I5 2500K Vs. RAM frequency and latency

[ToucHDowN 0]

After many days of research and reading, after having considered all restrictions set by my local retailer (some selected brands only available, etc), I'm ready to order about what you see in my signature. My problem, as the topic suggests, has to do with the proper choice of RAM in terms of frequency and latency.I'm not concerned with RAM performance that much. This benchmark is quite self-explanatory about actual, "tangible" difference in performance between different types of RAM. What really bothers me is the possible barrier, a not wisely chosen RAM, may impose over any CPU overclocking attempt. So I kindly ask: (1) Am I ok with a 1333MHz DD3 CL7 1.65V RAM in terms of being able to overclock the i5-2500k up to-say-4,5GHz or more? (125 euros).(2) If the answer is YES, is there any chance these 1.65 volts to create any kind of problems in the near future? How will they be affected by CPU's overclocking? Should I cool the memory? (3) If the answer to (1) is NO, what do you think of a 1066MHz DD3 CL9 1.5V solution? (99 euros).(4) Or equally, what do you think of a 1066MHz DD3 CL7 1.5V solution? (99 euros as well).Before you answer, please consider the following F.A.Q. from Intel site:

What are the Intel® Core™ i7 processor and the Intel® Core™ i5 processor DDR3 memory voltage limitations?Intel® recommends using memory that adheres to the Jedec memory specification for DDR3 memory which is 1.5 volts, plus or minus 5%. Anything over this voltage can either damage the processor or significantly reduce the processor life span.

Finally, I guess 1600MHz RAM is out of the question along with an i5-2500K, since Intel also states the following:

What is the maximum frequency for DDR3 memory when used with the Intel® Core™ i7-2xxx processor series, the Intel® Core™ i7-800 processor series, the Intel® Core™ i5-2xxx processor series, the Intel® Core™ i5-700 processor series, and the Intel® Core™ i5-600 processor series?These processors support DDR3 memory with a maximum frequency of 1333 MHz. If faster DDR3 memory is used (such as 1600 MHz or higher), it will be down-clocked to operate at 1333 MHz.

Thank you in advance.

[cmeeks 8]

(1) Sandy Bridge only officially supports up to 1333, but motherboards support up to 2133. Get a minimum of 1600MHz.(2) 1.65v is fine on Sandy Bridge, but I wouldn't recommend going any higher. Most of the Sandy Bridge compatible memory is even 1.65v(3) 1600 or higher is the only solution(4) terrible solutionTo sum it up - get memory with a minimum speed of 1600MHz. Try to get as tight of timings as possible (basically all those numbers, 6-8-6-24, for example... you want them all as small as possible, but the first number is the most important). If you want to avoid any potential hassles, check to make sure your memory selection is compatible by checking your motherboard's qualified vendor list (QVL). To further prevent any potential hassle, only populate four ram slots. Some people populate all four slots with no problems, but that's not always the case. Memory also requires less voltage when you populate fewer slots.Intel is correct that faster memory will automatically be down-clocked to 1333. It just switches the memory multiplier to 13.33. All you have to do is go into the BIOS and manually set the memory multiplier to 16, 18.33, 21.33, etc depending on your memory speed. The memory multiplier is multiplied by the BCLK (100MHz).Read the conclusion to this article

[dazz 69]

I just tested a couple different configs:2133 CL8Frames, Time (ms), Min, Max, Avg 10670, 300000, 23, 50, 35.5672133 CL9Frames, Time (ms), Min, Max, Avg 10669, 300000, 23, 49, 35.5631600 CL7Frames, Time (ms), Min, Max, Avg 10428, 300000, 23, 49, 34.7601600 CL9Frames, Time (ms), Min, Max, Avg 10099, 300000, 22, 47, 33.6631333 CL7Frames, Time (ms), Min, Max, Avg 10047, 300000, 22, 46, 33.490

[ToucHDowN 0]

I just tested a couple different configs:2133 CL8Frames, Time (ms), Min, Max, Avg 10670, 300000, 23, 50, 35.5672133 CL9Frames, Time (ms), Min, Max, Avg 10669, 300000, 23, 49, 35.5631600 CL7Frames, Time (ms), Min, Max, Avg 10428, 300000, 23, 49, 34.7601600 CL9Frames, Time (ms), Min, Max, Avg 10099, 300000, 22, 47, 33.6631333 CL7Frames, Time (ms), Min, Max, Avg 10047, 300000, 22, 46, 33.490

Thanks for sharing this. You're actually fully aligned with these aforementioned results from another source, indicating that RAM speed makes no (visible) difference in gaming, fsim in this case. The "humble" 1333MHz shows some teeth by holding 22fps Min, compared to the "mighty" 2133MHz with its 23fps Min. value. The rocketing cost (euros/$) from 1333 to 2133 does simply not worth it.Blue Skies.

[ToucHDowN 0]

(1) Sandy Bridge only officially supports up to 1333, but motherboards support up to 2133. Get a minimum of 1600MHz.(2) 1.65v is fine on Sandy Bridge, but I wouldn't recommend going any higher. Most of the Sandy Bridge compatible memory is even 1.65v(3) 1600 or higher is the only solution(4) terrible solutionTo sum it up - get memory with a minimum speed of 1600MHz. Try to get as tight of timings as possible (basically all those numbers, 6-8-6-24, for example... you want them all as small as possible, but the first number is the most important). If you want to avoid any potential hassles, check to make sure your memory selection is compatible by checking your motherboard's qualified vendor list (QVL). To further prevent any potential hassle, only populate four ram slots. Some people populate all four slots with no problems, but that's not always the case. Memory also requires less voltage when you populate fewer slots.Intel is correct that faster memory will automatically be down-clocked to 1333. It just switches the memory multiplier to 13.33. All you have to do is go into the BIOS and manually set the memory multiplier to 16, 18.33, 21.33, etc depending on your memory speed. The memory multiplier is multiplied by the BCLK (100MHz).Read the conclusion to this article

Thanks Corey for a thorough reply.My question remains though: is it ok to use 1600MHz and 1.65v while overclocking an i5-2500k from stock 3.30 to -say- 4,8GHz? Does CPU overclocking affects (raises) RAM's voltage? IF this is a YES, then I guess OC is out of the question since 1.65v is already marginal for the i5-2500k/P67 combo. Right?Given so, next available option is an 1.5v RAM. How does this voltage affects (restrains) the maximum OC result frequency I can get to my CPU?Pls excuse my ignorance. I've been sweeping all relevant forums about this, with no luck so far...Blue Skies.

[alainneedle1 97]

Thanks Corey for a thorough reply.My question remains though: is it ok to use 1600MHz and 1.65v while overclocking an i5-2500k from stock 3.30 to -say- 4,8GHz? Does CPU overclocking affects (raises) RAM's voltage? IF this is a YES, then I guess OC is out of the question since 1.65v is already marginal for the i5-2500k/P67 combo. Right?Given so, next available option is an 1.5v RAM. How does this voltage affects (restrains) the maximum OC result frequency I can get to my CPU?Pls excuse my ignorance. I've been sweeping all relevant forums about this, with no luck so far...Blue Skies.

If I may, 1.5v or 1.65v will be the rated voltage (on the sticker) for the spesific stick you will buy regardless of the overclock, if one set say 1.5v you have to run them at 1.5v or i.65v if that's what they are rated at, some can be rated at 1.65v and can run at 1.6v depending on how your rig will be tuned but if you are overclocking you will probably have to run them at spec. or raise the voltage a little over what they are rated at to stay stable, no set will react the same way, exp: mine are rated at 1.65v but because of my overclock (BCLK at 201) I have to run them at 1.7v., also because they are rated at 2000MHz I have to raise the QPI/vtt at 1.47v to stay stable with my high overclock but since you are asking about the SB I don't know about what kind of voltage you'll need on the QPI/vtt.So if your question is "will it be safe to run them at 1.65v" yes it will if they are rated at 1.65v, on paper (intel spec.) they can run at up to 1.8v but you will not need that.http://www.newegg.co...NodeId=1&name=6Getting more specific here....Now as far as MHz vs timing, what make the 1333NHz stick run as good (or almost as good) as the 2133MHz is the tight timing on them, if you were running the 2133MHz at 7-7-7-21 you'll see a better result, not as far as FPS but the stutterings will be reduced, trust me on this, I tried it myself.I don't like the middle number on these but that is to show you something else... http://www.newegg.com/Product/Product.aspx?Item=N82E16820231355 http://www.newegg.co...N82E16820231451I tried some 200MHz at 9-9-9-24 on my rig, maybe I lost 1 or 2 FPS compare to the one I have right now but I saw more stutterings with them. Tighter timing will reduce the stutterings MHz for Mhz. so if you are buying the 1600MHz buy the one with the lowest number you can find on the timing, 1600MHz at 6-6-6-21 will probably be more expensive that the one at 1600MHz 8-8-8-24 but you pay for what you get.Another piece of advise if I may, if you are thinking about overclocking your rams (by the way anything over 1333MHz is already overclocked) make sure you have a stable rig and do so taking baby steps, my rams are rated at 2000MHz 7-7-7-21 2T trcf (112) and I am now running them at 2010MHz 7-7-6-20 1T trcf (95), I will spare you about other settings you'll have to consider to do that but if your intention is to overclock them down the road go online and read a lot before doing so.The 2600K with 2100MHz cas 8 will do good.

[dazz 69]

It's true that those fraps results don't tell us anything about stuttering. I'm gonna try to do this with the frametimes files (e.g. fsx 2011-03-07 12-40-09-15 frametimes.csv)

Frame, Time (ms)   01, 	0.000   02,	23.286   03,	43.599   04,	64.371   05,	84.378   06,   104.803   07,   124.312   08,   144.348   09,   165.403   10,   185.910   11,   205.855   12,   226.313   13,   245.659   14,   265.715   15,   287.381   16,   307.440   17,   328.716   18,   349.401   19,   369.192   20,   389.759   21,   409.154   22,   428.957   23,   449.540   24,   469.359   25,   489.477   26,   510.355   27,   530.324   28,   551.058   29,   570.831   30,   590.613   31,   611.401   32,   631.643   33,   652.957   34,   674.269   35,   697.769   36,   719.903   37,   739.957   38,   760.066   39,   780.632   40,   800.123   41,   820.041   42,   840.522   43,   861.769   44,   884.972   45,   905.429   46,   928.671   47,   951.178   48,   971.695   49,   994.101   50,  1013.844

Maybe using the time-in-betwwen-frames can give us an idea of how smooth the test was: inverting that value should provide an instant FPS value. Long pauses / low instant FPS should denote stutters

[alainneedle1 97]

Hey! Dazz, to me the best thing to do to show how FSX run from one rig to another is to make a vid. of the screen when running FSX from an external source...exp: filming your FSX screen when flying with the FPS counter on with a camera over your shoulder.....like that one can show how his FSX run for real without loosing any FPS from using fraps or any other source like that.

[dazz 69]

Very true Alain. That's the ultimate way to gauge smoothness. I'm going to try that anyway, just in case it proves useful to help obtain analytic / objective data from those fraps files in regards to smoothness & stuttering ... that and because I don't have a decent video cam hehe

[ToucHDowN 0]

Hey! Dazz, to me the best thing to do to show how FSX run from one rig to another is to make a vid. of the screen when running FSX from an external source...exp: filming your FSX screen when flying with the FPS counter on with a camera over your shoulder.....like that one can show how his FSX run for real without loosing any FPS from using fraps or any other source like that.

Dazz,This seems a bright idea regarding the (possible) correlation between RAM frequency and Stuttering, combined with Corey's tool. But even so, any findings will necessarily point to the specific scenario set by FSX Benchmark11. Different user settings, various textures added by numerous add-ons (not just scenery add-ons), may lead to different results. Please consider this before you engage in such a time-consuming task (although I'd be really interested -after all these years of stu-stu-stu-ttering- to see finally some proof about it).Blue Skies.

[dazz 69]

It's ok George. I'll be using the same files I posted the max/min/avg from earlier today. It shouldn't be that hard with excel

[dazz 69]

For wht it's worth here are the resultsIn short:2133 CL8<30 1425 <20 32 <10 4 2133 CL9<30 1450 <20 38 <10 4 1600 CL7<30 1707 <20 36 <10 5 1600 CL9<30 2226 <20 50 <10 6 1333 CL7<30 2366 <20 61 <10 6That's the number of dips below 30, 20 and 10 FPS. There seems to be a pattern of increasing stutters with slower RAM. Not sure if it's a valid test or the differences are significant, but that's what I could gatherThe benchmark is the one I always use in UK2000 EGLL + PMSG744 in the VC, 100% MyTraffic airliners, Extremely dense AG & scenery, aircraft shadows, all sliders maxed out except for water (Low 2.x), 30% cars, REX2, GEX, UTX, 8xSQ FSAA, App controlled AF, Vsync off, RejectThreshold=131072

[alainneedle1 97]

This is "kinda" confirming what I always said, high overclock (4.4GHz +) will make good use of high MHz memory, tight timing is also better.2010MHz at 7-7-6-21 1T is a lot better than 2133MHz at CL8, don't know what are the other timing numbers 8-10-8? The middle number is also very important as it is the amount of time in cycles for issuing an active command and the read / write commands.

[dazz 69]

I'm not sure I would break the bank for 1600 CL6 or 2133 CL7 having 1600 CL7 or 2133 CL8 so much cheaper, but that's just me. I don't think anyone would notice the difference in two unlabeled vids to be honest, but who knows. Would be cool to do that test so I'll try to get me a video cam and set up a poll or something.I'll see if I can run one more test @ 2133 CL7 now.The complete timings were as follows:CL9 = 9-11-9-30CL8= 8-10-8-27CL7 = 7-9-7-24

[alainneedle1 97]

The thing is FSX love low latency, like I said before FPS wise you will not see a big difference but stutterings wise it does from what I've tested so far.I'll be waiting to see if you can run 2133MHz at 7-9-7-24 without BSOD....you'll probably need to play with QPI/vtt voltage but if you can run at that speed and timing be sure to post the result.