ROG Zenith Extreme / 1920X - poor game performance & sluggish?

Associate
Joined
22 Jun 2013
Posts
481
Location
Zomerset (Glastonbury)
Hi All,

Not quite sure where to put this but I suspect it maybe a mobo issue but not sure. Basically running games on the ROG Zenith Extreme / 1920X / 1080 FE SLI (AIO watercooled - both CPU & GPU's) / Samsung 960 EVO NVME - running the first game I came across - WOW - and I'm getting about 1/2 FPS I was getting on a 5930K / 1080 FE SLI and about 1/3 less than a 4970K and 9 series titan X (have a very set routine for testing FPS, so the variance should be minimal).

3d mark spy hunter gives about 12,000 SLI and 7,000 single card - seems well below average.

The system feels sluggish even in windows 10.

More details -

- 1080 FE in SLI with Gsync at 100Mhz to Asus 3440x1440p monitor
- Tried all BIOS revisions including beta 0503 (0211 doesn't work due to requirement to use DIMM2 and SLI)
- Cooling and temps are fine, nothing *appears* to be thermal throttling
- RAM is 2133 in bios, can push to 2400 but it becomes unstable
- Tiried stock 3.5 and o/c 3.8,3.9 and 4.0 - whilst all are stable the same problem persists
- brand new install of Windows 10
- latest drivers from Asus website
- GPU's running in PCIE 3.0

Things I'm going try -

- Another game (obviously) just downloading Doom which was a very solid 100FPS
- Unplug the power of a GPU so it uses a single card only
- Revert to a SLI bridge - I was using a EVGA HB SLI bridge but got the more aesthetically pleasing ROG HB bridge to sync with the mobo (grasping at straws)


Any other things I could try / check? Is there some AMD tools I should look at to check things, seem to recall some people mentioning them in reviews - has anyone has to do that?
 
Soldato
Joined
16 May 2007
Posts
3,220
My guess based on ryzen is running the ram at 2133 is having a significant impact on performance. It is suggested going from 2133 to 3200 ram gives a 40% performance increase in the right circumstances for ryzen as cpu performance is directly linked to ram speed.

I assume like ryzen it performs best with Samsung b die based ram but someone will be able to confirm. It may just be a case of waiting for bios updates, after a few updates ryzen ram speeds increased.
 
Associate
Joined
8 May 2010
Posts
426
Location
London
40% performance hit is a lot! I'm glad I read this before ordering memory for my planned Threadripper build. I will be watching this thread, I hope the OP gets the performance where it is expected to be.
 
Associate
OP
Joined
22 Jun 2013
Posts
481
Location
Zomerset (Glastonbury)
So some good news, Doom is running silky smooth and like a dream - lots of cores :)

Good USB problems though - devices are sometimes not recognised i.e. keyboard, HUE+ controller etc. just seems to be 1 per boot or something. Looking for issues around this.

Also BIOS 0503 seems to drop the Nvidia cards, get the message the display is not plugged into a Nvidia GPU - although obviously it is, dropping back to 0305 seems to resolve this although on 1 boot it Windows about 5 minutes to recognise the GPU's. Bit odd.
 
Associate
Joined
28 Nov 2012
Posts
668
So some good news, Doom is running silky smooth and like a dream - lots of cores :)

Good USB problems though - devices are sometimes not recognised i.e. keyboard, HUE+ controller etc. just seems to be 1 per boot or something. Looking for issues around this.

Also BIOS 0503 seems to drop the Nvidia cards, get the message the display is not plugged into a Nvidia GPU - although obviously it is, dropping back to 0305 seems to resolve this although on 1 boot it Windows about 5 minutes to recognise the GPU's. Bit odd.

Out of interest how is SLI 1080's treating you on threadripper?
 
Associate
OP
Joined
22 Jun 2013
Posts
481
Location
Zomerset (Glastonbury)
How did you go with the mems?? 3466 c16 or 3333 c14 your best option?? Few cpu can hit 3600 Imc not capable. But those mems can do 4450Mhz on Intel platform all day long.

Just loaded put them in, loaded a samsung b die profile from the bios sub menu's and got a clean boot at 3466.

Now you'll have to bear with me in terms of C16 / C14 etc. as I've never dived into memory overclocking before just followed the intel XMP.

CPU-Z is reporting -

HKwxQo.jpg
 
OcUK Staff
Joined
20 Feb 2012
Posts
10,178
Location
John Smiths Stadium
Just loaded put them in, loaded a samsung b die profile from the bios sub menu's and got a clean boot at 3466.

Now you'll have to bear with me in terms of C16 / C14 etc. as I've never dived into memory overclocking before just followed the intel XMP.

CPU-Z is reporting -

HKwxQo.jpg
Looks fine.

Intel doesn't benefit as much as Ryzen/ TR interms of performance for afew reasons. But the main one is when you push mems on ryzen you push cache also.

On Intel the cache and memory are overclocked separately. To get the best out of Intel oc mems and cache separately.
 
Associate
OP
Joined
22 Jun 2013
Posts
481
Location
Zomerset (Glastonbury)
Couple of issues at 3466 - a few failed tests (using memtest in windows, followed by failed restarts), just dropping back to 3333 to see if things stabilise a bit more - using the "safe profile". I'm sure with some tweaking 3466 should be fine but want to get a baseline in there.

Using 3333 CAS 14 for a bit.
 
Associate
OP
Joined
22 Jun 2013
Posts
481
Location
Zomerset (Glastonbury)
what are you getting for aida64 cache and memory bench?

I'll check for you and update later.

A further update on the system -

-> 1920x with a 4.0Ghz overclock, cooled via Kraken X52
-> 3333Mhz memory @ Cas 16
-> 1080 FE in SLI

A buttery smooth but with a couple of questions / issues -

1. SLI, when using a non SLI game i.e. world of warcraft the FPS will suddenly half every few seconds even when I've set the game to a single GPU in the Nvidia CP, disabling SLI in CP and the game is fine. SLI games work well in SLI i.e. Doom. I suspect this is a BIOS issue but I'll pop up a separate post asking other X399 users.
2. In terms of memory would 3333 @ CAS16 be as good as 3200 @ CAS14? I suspect there is no perceivable difference
 
Back
Top Bottom