RAM Latency

[kongwane]

Which ram latency and type is best out of the following??


1) Cosair DDR400 512mb XMS Pro 2-3-3-6
2) Kingston DDR400 512mb hyperX 2-2-2-5-1 (CL2)
3) Transcend DDR400 512mb
4) Kingston DDR400 512 ECC 3-3-3 (CL3)

 

[Person]

Cosair DDR400 512mb XMS Pro 2-3-3-6

 

[Beta]

2 2 2 5 1t
Option 2
Lower equals better
As for which is the best brand,debatable....you need to do some google for reviews.

 

[kongwane]

so, its between the first to options? which one?, the kingston hyperX is looking good tho

 

[Beta]

XMS is the version with the funny flashy lights on it,dunno what lights have to do with RAM but hey....apparently they overclock terribly.
Wait a couple hours for someone like ShockG to provide you with experienced based knowledge,he can then tell you bout the actuall chips on each model etc

 

[mic_y]

I have had some really bad experiences with the Kingston HyperX... When buying fast RAM you want to OC it, and for that there is nothing better than OCZ. Quite hard to come by in SA, but apparently Person has got, so consider giving him a shout.

Otherwise, the Corsair is a better choice, even tho on paper it seems slower, will give u more of an overclocking margin...

 

[Beta]

Hi,for discussion sake,you can't say a brand is bad because of your personal experience....Kingston is highly regarded RAM in many circles.
I'm not sure if it's right to assume that he's gonna OC the RAM.
Before I decided on a purchase just recently,I read a review with those XMS sticks and apparently they had hardly any headroom,I went for Xtreme RAM from the Team group.

 

[useless]

o have a couple of those XMS Corsair sticks and on a ASUS SLI Premium mobo no issues with head room. A bigger PC case might help for the reviewer of that article.

 

[Beta]

I give up ,lol
Buy whatever you want!!!Hell buy 128Megs of value RAM!

 

[kongwane]

DUDes is it that hard to answer a simple question, REALLY NOW!!! C'MON............

 

[pupa]

DUDes is it that hard to answer a simple question, REALLY NOW!!! C'MON............

It a tuff one to answer. The only way to really know is to check the brand for review and comparison on google as there is no way to check and advise you unless the rams were bench-marked. Those latency's means nothing unless bench-marked in a Mobo that is not affected by the latency settings as if you overclock the ram can seem to perform well but the performance of the Mobo can drop. Thats why I do not think you will get advice Not from me anyway. Those Kingston's look tempting but I will spend the extra few rand and get DDR II 800's. The units can always be used in better mobo's later but you cannot use DDR 400's for performance in DDR 800 Boards! IMHO!

 

[werner]

your question was

Which ram latency and type is best out of the following??

the answer is #2

your question has since mutated into: which ram stick is best for OC etc? answer: whichever is cheapest. synthetic benchmarks may be able to tell the difference, however you wont. ever. conversely, the extra money will still be tangible, and in your wallet. go buy a hooker with the savings

 

[ShockG]

Hold on now.
Always and I mean always try and get the lowest latency you can get, then followed by the highest speed.
If Hynix had chips that could do 2-2-2-5 those would be faster than any XMS 2-3-3-6 anyday.

Since the BH is out of circulatio Im guessing that HyperX is Samsung TCCD/TCC5 in which case it'll give you tight timings at 400~433MHz maybe 440MHz, but ater that you'll have to relax the timings to say 2-3-3-6 or 3-4-4-10 etc...

Now if you start out with the XMS at 2-3-3-6 @ 200MHz, you obviously don't have as much headroom. But that also depends on the chips used.
If they are th Hynix EL4 or D5 they are good for upto DDR550 depending on the PCB/SPD used. If its the generic 6/4-layer stuff then you are unlikely to get high speeds, unless you get really good chips. But if you get Brainpower PCB's then youll make the DDR533/550.

More than the timings, it may be better if you just go for high speeds esp considering that getting BH/UTT chips these days will be like finding a TNT that plays Oblivion.

2) Kingston DDR400 512mb hyperX 2-2-2-5-1 (CL2)
Your best option in that list!

 

[pupa]

Tight Timings vs High Clock Frequencies: MB FSB and Money wins! IMHO

As I said its an complex issue and as I wanted to do some refreshing myself did some research of more recent tests and findings. Here goes! Toms Hardware explains and confirms our sentiments as posted here! This is a mouth full but worth you while if you ponder!
tight_timings_vs_high_clock_frequencies
Intro:
http://www.tomshardware.com/2006/03/31/tight_timings_vs_high_clock_frequencies/index.html
Conclusion:
http://www.tomshardware.com/2006/03/31/tight_timings_vs_high_clock_frequencies/page10.html

Benchmark Conclusions
We conclude the following from our testing:

There are very small real-life differences in performance between low clock frequency/fast timing Winbond memory and high clock frequency/relaxed timing Samsung memory. This is true in 3DMark01 and SuperPI even for CPU/memory intensive applications.
To accurately answer the question we asked earlier in this article - namely, whether to go for tight timings or high clock frequencies - one should conduct the tests using a very fast CPU to eliminate bottlenecks. It is our opinion that even our overclocked 2610 MHz dual core Opteron wasn't really fast enough to do more than hint at a possible victory for tight timings at even higher CPU clock frequencies.
Given the two facts above, our conclusion must be that our testing using only Samsung memory and extrapolation, instead of comparison to actual Winbond memory, does not result in data accurate enough to give an entirely foolproof answer. According to our calculations, the difference between CL2.0-2-2-6 at 270 MHz and CL3.0-4-4-7 at 320 MHz in SuperPI 8M swayed from 0.7% to 0.5% in both directions, depending on how performance scales.
Testing indicated that tight timings become more important as CPU clock is raised. This could potentially lead to Winbond memory performing 1-2% better than Samsung memory at Athlon 64/Opteron CPU speeds over 3 GHz. Of course, Winbond is no longer an active player on the DDR1 market, so getting a hold of this kind of memory is so hard that the issue becomes moot.
Depending on CPU clock frequency, tight timings have a performance advantage over relaxed timings in CPU/memory intensive applications, ranging from 2% at 2 GHz to 6% at 2.6 GHz.
When leaving the timings untouched at a CPU clock of 2 GHz or 2.6 GHz respectively, DDR600 performs 2% or 5% better than DDR400 in CPU/memory intensive applications.
For Samsung TCC5/TCCD memory, like the G.Skill F1-4400DSU2-1GBFC used for this article, DDR600 at medium timings outperforms DDR400 at tight timings. DDR600 at relaxed timings performs about the same as DDR400 at tight timings (+/- 1%).

the performance difference seems to be minuscule
we would not advise anyone to spend a lot of money on highest end memory in the hope of improving computer performance by increasing memory speed. As noted earlier in the article, keeping the timings unchanged and at a steady CPU clock of 2.6 GHz, DDR600 performs 5% better than DDR400 in CPU/memory intensive applications. These are very weak gains given that there is a 50% increase in memory speed, and these gains are even smaller at lower CPU speeds. In modern games, which are mostly limited by the graphics card, the performance increase would be zero, as even big changes in CPU speed can go by unnoticed.

The bottom line is that as long as you have enough memory - preferably 2 GB - the extra money you pay for more memory speed would be better invested in a faster graphics card. And if you don't play games, then the CPU and hard drive offer more room for improvement than the memory.

So again IMHO its an complex issue and no straight answer exists, too many dependables, When I overclocked the last thing I played with was the ram timing to see how it effects the mobo stability at best timings without droppng the MBoard benchmark. Ditto! So:

- Spend the money on higher MB FSB, Higher level CPU (OC) and VGA card
- Choose DDR2 800/1066 standard ram to suit MB above the "specialised" expensive ram as the true gains will be minimal.

Motherboard / RAM compatibility is more important! (Ram is cheap by comparison anyway!)

Some more on ram performance:
Intro:
http://www.tomshardware.com/2006/04/11/live_memory_test/index.html
Conclusion:
http://www.tomshardware.com/2006/04/11/live_memory_test/page20.html

The the outcome also shows that there is a sweet spot for current DDR2 memory. All products reached at least DDR2-800 speeds; most of the DIMMs were capable of running at DDR2-900 level. While you should never hope for retail memory to reach the same clock speeds as review samples, you can certainly expect all available quality DDR2 memory to hit the 400 MHz mark (DDR2-800) at slightly increased voltages.

Thus Buy cheap pray and overclock, or just buy the best value for money Highest spec DDRII MHZ for $ and overclock!

Live Memory Test: Overclock 'Em Till They Crash

Intro:
http://www.tomshardware.com/2006/04/11/live_memory_test/index.html
Conclusions:
http://www.tomshardware.com/2006/04/11/live_memory_test/page19.html

The outcome of this stress test run was surprising, because the Corsair test system crashed at 456 MHz memory clock. This equals DDR2-912, which is 8.9% below the rated speed of DDR2-1000. A-Data's memory was able to support up to 476 MHz, which corresponds to DDR2-952. But this is still well below the DDR2-1000 specification! However, both memory products were capable of running at their rated speed on an Asus P5WD2-E, which shows that you should pay close attention to the motherboard recommendations of your favorite memory vendor.

If you still decide to go for the highest-end you should bear in mind that it might not work the way you want it.

Very Important: A data is available here! Rough estimates!
Gaming: Desktop Memory : 6-Layer PCB, Life Time Warranty, Retail Package, W/Heat Spreader!!!

ME-A512MB667 R 440 AData® DDR II-667 512MB 240-Pin
ME-A1024MB667 R 850 AData® DDR II-667 1024MB 240-Pin
ME-A512MB800 R 620 AData® DDR II-800 512MB 240-Pin
ME-A1024MB800 R 1250 AData® DDR II-800 1024MB 240-Pin
ME-A1024MB1066 R 2508 SA No.1 AData® DDR II-1066 1024MB 240-Pin

Kingston HyperX - High Performance Desktop Memory Module: DDR2-800 - Life Time Warranty - NEW
KHX1G800D2/2 R 2000 DDR2 800 Kingston® HyperX™ Series 1GB DDR2-800 Memory Kit, 512MB Matched Pair Twin Pack, 240-Pin Module:CL5 Latency
KHX2G800D2/2 R 3250 DDR2 800 Kingston® HyperX™ Series 2GB DDR2-800 Memory Kit, 1GB Matched Pair Twin Pack, 240-Pin Module:CL5 Latency

How Much RAM Do You Really Need?
Some memory requirement advise: VISTA is here! More ram is required!

The Need To Invest In 2 GB Of RAM Is Not Marketing Hype! Consider that the increased use of memory-intensive applications such as video encoders has already caused 1 GB configurations to go mainstream. With Microsoft's upcoming Windows Vista OS release needs ram!

http://www.tomshardware.com/2005/12/13/how_much_ram_do_you_really_need/
Conclusion: Very interesting for all!!!!!!!!!!
http://www.tomshardware.com/2005/12/13/how_much_ram_do_you_really_need/page12.html

2 GB?[/B]
Still there are situations where more than 1 GB is what you want.

If you are a professional user, you might need more than 1 GB for really heavy applications.
If you intend to do heavy multitasking, especially if you have more than one CPU or CPU core. Running RAM intensive games such as World of Warcraft, downloading files via high speed FTP or encrypted protocols, Bittorrent or any P2P program; decompressing large archives and playing large size video files in a window or on second monitor all at the same time can max out your system memory pretty fast - if your CPU can handle it

Upgrading RAM?
http://www.tomshardware.com/2006/03/24/navigating_the_memory_upgrade_jungle/index.html

Some comments from posters here: http://forumz.tomshardware.com/hardware/Tight-Timings-High-Clock-Frequencies-ftopict180531.html

-Latencies sure help, but I think it's better to seek higher frequencies first and then try to tighten the latency
-While it's true that tighter timing will have that edge over high Mhz during overclocking, one must not ignore the clear advantage of high Mhz when both cores are used at higher than a frequency of 2.6Ghz.
-do realise that when you increase frequency without changing the latency cycles, you're actually decreasing latency timings (if it takes 3 cycles to perform an operation at 200 MHz, if you increase frequency to 300 MHz the operation will take the same time as 2 cycles at 200 MHz. So, increasing frequency is a dual effect operation (more throughput, faster responsiveness).

The best being, of course, faster bus speed and tighter clock timings

There is a risk however, is that the RAM can't take the decreased timing and skip operations: if it can't go faster than 3 CL, setting it to 2 CL may have it skip every odd operation (which needs to be delayed) and actually increase timing to 4 CL!

Don't forget, too, that increasing frequency may lead to asynchronous HT/FSB and RAM clock - big performance hit.

 

[ShockG]

Kongwane seems to have ask a very secific question about latency and really size etc...

Maybe he could be doing it for 3DM01/03/06/AM3/SPi scores in wich case he is still better of with tight timings.

I don't believe high frequencies can without limitation make up for tight timings.
I've seen it myself.
Ie Just reently I managed to get a set of OCZ TCC5 ram to 680MHz! at 3-5-4-10, giving around 8.8GB/sec on the S939 4800+. Now another guy was/is running HyperX at 2-2-2-5 @ 540MHz and he was/is getting 8.2GB/sec.
I have a 600MB/sec advante but I need to be 140MHz faster. How many DDR1modules make 600 let alone 667/680MHz?
540MHz is a stretch, but 250MHz 2-3-3-6 is a lot more reasonable than a DDR600 set at 3-5-5-10.

Esp where 2GB sets are concerened since these top out at the DDR560/580 mark at most with usually sad timings.
Rather DDR500 at 3-3-3-8 than DDR600 at 3-5-5-12.

 

[pupa]

Kongwane seems to have ask a very secific question about latency and really size etc...

Maybe he could be doing it for 3DM01/03/06/AM3/SPi scores in wich case he is still better of with tight timings.

I don't believe high frequencies can without limitation make up for tight timings.
I've seen it myself.
Ie Just reently I managed to get a set of OCZ TCC5 ram to 680MHz! at 3-5-4-10, giving around 8.8GB/sec on the S939 4800+. Now another guy was/is running HyperX at 2-2-2-5 @ 540MHz and he was/is getting 8.2GB/sec.
I have a 600MB/sec advante but I need to be 140MHz faster. How many DDR1modules make 600 let alone 667/680MHz?
540MHz is a stretch, but 250MHz 2-3-3-6 is a lot more reasonable than a DDR600 set at 3-5-5-10.

Esp where 2GB sets are concerened since these top out at the DDR560/580 mark at most with usually sad timings.
Rather DDR500 at 3-3-3-8 than DDR600 at 3-5-5-12.

I share this sentiments, But if the motherboard can handle a higer FSB then I would opt for the Higher MHZ RAM (Cost difference is low) and then Optimize the timing for performance. Unless he is brave and run at higher voltages with Ram Heatsinks? NOT? I will not waste money on the High performance sticks if you can get higher gains out of Adata or sticks with Samsung chips and then rather increase ram to 2G?

I am to upgrade two gaming systems soon and thinks the DDRII 800 overclocked with Motherboard FSB 1066 (AM2) or P4 930D overclocked hopefully would give a good trade off?

 

[pupa]

Pushing Your DDR2 RAM To The Max: AData® DDR II-1066 1024MB

Proof? that Money for speed in ram is money wasted?

http://www.tomshardware.com/2005/07/18/pushing_your_ddr2_ram_to_the_max/

Nice hardware forums: Overclocking!

http://www.xtremesystems.org/forums/forumdisplay.php?f=164

 

[JTech]

As expected, the speeds are very similar. You can only read a DRAM cell as fast as the design speed allows with current technology.

 

[ShockG]

As expected, the speeds are very similar. You can only read a DRAM cell as fast as the design speed allows with current technology

Cool. Explain.

 

[JTech]

I surely will:

If you compare DRAM to SRAM, you will notice that SRAM will always outperform DRAM and the reason is not difficult to understand. Your CPU's cache is constructed out of high-speed SRAM.

SRAM is too expensive and power-hungry to implement in any significant size. a SRAM cell is 4x as complicated as a DRAM cell. I could explain how it works here but somehow I don't think you'll understand all of it.

Whether you like it or not, your fancy CPU waits for the DRAM, the DRAM controller, depending on what it is doing, may assert a wait cycle to the CPU before the CPU can actually read or write location X in memory. And why must the CPU wait?

DRAM requires operations to work, regardless of the type DRAM/SDRAM/DDR SDRAM/DDR2 SDRAM they all have quite a few things in common:

The entire memory array needs to be refreshed, this means within 64uS every single bit in that memory chip has to have been refreshed and this is the same as a read/write cycle except you're not reading or writing anything, you're simply recharging the capacitors that hold your data. The CPU cannot interrupt a refresh cycle so somehow the memory controller has to ensure that a 50/50 rule applies that the CPU has enough time to do it's thing with the memory and the controller has enough time to refresh the memory to prevent data loss. It's a tradeoff and always will be!

DRAM cannot be addressed simply with a number like you can with SRAM, the DRAM controller has to translate your physical address 0x42434566 into row address, column address and bank address. This requires at least two memory clock cycles to latch into the registers within the DRAM chips themselves before the memory will even consider putting out data.

Nearly all generation-specific DRAM technologies are at the same level, meaning that most manufacturers are pushing the same boundaries of technology. In general, for a given DRAM memory cell the speed is generally the same, as most chips are manufactured to meet standard speed grades i.e. 7ns, 10ns, etc... As revealed by the article, playing with the timings will yield miniscule improvements, unless of course one was very clever and you arranged all your code to fit into the SDRAM in very well organized structure so that the SDRAM controller wouldn't have to sit and precharge rows all over the place, it could precharge one row and execute code in a linear fashion out of that row. That would speed things up nicely. However in reality code is code and it is allocated a chunk of RAM in an arbitary manner by the OS and that's it, the MMU and the SDRAM controller are left to sort out the mess.

In reality, there is something else that would also affect your speed, the chipset, the quality of the layout of a given motherboard (trust me, DDR2 PCB layout is no picnic and is not easy at all) and various other factors such as noise.

I have designed a DDR memory interface in the recent past and the above is information I have learnt in the process.