Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

3680

October 31st, 2016 08:00

UPDATED!! Poor Area 51 R2 6900k CPU results

This is an appeal to the community of Area 51 R2 6900k owners out there to please let me know your results, as you can see below the performance buff after 'upgrading' from the 5930k is less than stella, and as Dell support are unable to assist me in confirming the results below, or give assistance in anyway to improve or check the performance, they have advised me to seek feedback from the community...  

Cant say I'm impressed with the Alienware Area 51 R2 and how it utilises the 6900k... below is a comparison between the 5930k and the 6900k

5930v6900.PNG

so that is quite disappointing for such a substantial upgrade, so I thought I would compare it like for like against another 6900k and set the Asus as a baseline

AsusvDell.PNG

HOW can there be THAT much of a difference, I understand there could be OC'ing going on with the Asus but still... More worryingly the drop in GPU performance which after the CPU upgrade I thought would give a subtle boost unleashing throughput but the opposite was the case.

908ti-5930v6900.PNG

I checked bios and everything seems to be showing ok, under CPU power management config cores 7 and 8 ratio limit is set at 0 compare to 39 for 1&2 and 38 for 3-6. Is that correct?

EDIT: changing ratio to 38 yields results

OC'ing to level 2 yields marginal improvements to 19800ish but still miles of any kind of worthy buff, Am I being stupid and there is something I'm missing here??? Any feedback would be great, if not ill have to live with the poor choices I have made...

41 Posts

November 2nd, 2016 01:00

I'm SO glad I didn't pull out all the stops and get the 6950x (feel sorry for the person who benched this!!!):

6900v6950.PNG

NOTE to Alienware.... you HAVE to fix this, the area 51 is the Alienware Flagship machine, what are we to do when an industry recognised single core and multi core benchmark software is consistently showing Alienware flagging behind the rest??

314 Posts

November 2nd, 2016 07:00

Geekbench is ***. You need to tell me the new i5 U (laptop cpu) series scores the same or higher?

Use Cinebench for CPU

41 Posts

November 2nd, 2016 08:00

Thanks for your input, Geekbench scores systems/CPU's against the same tests... so like for like my 6900 should bench the same as another - in theory if un-tampered with (OC). The thing I am making a point of is that the results against another system with SAME CPU is massively retarded for the area 51.

'Each CPU workload models a real-world task or application, ensuring meaningful results. These tests are complex, avoiding simple problems with straightforward memory-access patterns, and push the limits of your system'

For me it gives an indication as to the performance I should get back in RW tasks.

'Cinebench CPU test scenario is designed to use all of your system's processing power to render a photorealistic 3D scene. The test puts up a number of render requirements calculated to drive any processor close to insane, with sharp and blurry reflections, area lights, shadows, procedural shaders, anti-aliasing, and so on.'

Great for rendering... what about PDF export? JPEG export? Speech recognition? Facial recognition? all tested in Geekbench along with other real world tests.

209 Posts

November 2nd, 2016 09:00

 I was going to say that- in you're example of machine vs. machine the multipliers are SET in the asus and AUTOMATIC  in the AlienWare.  OC modes are different in 1 vs 2.  I think this is where the power struggle takes hold,  Cores management vs core STATES.  the size of the Data set during the tests for marks probably didn't  even "wake up"  the AW .  so is nearly at idle states.  The guy that said use the bencmarks that render, serves a pupose in that theres more load, and mapping going on in the computers instructionset, and probably would require the cpu to 'kick it In". ( Im a fan of manual OC anyway)  

41 Posts

November 2nd, 2016 10:00

Thanks @DBREWSTER

OK... how do you know the Asus was manual and Alienware Automatic? i cant see a difference in the comparison? 

209 Posts

November 2nd, 2016 12:00

 As u posted the cpu ratios ( must be some level of OC )  are 0 for some cores.. this is indication of automatic. Usually they are assigned some value, like CPU 'COUNT' is probably 0 as well. This is automatic, and However they do it  CPU states is complicated, anf is configurable,, but usually no choices for it. U config this via registry attributes, and power management modes.( I assume power management in your case is set to Balanced?)

www.tenforums.com/.../8002-how-do-i-disable-core-parking-windows-10-build-10162-a.html

Read through the above POLO6RGTI  has invested some knowledge of the registry  involved with power states and cpu states.etc.  Its very professionally based section there.

41 Posts

November 2nd, 2016 12:00

CPU power management was set for 0 for cores 7+8, I have set them at 38 now. Alienware OC then changes them to 40 for cores 1+2 and 39 for others. This doesn't change the result by any tangible measure, how do you know the Asus is manual and not Automatic? I think it's safe to say Asus bench is with OC to the max, but even looking down to 6900k results from other manufacturers they are still in mid 20k's. I am worried that the architecture and components of the area 51 machine are restricting the power of the CPU, something just does not add up

209 Posts

November 2nd, 2016 13:00

 U might be right.. I just think the asus is OC'd differently  I am sure though the AW will perform as expected with the right settings.  I still think it is sleeping(not enough load)  Im no expert. The Link to the regedit cpu config  would eliminate all low power settings,  (settings?) but I would do it only if recorded the default entries for the power scheme I wanted. ( I guess it is restorable from commandline or power shell, or windows repair.) I think itsa windows problem .    

  I was just guessing from the numbers I read in your post about the cpu config being in auto, etc..

  Asus sometimes is installed with their own OC panel, I forget the name of the asus one. But it IS different.  You need to play with the AW  to get it to clock to your satisfaction.  ( I still have the 5930K

level2 oc with custom inputs.. new win10 install, I use powermanagement "highperf" , In the reg I see just a few things that could be changed , but I think Im running fine for now. I could only change the cpu states, but really the max/min cores is hard set in the powerscheme.. Staes (as shown in the post)  would take a little more time.

41 Posts

November 2nd, 2016 13:00

Ok thanks BDREWSTER , that's a lot to look into, obviously I don't want to stress the machine to the limit, which is possibly what the Asus is doing, but I'm just worried the bios settings or key settings are set so low the system never reaches its potential, I know the 6900 can do better, and your right when I start rendering 4k video in De Vinci resolve and exporting I will hopefully see some gain, I use a bench like the one ref above to compare like for like, not my workflow .

But thanks again you've been the most help so far and that includes alienware support.

209 Posts

November 2nd, 2016 13:00

Even with core multipliers set to 44 and turbo / per core at same, I get a good OC with mailbox tweaks.(bios)

But I have left STATES alone, and # of cores always says 0 in bios. So it does sleep some at 2.9ghz, then 3,5-3.8 under some strain.. but when I load it or do a test I see The Turbo(enabled 3.0 std) works to give the 4.4 ghz . And it doesn't takemuch.. If set the core multiplier in CC different from AUTO The lower GHZ is More/ proportional to the multiplier.  I didn.t want to 'force run' at 4.4 ghz.  

209 Posts

November 2nd, 2016 14:00

ya sir

May 9th, 2017 13:00

Just wanted to add that with RAM at 2666 MHz and Level 1 OC, these are the 5930K benchmarks. I have no doubt that 6950X will be better at multitasking.

Alienware Alienware Area-51 R2 - Geekbench Browser 

8 Wizard

 • 

47K Posts

May 12th, 2017 09:00

Windows 7 vs 10 is also a significant difference.   Windows 10 has tons of background tasks including cortana that must be deleted from the drive in order to not be running all the time.

You also get less junk when you DISABLE ISATAP and Teredo and 6 to 4 and then Disable IP6 services.

Then open an administrator CMD window Before and After.

Run NETSTAT -b -o -a -r -n 

The whole update via windows update strangers on the internet is a horrible default as are Active Tiles on the start menu.

   
netsh int ipv6 isatap set state disabled
netsh int ipv6 6to4 set state disabled
netsh interface teredo set state disable

To re-enable isatap teredo and 6to4,
just replace the disabled with type=default

No Events found!

Top