Mr. Fox

All Memory is Not Created Equal - HyperX 2133 CL12 versus Ripjaws 2133 CL11

82 posts in this topic

I already knew (or so I thought) that gains or losses from different memory speeds and timings made only a small difference in real-world performance, but the allure of something faster is a hard bug to shake. I grabbed some G.SKILL Ripjaws 2133 CL11 memory to replace my Kingston PnP 2133 CL12 fully expecting to see a very modest improvement in performance for my already insanely powerful M18xR2. Unfortunately, the opposite was true. There was a substantial decrease in performance... enough to be seen and felt.

I have used G.SKILL memory many times in the past and it has always been great. The Ripjaws was 100% compatible and rock-solid stable. It booted with no BIOS adjustments (all "auto" settings) and ran at advertised speed without any drama. The decline in performance was a disappointment, but NewEgg issued an RMA promptly. I am sharing the results so nobody else makes the same mistake. The images in the album below speak louder than words.

Tests were run repeatedly to confirm there was no anomaly. All BIOS settings were identical between the different brands. I alternated HyperX > Ripjaws > HyperX > Ripjaws and the differences were consistent in all tests.

4 people like this

Share this post


Link to post
Share on other sites

Very nteresting. Did you compare all the timings? Secondary and tertiary timings might not be set that well...

1 person likes this

Share this post


Link to post
Share on other sites

Quite some difference indeed.... thanks for sharing. Did you try to reset BIOS defaults and then load the XMP profile? I wonder if it might have kept some settings from the kingston configuration...

1 person likes this

Share this post


Link to post
Share on other sites

Was the g.skill only single rank? If so that directly explains the bandwidth loss. If it's single rank then the memory sticks will only have 4 modules per side, but dual will have 8 per side.

3 people like this

Share this post


Link to post
Share on other sites

post-11295-14494997185214_thumb.jpg

Was the g.skill only single rank? If so that directly explains the bandwidth loss. If it's single rank then the memory sticks will only have 4 modules per side, but dual will have 8 per side.

You right. Only single rank, and that explains the speed loss. Samsung low voltage chips Samsung B-Die SDRAM package is SEC 340 K4B4G0846B HYK0.

post-11295-14494997185039_thumb.jpg

2 people like this

Share this post


Link to post
Share on other sites
Very nteresting. Did you compare all the timings? Secondary and tertiary timings might not be set that well...
Yes, all of the timings were similar to those that I set manually for enhanced performance.
Quite some difference indeed.... thanks for sharing. Did you try to reset BIOS defaults and then load the XMP profile? I wonder if it might have kept some settings from the kingston configuration...
Yes, I cleared the CMOS just in case because I thought that might be what was happening. There are no XMP profiles on this memory.
Was the g.skill only single rank? If so that directly explains the bandwidth loss. If it's single rank then the memory sticks will only have 4 modules per side, but dual will have 8 per side.
[ATTACH]10905[/ATTACH][ATTACH]10906[/ATTACH]

You right. Only single rank, and that explains the speed loss. Samsung low voltage chips Samsung B-Die SDRAM package is SEC 340 K4B4G0846B HYK0.

Ah, I learned something new. I was not aware of this. The Kingston is 8 per side. The Ripjaws worked flawlessly, very stable... just unacceptably slower.

post-119-14494997185403_thumb.jpg

Might have to try some of this to see how it compares to the Kingston. By having only two sticks the CR is 1, it jumps to 2 with 4 sticks and manually setting 1N mode in the BIOS with 4 sticks is less stable.

CORSAIR Vengeance Performance 16GB (2 x 8G) 204-Pin DDR3 SO-DIMM DDR3 2133 Laptop Memory Model CMSX16GX3M2B2133C11

2 people like this

Share this post


Link to post
Share on other sites

[MENTION=119]Mr. Fox[/MENTION] also what voltage were you running the chips at? I am running 8GB sticks with the same memory chips (so I have dual rank), but am running into overheating issues at 1.5V. They also only overclock to 2000MHz at 10-10-10 even at 1.5V, so I figure I'll look into dropping the voltage some since it seems to have little impact on performance. I absolutely cannot get 2133 to run in dual channel at any timings, and that's with hardmods to increase vccsa and vccio.

1 person likes this

Share this post


Link to post
Share on other sites

I ordered a 16GB (2x8) kit of the Vengeance 2133 CL11 today. I noticed with 2 sticks versus 4 that the CR is 1T with both the Kingston and Ripjaws. It automatically sets itself for 1T with 2 sticks and changes to 2T with 4 sticks, so I am hoping the 2x8 CL11 Vengeance gives me at least a small bump in benchmarks compared to my 4x4 Kingston.

The drop in 3DMark11 Physics Test is what shocked me the most. I did not expect the Physics score to take such a heavy hit from the Ripjaws memory performance. Looks like 3DMark11 Physics test is a really good one for a number of things. My 4930MX also has a hard time with it compared to the 3920XM. Now I am wondering if the DDR3L 1600 memory is holding back the 4930MX from achieving good results. I may drop the Vengeance RAM into the Alienware 18 for testing first, before I move it to the M18xR2. That should be interesting.

@Mr. Fox also what voltage were you running the chips at? I am running 8GB sticks with the same memory chips (so I have dual rank), but am running into overheating issues at 1.5V. They also only overclock to 2000MHz at 10-10-10 even at 1.5V, so I figure I'll look into dropping the voltage some since it seems to have little impact on performance. I absolutely cannot get 2133 to run in dual channel at any timings, and that's with hardmods to increase vccsa and vccio.

I initially ran with everything in the BIOS set to default/auto and everything worked fine other than low performance. I changed to manual settings and 1.60V, then 1.65V and all of the performance tests were identical... same low performance regardless of voltage. It runs at 1.50V in the M18xR2 by default, There is no lower voltage option. There did not seem to be any memory overheating issues. I did try to tighten up some of the timings to improve the performance and it did not change anything.

1 person likes this

Share this post


Link to post
Share on other sites

@J95 - even so, your 2133 RAM running at the lower 1866 CL9 is still faster than the Ripjaws was running at 2133 CL11. As @Khenglish pointed out, this is likely due to the Ripjaws being single rank. Your Vengeance is dual rank.

I ordered Vengeance 2133 CL11 this morning from the same place as the link you posted. NewEgg has been out of stock for a while now.

1 person likes this

Share this post


Link to post
Share on other sites

With the M18xR2/XMP profiles CL10 2133MHZ CL12/13 2400MHz rated L (1.5V) headroom...XMP will work.

1 person likes this

Share this post


Link to post
Share on other sites

2x4Gb Samsung M471B5273DH0 in MSI GE60 0NC laptop.

post-11295-14494997186501_thumb.png

2 people like this

Share this post


Link to post
Share on other sites
2x4Gb Samsung M471B5273DH0 in MSI GE60 0NC laptop.

[ATTACH]10912[/ATTACH]

Yeah I was thinking that 1866 10-10-10 1.35V chips could match, if not beat that memory at 1.5V, but I was wrong. I should have gone with the 3 year old samsung sticks :(

1 person likes this

Share this post


Link to post
Share on other sites

I can hardly wait to find out if the 2x8 Vengeance 2133 CL11 will beat this (which is still very good) 4x4 Kingston HyperX PnP 2133 CL12 :Banane47:

post-119-14494997188148_thumb.jpg

1 person likes this

Share this post


Link to post
Share on other sites

post-3012-14494997189425_thumb.png

With the R4 the trick is matching the BIOS XMP profile timings/settings first (1.65V enabled) then set the XMP profile, reboot -> disable memory override support(XTU), 1.65V/set now XTU will read timings properly.

2 people like this

Share this post


Link to post
Share on other sites
I can hardly wait to find out if the 2x8 Vengeance 2133 CL11 will beat this (which is still very good) 4x4 Kingston HyperX PnP 2133 CL12 :Banane47:

[ATTACH]10916[/ATTACH]

You lucky. You have Kingston HyperX PnP 2133 CL12 with chips Samsung K4B2G0846D HYK0. It is the best chips in the world! Never 2x8 Corsair Vengeance 2133 CL11 will not win them (if you overclock your Kingston HyperX). Because Corsair Vengeance 2133 CL11 have chips Samsung K4B4G0846Q HYK0. In laptops max for this chips 2133 MHz with timings 10-11-11-27.

But you can now overclock your Kingston HyperX PnP 2133 CL12 with chips Samsung K4B2G0846D HYK0 to 2133MHz with timings 9-10-10-24, or if not work 10-11-11-24. Just set in bios setup from "Default DIMM Profile" to "Custom Profile" and set correct timings.

post-11295-14494997190988_thumb.png

post-11295-14494997189865_thumb.png

post-11295-14494997190459_thumb.png

1 person likes this

Share this post


Link to post
Share on other sites

Revision (Dec. 2012)

M : 1st Gen.

A : 2nd Gen.

B : 3rd Gen.

C : 4th Gen.

D : 5th Gen.

E : 6th Gen.

F : 7th Gen.

G : 8th Gen.

H : 9th Gen.

K4B2G0846D

2 density 2Gb

Revision: 5th Gen. 'D'

Speed

F8 DDR3-1066 (533MHz @ CL=7, tRCD=7, tRP=7)

H9 DDR3-1333 (667MHz @ CL=9, tRCD=9, tRP=9)

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

MA DDR3-1866 (933MHz @ CL=13, tRCD=13, tRP=13)

Power

C Commercial Temp.( 0C ~ 85C) & Normal Power(1.5V)

Y Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

K4B4G0846Q

4 density 4GB

Revision: 18th gen.? 'Q'

Speed

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

MA DDR3-1866 (933MHz @ CL=13, tRCD=13, tRP=13)

Power

C Commercial Temp.( 0C ~ 85C) & Normal Power(1.5V)

Y Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

----------------------------------------------------------------------------------------

HYK0

H Package Type FBGA (Halogen-free & Lead-free)

Y Temp & Power Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

1 person likes this

Share this post


Link to post
Share on other sites
Revision (Dec. 2012)

M : 1st Gen.

A : 2nd Gen.

B : 3rd Gen.

C : 4th Gen.

D : 5th Gen.

E : 6th Gen.

F : 7th Gen.

G : 8th Gen.

H : 9th Gen.

K4B2G0846D

2 density 2Gb

Revision: 5th Gen. 'D'

Speed

F8 DDR3-1066 (533MHz @ CL=7, tRCD=7, tRP=7)

H9 DDR3-1333 (667MHz @ CL=9, tRCD=9, tRP=9)

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

MA DDR3-1866 (933MHz @ CL=13, tRCD=13, tRP=13)

Power

C Commercial Temp.( 0C ~ 85C) & Normal Power(1.5V)

Y Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

K4B4G0846Q

4 density 4GB

Revision: 18th gen.? 'Q'

Speed

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

MA DDR3-1866 (933MHz @ CL=13, tRCD=13, tRP=13)

Power

C Commercial Temp.( 0C ~ 85C) & Normal Power(1.5V)

Y Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

----------------------------------------------------------------------------------------

HYK0

H Package Type FBGA (Halogen-free & Lead-free)

Y Temp & Power Commercial Temp.( 0C ~ 85C) & Low VDD(1.35V)

K0 DDR3-1600 (800MHz @ CL=11, tRCD=11, tRP=11)

Yes that right. Corsair just overclock his memory, just writed overclocked frequency and timings in SPD. And it work nice, without any problems.

G Skill and Kingston did the same.

1 person likes this

Share this post


Link to post
Share on other sites
Yes, those timings do work. But, sometimes I get a PFN_LIST_CORRUPT bug check when using them. The more aggressive custom timings also sometimes cause a BSOD with my WiFi and NVIDIA drivers, or browser crashes that never happens with the stock timings. If the Vengeance runs CL11 natively without those issues while running at comparable speeds then I will be satisfied with it.
Spoiler

[FONT=Segoe UI]

On Wed 2/12/2014 1013 AM GMT your computer crashed crash dump file: C:WindowsMinidump21214-14757-01.dmp This was probably caused by the following module: [FONT=Segoe UI]

ntoskrnl.exe

[/FONT] (nt+0x75BC0)

Bugcheck code: 0x4E (0x99, 0x38A25F, 0x2, 0x389A5E) Error: [FONT=Segoe UI]

PFN_LIST_CORRUPT

[/FONT] file path: C:Windowssystem32 toskrnl.exe product: [FONT=Segoe UI]

Microsoft® Windows® Operating System

[/FONT] company: [FONT=Segoe UI]

Microsoft Corporation

[/FONT] description: NT Kernel & System Bug check description: This indicates that the page frame number (PFN) list is corrupted. This appears to be a typical software driver bug and is not likely to be caused by a hardware problem. This might be a case of memory corruption. More often memory corruption happens because of software errors in buggy drivers, not because of faulty RAM modules.

The crash took place in the Windows kernel. Possibly this problem is caused by another driver that cannot be identified at this time.



[/FONT] [FONT=Segoe UI]

On Wed 2/12/2014 1043 AM GMT your computer crashed crash dump file: C:Windowsmemory.dmp This was probably caused by the following module: [FONT=Segoe UI]

ntkrnlmp.exe

[/FONT] (nt!KeBugCheckEx+0x0)

Bugcheck code: 0x4E (0x99, 0x38A25F, 0x2, 0x389A5E) Error: [FONT=Segoe UI]

PFN_LIST_CORRUPT

[/FONT] Bug check description: This indicates that the page frame number (PFN) list is corrupted. This appears to be a typical software driver bug and is not likely to be caused by a hardware problem. This might be a case of memory corruption. More often memory corruption happens because of software errors in buggy drivers, not because of faulty RAM modules.

The crash took place in the Windows kernel. Possibly this problem is caused by another driver that cannot be identified at this time.

[/FONT] [h]Bug Check 0x4E: PFN_LIST_CORRUPT[/h][FONT=Segoe UI] The PFN_LIST_CORRUPT bug check has a value of 0x0000004E. This indicates that the page frame number (PFN) list is corrupted. Important Info If You Have Received a STOP Code
If you received a blue screen error, or stop code, the computer has shut down abruptly to protect itself from data loss. For information about how to recover from this error, see Resolving STOP (Blue Screen) Errors in Windows.
[h]PFN_LIST_CORRUPT Parameters[/h]The following parameters are displayed on the blue screen. Parameter 1 indicates the type of violation. The meaning of the other parameters depends on the value of Parameter 1. Parameter 1 Parameter 2 Parameter 3 Parameter 4 Cause of Error
0x01 The ListHead value that was corrupted The number of pages available The list head was corrupted.
0x02 The entry in the list that is being removed The highest physical page number The reference count of the entry being removed A list entry was corrupted.
0x07 The page frame number The current share count A driver has unlocked a certain page more times than it locked it.
0x8D The page frame number whose state is inconsistent The page-free list is corrupted. This error code most likely indicates a hardware issue.
0x8F New page number Old page number The free or zeroed page listhead is corrupted.
0x99 Page frame number Current page state A page table entry (PTE) or PFN is corrupted.
0x9A Page frame number Current page state The reference count of the entry that is being removed A driver attempted to free a page that is still locked for IO.
[FONT=Segoe UI Semibold]Cause[/FONT] This error is typically caused by a driver passing a bad memory descriptor list. For example, the driver might have called MmUnlockPages twice with the same list. If a kernel debugger is available, examine the stack trace.

:%20Bug%20Check%200x4E:%20PFN_LIST_CORRUPT%20%20RELEASE:%20(1/27/2014)&body=%0A%0APRIVACY%20STATEMENT%0A%0AWe%20use%20your%20feedback%20to%20improve%20the%20documentation.%20We%20don%27t%20use%20your%20email%20address%20for%20any%20other%20purpose,%20and%20we%27ll%20remove%20your%20email%20address%20from%20our%20system%20after%20the%20issue%20that%20you%27re%20reporting%20is%20fixed.%20While%20we%27re%20working%20to%20fix%20this%20issue,%20we%20might%20send%20you%20an%20email%20message%20to%20ask%20for%20more%20info.%20Later,%20we%20might%20also%20send%20you%20an%20email%20message%20to%20let%20you%20know%20that%20we%27ve%20addressed%20your%20feedback.%0A%0AFor%20more%20info%20about%20Microsoft%27s%20privacy%20policy,%20see%20http://privacy.microsoft.com/en-us/default.aspx."]Send comments about this topic to Microsoft [/FONT]
[FONT=Segoe UI]Build date: 1/27/2014[/FONT]





1 person likes this

Share this post


Link to post
Share on other sites
Yes that right. Corsair just overclock his memory, just writed overclocked frequency and timings in SPD. And it work nice, without any problems.

G Skill and Kingston did the same.

K4B4G0846Q Samsung's latest revision/NAND, speed K0/MA.

Yes, those timings do work. But, sometimes I get a PFN_LIST_CORRUPT bug check when using them. The more aggressive custom timings also sometimes cause a BSOD with my WiFi and NVIDIA drivers, or browser crashes that never happens with the stock timings. If the Vengeance runs CL11 natively without those issues while running at comparable speeds then I will be satisfied with it.

2133MHz 11-11-11-27 1.50V~1.55V - 1866MHz 10-10-10-27 1.35V Max (AW18), even when using an XMP profile custom BIOS settings must be identical, 1.65V will stick after disabling memory override support, if XTU disables mem override support it will mostly set default voltage too.

post-3012-14494997192082_thumb.png

RAM stress test, BIOS power lock 55W-68W (6 months old paste job/30C room temp) it's getting better :)

Share this post


Link to post
Share on other sites

Mr. Fox

Set timings:

CL 10

RCD 11

RP 11

RAS 24

WR 10

RFC 120

RRD 5

WTR 6

RTP 6

RC 15

FAW 25

And memory voltage 1,5V.

2 people like this

Share this post


Link to post
Share on other sites

@Klem - Thanks for the suggested settings. My machine does not stay turned on with 1.5V and those timing. It does not attempt to POST with beeps, simply turn itself off before the Intel OROM screen using those settings with 1.5V.

I set it back to 1.65V and then it POST fine, but Windows Explorer crashes repeatedly with those timings. I put tWR back to 16 again, but left all of the other settings that you suggested. Will see if that proves stable with no BSOD like my previous settings.

@J95 - no problem getting 1.65V to stick. I have been running set at 1.65V for almost a year now.

Share this post


Link to post
Share on other sites
@Klem - Thanks for the suggested settings. My machine does not stay turned on with 1.5V and those timing. It does not attempt to POST with beeps, simply turn itself off before the Intel OROM screen using those settings with 1.5V.

I set it back to 1.65V and then it POST fine, but Windows Explorer crashes repeatedly with those timings. I put tWR back to 16 again, but left all of the other settings that you suggested. Will see if that proves stable with no BSOD like my previous settings.

Well, try as you wrote. Your main mistake before: timing RC = 0.

And for quick testing of stability with new timing configuration you can use very nice program TM5.

Share this post


Link to post
Share on other sites

I have my chips at 1.65v constantly in order to achieve 1866Mhz CL9. I don't get as good transfer speeds but my latency is in the 40s.

1 person likes this

Share this post


Link to post
Share on other sites

Samsung K4B2G0846D can very stability work on 1866 MHz with timings 9-9-9-15 and with voltage 1,5V. I tryed much times. But on modern platforms timings are much less important than the frequency.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now