That was my understanding.

Something about the only reason that they could BCLK OC locked SKL CPUs, was that Intel let slip a "debug microcode" that shut off the PMU (power-management unit - it manages clocks, among other things), and that by using that debug microcode, the mobo makers could then increase the external bus clock, to overclock the SKL locked CPUs.

Unless they've got a hold of another "debug microcode" for KBL locked CPUs, I don't think that we'll see a repeat. Unless of course I'm wrong, and it doesn't depend on debug microcode. (Which, then, why is the PMU shut off to BCLK OC SKL CPUs?)

Something seems to be set to not allow a bus speed of 103 mhz or more, at least with my asus z270. 102.9 works. The setting goes from 40 to 650. I've not tried underclocking. Asrock or Supermicro may have done better. Maybe the full range of bus clocks will work with my 7700k? (Worthless.) I do have a small overclock on my 6500. 7700k will be here soon.

 

[VirtualLarry]

How was the webpage scrolling on your 4K TV once the page was fully loaded?

I'm trying to judge that. With Waterfox full-screen, and using the middle-mouse-wheel-click smooth-scrolling feature, it's mostly smooth, but it has a little bit of judder.

I tried setting Power Profile to High Power, still not any change (unless that requires a reboot to take effect).

It does seem smoother than with only single-channel RAM (higher frame-rate), but still not quite "glass smooth", like an Android phone scrolling.

One thing that I forgot about when I bought these two G4600 CPUs, is that Waterfox actually DOES, I think, have some AVX optimizations that it can utilize. So I would be interested in comparing it on a real i3 (SKL or KBL, either one).

Another thing, I went to Intel's site trying to download drivers for my iGPU on this G4600, they gave two downloads, one beta one not, but both of them are dated prior to Jan 2017 when KBL desktop was released, and neither driver package will install, they both claim that I don't meet the minimum requirements for some reason. Not sure what is going on.

Currently using whatever driver Win10 1607 64-bit auto-installs. Seems to have at least basic functionality, including audio via DisplayPort. (Yes, it has the Intel Graphics Control Panel.)

 

[coercitiv]

Another thing, I went to Intel's site trying to download drivers for my iGPU on this G4600, they gave two downloads, one beta one not, but both of them are dated prior to Jan 2017 when KBL desktop was released, and neither driver package will install, they both claim that I don't meet the minimum requirements for some reason. Not sure what is going on.

Currently using whatever driver Win10 1607 64-bit auto-installs. Seems to have at least basic functionality, including audio via DisplayPort. (Yes, it has the Intel Graphics Control Panel.)

Look for you GPU name, not CPU name when downloading drivers. And for your second problem with driver packages refusing install:
1. download desired driver package
2. cut internet connection (optional)
3. uninstall device driver from Device Manager - choose to also delete the driver software (repeat this step in the unlikely event the system installs an even older driver version)
4. reboot and install new driver from step 1

To explain some of the above - Intel seems to have some checks put in place to avoid installing their generic drivers over OEM optimized drivers. Sometimes this strategy backfires on their own software, including system auto-updates (this was my case with a Skylake desktop, iGPU only). Cutting internet connection is optional, normally a simple uninstall + driver delete at step (3) will allow Windows to install new auto-update driver, which in turn will not stop you from manually installing newer drivers.

 

[cbn]

I'm trying to judge that. With Waterfox full-screen, and using the middle-mouse-wheel-click smooth-scrolling feature, it's mostly smooth, but it has a little bit of judder.

I tried setting Power Profile to High Power, still not any change (unless that requires a reboot to take effect).

It does seem smoother than with only single-channel RAM (higher frame-rate), but still not quite "glass smooth", like an Android phone scrolling.

One thing that I forgot about when I bought these two G4600 CPUs, is that Waterfox actually DOES, I think, have some AVX optimizations that it can utilize. So I would be interested in comparing it on a real i3 (SKL or KBL, either one).

Another thing, I went to Intel's site trying to download drivers for my iGPU on this G4600, they gave two downloads, one beta one not, but both of them are dated prior to Jan 2017 when KBL desktop was released, and neither driver package will install, they both claim that I don't meet the minimum requirements for some reason. Not sure what is going on.

Currently using whatever driver Win10 1607 64-bit auto-installs. Seems to have at least basic functionality, including audio via DisplayPort. (Yes, it has the Intel Graphics Control Panel.)

Virtual Larry,

Are you using any DPI scaling?

 

[VirtualLarry]

will allow Windows to install new auto-update driver, which in turn will not stop you from manually installing newer drivers.

The only video driver on here IS the auto-update driver, and yes, it IS stopping me from installing the Intel drivers. That said, the newest Intel driver I could find in a search ("7th generation Intel graphics"), were from 12/2016, not even Jan 2017. So I'm still waiting for updated drivers for my iGPU.

 

[VirtualLarry]

Are you using any DPI scaling?

Nope.