Faster Mem?

ransak

Distinguished
Aug 28, 2011
3
0
18,510
Hello,
I am getting the P8Z68V-Pro and an i7 2600k. The i7 mem controller supports ddr3 1333. Is there any advantage to getting faster ddr3 if using this i7?
 
Solution
Not true, the Sandy Bridge has a robust IMC and the P8Z68-V PRO can run up to DDR3 2200. However, in gaming the there's no discernible advantages above DDR3 1600 CAS 8/9, but in Multi-Tasking there is an advantage to higher frequency. Though the cost of ultra high frequency RAM gets expensive with diminishing returns plus a degradation in stability.

Great article -> http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/1
Specs P8Z68-V PRO -> http://www.asus.com/Motherboards/Intel_Socket_1155/P8Z68V_PRO/#specifications

Further since the cost per GB of RAM has sharply dropped I only recommend 4GB/stick density or 8GB Dual Channel 1.50v kits.

My favorite (2) kits:
2x4GB G.SKILL Ripjaws X Series 8GB (2 x...
Not true, the Sandy Bridge has a robust IMC and the P8Z68-V PRO can run up to DDR3 2200. However, in gaming the there's no discernible advantages above DDR3 1600 CAS 8/9, but in Multi-Tasking there is an advantage to higher frequency. Though the cost of ultra high frequency RAM gets expensive with diminishing returns plus a degradation in stability.

Great article -> http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/1
Specs P8Z68-V PRO -> http://www.asus.com/Motherboards/Intel_Socket_1155/P8Z68V_PRO/#specifications

Further since the cost per GB of RAM has sharply dropped I only recommend 4GB/stick density or 8GB Dual Channel 1.50v kits.

My favorite (2) kits:
2x4GB G.SKILL Ripjaws X Series 8GB (2 x 4GB) DDR3 1600 F3-12800CL8D-8GBXM - http://www.newegg.com/Product/Product.aspx?Item=N82E16820231445
4x4GB CORSAIR Vengeance 16GB (4 x 4GB) DDR3 1600 CMZ16GX3M4A1600C9 - http://www.newegg.com/Product/Product.aspx?Item=N82E16820233143
 
Solution



A 'robust' IMC?? Explain that... Because the fact that its robust won't make it run any faster???

And yes, if you are an enthusiast you can force/overclock anything to higher than specified, but for 90% of users, 1333MHz is what it will run at.
 
I disagree with you're saying, most enthusiast run DDR3 1600 and lately in 4GB/stick density.

The typical Intel build is more expensive and in general has 'better' components added to it like RAM, GPU, etc. More budget orientated builds are indeed AMD with DDR3-1333, and OEM builds most diffidently use DDR3-1066 or DDR3-1333 and offer no BIOS options for faster RAM i.e. faster RAM is forced to run slower.

IMO - more than half the folks that have purchased DDR3-1600 or higher frequency RAM have it actually running at 'Default Frequency' and don't even realize the problem.

Robust Examples:

AMD 4 or 6 core:
DRAM Frequency -> DDR3-1600MHz ; Errors
DRAM Frequency -> greater) >DDR3-1600MHz e.g. 1800, 2000, etc ; Post Failure
DRAM Voltage -> <1.50v ; Errors or Post Failure
OC REQUIRED >DDR3-1600MHz ; FSB -> 240MHz+ OR CPU-NB Frequency -> 2400MHz ; Is Required
4GB/stick density ; Requires BIOS update most of the time

Intel Sandy Bridge:
DRAM Frequency -> DDR3-1066~DDR3-2200 ; Works
DRAM Voltage -> 1.25v~1.70v ; Works
NO OC REQUIRED!!!
4GB/stick density ; never a problem

The current AMD IMC is and has been horribly weak, touchy and unstable in comparison to Intel. My hope is that this will be corrected shortly the AMD Bulldozer, I have seen the tests but cannot draw a complete conclusion.
 
I've got an AMD 1090T six core and I've overclocked my RAM to 1600MT/s with 8-8-8- timings at 1.6v and had it working even higher. My board doesn't support faster RAM than 1600 but t still overclocked easily.

I realize that AMD can't always overclock the RAM as high as Intel with as much ease but even among enthusiasts a lot of people don't try to go much faster than 2000MT/s.

What do you mean sandy bridge doesn't need to overclock to get higher speeds than 1333MT/s? that isn't true because ALL speeds for DDR3 above 1333MT/s are overclocked speeds. With things like XMP it may be easier to overclock the RAM but it's still overclocked.

I haven't had any RAM problems overclocked or not with my motherboard and its an old crap board compared to current boards.
 
I'm confused here, are you trying to school me from WIKI? I'get' Sine wave frequency and 'double clocking' per cycle and its synchronization to BCLK/FSB - SDRAM.

You're simply not aware of "RAM problems overclocked or not..," problems. Overwhelming chances are unless you increased either the FSB or CPU-NB Frequency your RAM is throwing errors especially under stress. Yep, I run across 6-core AMD all the time with this 'hidden' problem until the OP runs some stress tests. Try Prime95 Blend for a few hours. You'll be jumping into you BIOS like flies to sugar.

I am NOT an advocate of DDR3-1866 or faster RAM in this forum; please feel free to Google my posts. However, on my 980X gaming rig I run DDR3-2000MHz, DDR3-1600 on my desktops, and RDIMM DDR3-1333 on my servers.

Further, you'll very RARELY see any of my posts recommending XMP - I always recommend manually setting both the Frequency and CAS Timings. XMP simply does the same, by setting Frequency & CAS Timings, using one of two profiles encoded into the RAM similar to 'SPD' {JEDEC}, but it does NOT OC the BCLK.

In fact, on the Sandy Bridge you need/should leave BCLK alone. Don't confuse your OC {CPU/FSB/BCLK} side to running RAM at their Rated Frequency; that same argument can be said with 800 vs 1066 vs 1333 Frequency; it's at best technically an OC relative to the CPU's RATED IMC; hence the 'Robust Intel IMC.' However, the CPU in just the SAME exacting way is an OC via the same Multiplier principle to BCLK/FSB and its Rated 'GHz' Speeds. Otherwise AMD CPUs would be running FSB speeds (200MHz) and Intel to its' BCLK.

The LGA 1366 i7 9XX IMC craps-out around 2200MHz (160 BCLK), the Sandy Bridge 2300MHz~2400MHz (100MHz BCLK), the AMD Phenom II x4 around 1800MHz, and the AMD Phenom II x6 around 2000MHz - BUT only with extensive FSB OC >240MHz.
 
My mobo doesn't let me change my BCLK so it's not a problem. I said that the AMD CPUs top out about 2000 and intel tops out at 2500 which is confirmed by many and you just agreed with me on the AMD side at least.

People aren't always able to get a decent overclock with their hardware but other people are able to. I don't have any problems with my minor RAM overclock on my 1090T six core and I realize not every one can get away with it but I fail to see how the AMD controller is horribly weak when it can get to 2000MT/s.

I never said I ran faster than 1600 on my pc and I can't run faster than that because my mobo won't allow it and my RAM probably wouldn't. Why would I have to change BIOS settings if I'm not even doing a major overclock and I'm running stable? I also never said that you mentioned XMP I mentioned it as a response to what you said. Your post said sandy-bridge doesn't need to overclock to get higher speeds (at least that is what it looks like you're saying) which is false and that was my point.

If I missed/misinterpreted something please explain what it was because when you said no overclock required for sandy-bridge I assumed that you meant no overclock required to increase RAM speeds.

 
I'm clear, IF your RAM is indeed 'running' DDR3-1600 aka CPU-z DRAM Frequency = 800 MHz -> http://www.cpuid.com/medias/images/en/softwares-cpuz-04.jpg then I can almost guarantee Prime95 + Blend will produce a lot of errors and/or fail with no FSB or CPU-NB Frequency manual tweaking (OC). The Command Rate alone will produce errors if not set per spec. If you simply installed DDR3-1600, 1800, etc and did nothing then 100% it's running at 1333MHz {667MHz} - period.

Stable vs Crashing vs Errors corruption; Stable to me is as error free as possible, and I find it odd you did not thoroughly testi vs ass-u-ming it's all fine. Ostrich with head in the sand. All my PC's are fully stress tested: Prime95 and Blend for 12~24 hours, Memtest 4-passes, Futuremark, ATTO, etc. Eventually, you'll have a document or something else that will be corrupt - real world. I assume nothing - ever.

In contrast to AMD, the Sandy Bridge can run 1600~2200 with a SIMPLE DRAM Frequency setting.
1333MHz ; Memory Multiplier = 13.33
1600MHz ; Memory Multiplier = 16
1866MHz ; Memory Multiplier = 18.66
Both the CPU and BCLK are 100% running STOCK/untouched.

Correct re: I assumed that you meant no overclock required to increase RAM speeds. Impossible, period, to do on the current AMD. I also said, hopefully this will change with the Bulldozer; I realize it's new IMC is rated for DDR3-1866. I haven't a clue how it will run at faster DDR3 Frequencies or at low voltage - time will tell.

As I stated, "OC relative to the CPU's RATED IMC" however, not an OC of BCLK or CPU Multiplier. Intel and AMD OC especially now in completely different ways.

I spend 50X the time helping folks with AMD to get stable DDR3-1600, in contrast my typical SB/i7 9xx DDR3-1600 is one post followed by "Good Luck!".
 
Lets get some nomenclature right, the 'MT/s' is internal to the RAM's IC. The MHz is Synchronous to MOBO's BCLK and CPU's IMC.

Example:
BCLK (or FSB) * Memory Multiplier = Memory Frequency (MHz)
200MHz * 8 = 1600MHz = 1.6GHz
100MHz * 16 = 1600MHz = 1.6GHz

The FSB Ratio:
200MHz : 1600MHz / 2 = 200 : 800 = 2 : 8 = 1 : 4

What you described above is BOTH an OC to the Rated Frequency AND CAS of the RAM PLUS an IMC 'OC vs Rated.' So running 1333MHz 8-8-8 is nearly equivalent to 1600MHz 9-9-9; my experience with AMD 1600 MHz + low CAS, less than 9, and CAS 8 or especially CAS 7 is damn near impossible without a FSB/CPU-NB Frequency adjustment. IF you have an e.g. ASUS and are using DOCP then you ARE OC the CPU-NB Frequency -- the BIOS is doing it for you.

No not really, Memtests job is to primarily test the RAM with little to no stress on the CPU other than Test #7 which does stress the IMC. This is half the picture, Prime95 Blend test stress BOTH the CPU AND the Memory which is why it's a good real world stress test instead of a 'is your RAM good' test. Memtest further completely avoids the OS Environment -- it's a RAM Integrity test.

I'd fall over if your can pass the Prime95 Blend...
 
You need to get some nomenclature right.. 1600MT/s isn't 1.6GHz it's double pumped 800MHz. Double pumping isn't the same as doubling the frequency even though most marketers wrongly advertise them as the same thing. Latencies aren't double pumped so they apply to half the transfer rate which means that CAS of 7, 8, or 9 apply to 800MHz of 1600MT/s DDR3 because that's the actual input/output clock speed of the ram. The input/output clock is 4 times the memory clock for DDR3 regardless of Intel or AMD so for DDR3 1600 it's 200MHz (memory clock) * 4 (bus multiplier) equals 800MHz (input/output clock) and data transfers are *2 again for double pumping.

The bus connecting the RAM and CPU is what the 100*16 and 200*8 for RAM speeds apply to. Those aren't FSBs for current Intel and AMD CPUs these buses are different. The FSB connects the CPU to the northbridge while the memory controller connects the rest of the CPU to the RAM. The connection between the controller and the RAM replaced the FSB Intel abandoned the FSB when they made the Core i CPUs in 2008 and AMD ditched the FSB earlier.

The bus connecting the RAM to the CPU is what the BCLK applies to, not the RAM. The bus may run at 1.6GHz but that is the bus running at 1.6GHz not the RAM which is double pumped 800MHz. You can say the MT/s is just internal and I would agree if everything about the RAM were double pumped but since not everything is double pumped (such as timings), MT/s is not the same as MHz internally or externally regardless of the bus.

I'll try the Prime95 Blend some time and we'll see what happens.
 
The calculations and nomenclature above are on the system bus side, not the memory itself -- look again. Duh, the Memory itself internally is running, see example, at 800MHz. However it expresses itself to the FSB/BCLK to MHz.

The maximum bandwidth is calculated below for DDR3-1600
DDR3-1600 = 800MHz * 2 (double data rate) = 1600MHz
1600MHz * 64-bit data bus * 1bit/8bytes = 12,800MB/s

More WIKI Crap:

1. Front-side bus - http://en.wikipedia.org/wiki/Front-side_bus - http://en.wikipedia.org/wiki/Front-side_bus
8B x 100 MHz x 4/cycle
= 8B x 100M x Hz x 4/cycle
= 8B x 100M x cycle/s x 4/cycle
= 3200MB/s

DO YOU SEE ANY MT/s in the calculations and nomenclature??

2. Transfer (computing) - http://en.wikipedia.org/wiki/Transfer_(computing)
"...and megatransfers per second (MT/s) are informal language that refer to the number of operations transferring data that occur in each second in some given data-transfer channel."

3. List of Intel chipsets - http://en.wikipedia.org/wiki/List_of_Intel_chipsets#Core_i_Series_chipsets
Error: States the PCIe is 2.0 when in fact it's PCIe 3.0 X79, Z77, Z75, and H77
See -> http://www.anandtech.com/show/4318/intel-roadmap-ivy-bridge-panther-point-ssds/2

I kinda knew this before I posted ;)
 
We don't use a front side bus anymore... Intel stopped using the FSB when Nehalem came out in 2008 and AMD stopped using it with the K8 family in 2003.

The RAM latencies operate on half the transfers per second and that s why I refuse to call it MHz. If everything worked either double the frequency or double pumped internally then it wouldn't matter if it were called MT/s or MHz because operationally they would be identical. Since latencies like CAS and such are measured in the actual clocks of the RAM regardless of the bus, MT/s and MHz are not identical and I will not call them identical if they aren't.
 
Your MT/s is like calling minutes 'quacks' or 60 quacks = 1 hour. I appreciate WIKI but it is filled with erroneous, incorrect and misleading information.

Otherwise SHOW ME how to get MT/s into MB/s Transfer Rate...I'm waiting.

The calculations and examples prove MHz over MT/s to calculate Transfer Rates.

AMD sure does have a 200 MHz FSB {Front Side Bus} and Intel used a 133.33MHz or 100MHz BCLK {Base Clock}.

Since WIKI is your Bible:
AMD "FSB frequency 200 MHz System clock HyperTransport up to 3.2 GHz" - http://en.wikipedia.org/wiki/Socket_AM3

Another WIKI contradiction - http://en.wikipedia.org/wiki/LGA_1155 vs http://en.wikipedia.org/wiki/List_of_Intel_chipsets#Core_i_Series_chipsets
 

I needed quote you on every post, you keep editing, adding, changing your posts - after the fact.

At this point I'm tired of our little debates because you 'CHEAT' your responses in fear of looking BAD. At this point, I'd have to reread every response just to keep up with some new twist or deletion.

You stated, "I refuse to call it MHz" because I just proved it's MHz. Call them 'Butterflies/s' or 'Blazorthon/s'
 
This isn't about looking bad.

You mentioned the FSB often even though we don't even use it outside of the outdated LGA 775 platform and MHz are not equal to MT/s. Only the data transfer of DDR3 RAM is double pumped which is why latencies apply to half the MT/s because half the MT/s is the actual MHz.

Your analogy doesn't work because a transfer occurs twice per Hz. The whole point of double pumping is too lower clock speeds while heightening transfer speeds.

The CPU/NB frequency is not an FSB. An FSB is a different type of bus and is not used on new Intel hardware or AMD hardware. I refuse to rename MT/s as MHz because they are different things in the context of all DDR RAM to date.
 
You are measuring the speed of the bus between the RAM and on-die controller and calling it the speed of the RAM. Just because the bus runs at 1.6GHz doesn't mean the RAM runs at 1.6GHz. The RAM still runs double pumped at 800MHz. The RAM can keep up with the bus with data transfers but the latencies work off the real clock speed of the RAM so the RAM still is limited by its clock between transfers.

A MHz is either the rising or the falling edge of a clock whilst MT/s are both the rising and falling edge of the clock for DDR3.

That is why MHz and MT/s are not the same for DDR3 RAM and why I refuse to call them such.
 

It sure the hell doesn't have a BCLK. You came in on an Intel {BCLK} discussion with your AMD {FSB} added side discussion, and your AMD sure the hell uses a FSB {Front Side Bus} in particular the 200MHz variety with your 1090T. WIKI's 'Front-side bus' is screwing you up. Example - Current AM3 http://en.wikipedia.org/wiki/Socket_AM3 next click the link 'FSB' frequency which takes you to http://en.wikipedia.org/wiki/Front-side_bus which is apprently what you're reading and Trolling this discussion.


Never said it was, to OC the RAM on your AMD there are as I said (2) methods.
"OC REQUIRED >DDR3-1600MHz ; FSB -> 240MHz+ OR CPU-NB Frequency -> 2400MHz ; Is Required"


Math multiplying HZ: 1*10Hz = 10Hz, 200*10Hz = 2,000Hz = 0.002MHz
Math Transfer Rate {CPU IMC <=> RAM peak rate}
The maximum bandwidth is calculated below for DDR3-1600
DDR3-1600 = 800MHz * 2 (double data rate) = 1600MHz
1600MHz * 64-bit data bus * 1bit/8bytes = 12,800MB/s ; the 12,800MB/s is exactly what WIKI DDR3 lists so apparently it is after all 1600MHz to the Bus!


Never said it did.

Bottom-line, you're not going to change my understanding by jumping around WIKI.
 
Then READ this again! {MT/s and MHz}

2. Transfer (computing) - http://en.wikipedia.org/wiki/Transfer_(computing)
"...and megatransfers per second (MT/s) are informal language that refer to the number of operations transferring data that occur in each second in some given data-transfer channel."

Definition:
in·for·mal/inˈfôrməl/
Adjective: Not formal; having a relaxed, friendly, or unofficial style, manner, or nature: "an informal atmosphere"; "an informal agreement"
 

ransak

Distinguished
Aug 28, 2011
3
0
18,510
Thanks to all 3 of you for addressing this question. I think I'm going to wait to see what Q1 of 2012 brings us. FWIW, I tend to agree with jaquith here. Again, thanks!