The Great Debate, Sync Ram 1:1?Post Date: 2008-10-11 |
Post Reply
|
Page 12> |
Author | |
Alex
Admin Group Digital Storm Supervisor Joined: 04 Jun 2012 Online Status: Offline Posts: 16314 |
Quote Reply
Topic: The Great Debate, Sync Ram 1:1? Posted: 11 Oct 2008 at 3:35am |
Let's get the assumptions out of the way.
Don't post anything unless you have results to back it up.
On Monday, I'll have Venom post screen shots of benchmarks of a system using a Q9550 (1333MHz FSB) using 1333MHz memory, 1600MHz memory, and 2000MHz memory.
Let's see those benchmarks. From my experience, you will notice a difference, but, obviously not a huge leap (price versus performance) but, if it's within your budget and you want that extra edge, go for the faster ram.
Let the benchmarks do the justice.
If there is anything else that we should test, bring it out.
Cheers,
Alex Edited by Alex - 11 Oct 2008 at 3:39am |
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 3:55am |
its all very simple:
for example if someby has a fsb of 1066MHz ( 266 x 4 = 1066 MT/s, each transfer is 64 bits for a total of 8,533 MB/s transfer rate) So you want your memory to match the 8,533 MB/s transfer rate otherwise what ever memory speed you have over that will be bottleneck by the fsb's 8,533 MB/s transfer rate. Now lets say that you have paired your RAM (2 x 533MHz) ), your memory controller is now able to talk to the RAM transferring up to 8,533 MB/s. Notice that this matches a 1066 FSB, so here the system is balanced, neither the FSB nor the RAM speed is a bottleneck for the data flow between the processor and the RAM. If you put in faster RAM, you can't see a gain in the memory bandwidth you can use, since the FSB becomes a bottleneck. Now on top of all that, all the q9550s I have seen on this forum Alex, all have come out with OC speed of 3.6Ghz tops. thats with systems that had: q9550 liquid cooling 1333MHz ram or 1600Mhz ram. For people that are on a budget or people that don't know how to OC themselves, does the latency really come out to be worth $210 dollars? But its real good of you guys to put the benchmarks out, its great to see regardless of what any of us think. Edited by DST4ME - 12 Oct 2008 at 6:52am |
|
Alex
Admin Group Digital Storm Supervisor Joined: 04 Jun 2012 Online Status: Offline Posts: 16314 |
Quote Reply Posted: 11 Oct 2008 at 3:59am |
We will put it to the test.
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 4:56am |
can't wait to see how much of a difference latency makes in real life.
but are we OCing the processor or its all gonna be stock fsb? I also like to say again, my advise of matching fsb to ram is for people who a. don't oc. b. don't know how to oc. c. are on a budget. I've stated that clearly many times to venom and if you search my post you will see that anytime I advised to match the FSB to ram was because it was one or more of the above 3 situations. Edited by DST4ME - 11 Oct 2008 at 4:59am |
|
Daygn
Senior Member Joined: 10 Oct 2008 Online Status: Offline Posts: 448 |
Quote Reply Posted: 11 Oct 2008 at 5:07am |
Hey Alex you always seem to be agitated or sound mad. Dont be! You are a DS manager and you must know what you are talking about or you would not be, right. So dont get upset just post the benchmarks sit back and have a laugh on him. K Its all just fun and games
|
|
BB
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 5:14am |
Excuse me, but what do you mean by have a laugh on him?
If your read Alex's post right up there he says:
So already I'm right. I don't know if its a language barrier but "have a laugh on him" means, he should be laughing at me, which is not nice. Edited by DST4ME - 11 Oct 2008 at 5:15am |
|
Daygn
Senior Member Joined: 10 Oct 2008 Online Status: Offline Posts: 448 |
Quote Reply Posted: 11 Oct 2008 at 7:01am |
No no no not at him on him. All ment was have some fun and dont be so serious. Isnt that what you are doing, laughing and having a good old time talking to people in the fourms?? When Alex called me yesterday he seemed stressed and uptight all I ment was have a good time. Sorry if I offended you. I was not trying to put you down.
|
|
BB
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 7:05am |
I figured its a language barrier don't worry about it, you meant to say "laugh with him"
Yes well he does need to take it easy, to be honest from what I can tell, he is an easy going guy, you don't find lot of places that allow you to be so honest with customers, so he must be easy going. |
|
Daygn
Senior Member Joined: 10 Oct 2008 Online Status: Offline Posts: 448 |
Quote Reply Posted: 11 Oct 2008 at 7:25am |
Yah, but just seems to tense. He was cool on the phone but something was eating at him
|
|
BB
|
|
Nomak
Senior Member Joined: 15 Aug 2008 Online Status: Offline Posts: 442 |
Quote Reply Posted: 11 Oct 2008 at 8:00am |
He was tense because i gave him a call that same night, i told him i have high-resolution screenshots (HD quality) of said thread(s) in question, which, if im not paid a handsome amount for + given a generous discount on my rig to keep silent, will end up on BB/FTC hands.
Hope that clears up why he was tense.
In other news... If you have the money and plan to have any OC done on your CPU/RAM definitely go with the best you can afford. Thats what ive gotten out of the discussion yesterday and now proven (almost, waiting for benches) by Alex.
|
|
|
|
Alex
Admin Group Digital Storm Supervisor Joined: 04 Jun 2012 Online Status: Offline Posts: 16314 |
Quote Reply Posted: 11 Oct 2008 at 11:44am |
Im not mad, .
I guess I seem too serious at times then. .
We won't overclock the CPU DST4ME.
Warm Regards,
Alex
|
|
venom
Newbie Joined: 08 Oct 2008 Online Status: Offline Posts: 0 |
Quote Reply Posted: 11 Oct 2008 at 12:26pm |
I still don't understand why you would have someone match it if they don't know how to overclock. Doesn't matter how many times I read it, just doesn't make sense. |
|
TomD
Groupie Joined: 22 Jun 2008 Online Status: Offline Posts: 289 |
Quote Reply Posted: 11 Oct 2008 at 12:34pm |
Here is a recent article with several benchmarks comparing 2GB vs 4GB, DDR2 s. DDR3 and various frequencies of each memory type.
The "bottom line" - which is consistent with just about every article with actual benchmarking data I have found - is that increased frequency (synced or not) gives a measurable but small performance advantage that for "most" people isn't worth the cost.
Of course "worth" it is subjective, especially given how much costs have come down on DDR3 over the last few months.
Thank you Alex for putting together some benchmarks for us to see.
I Look forward to seeing Venom's data!
|
|
Nomak
Senior Member Joined: 15 Aug 2008 Online Status: Offline Posts: 442 |
Quote Reply Posted: 11 Oct 2008 at 1:14pm |
Well in his defense - dont DSO rigs come with a standard 3-year warranty? It cant be that hard to walk a complete hack thru basic settings in BIOS over the phone now can it? So, even IF someone doesnt know how to OC jack can still have DSO do it with a peace of mind (it would seem) that he/shell be taken care of should a problem arise.
Or am i wrong in my assumption, because you know what thery say about assuming things..
Edited by Nomak - 11 Oct 2008 at 1:29pm |
|
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 2:53pm |
I showed the the simple math then I stated why it should be matched.
I not sure how else to explain it. |
|
Nomak
Senior Member Joined: 15 Aug 2008 Online Status: Offline Posts: 442 |
Quote Reply Posted: 11 Oct 2008 at 3:01pm |
Nomak. Say it with me..
Nomak.Nomak.
PS: I dont sell big macs or anything. Edited by Nomak - 11 Oct 2008 at 3:02pm |
|
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 3:32pm |
Nomak, its a matter of performance vs money
|
|
venom
Newbie Joined: 08 Oct 2008 Online Status: Offline Posts: 0 |
Quote Reply Posted: 11 Oct 2008 at 9:46pm |
http://episteme.arstechnica.com/eve/forums/a/tpc/f/77909774/m/571005387831
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 10:03pm |
those charts are right along what I'm saying, minimal difference, but as alex said if the extra bit of gain is worth it to you then go for it. @venom, that link not only says exactly what I'm saying but it also agrees with what I'm saying and what Alex says at the beginning. also just some of the charts from TomD's link: so is the 1600MHz ram running at 1600MHz yes, is that speed begin carried thru the FSB if the FSB is at 1333MHz? no cause its capped at 10666 MB/s transfer rate, so what ever is above 10666 MB/s, is not going to transfer with the FSB any faster then 10666 MB/s transfer rate. You can't transfer data at 12800 MB/s when the limit fsb can handle is capped at 10666 MB/s. The difference as I said earlier is in latency. Can't wait till monday to see how DS's benchmarks turn out Edited by DST4ME - 12 Oct 2008 at 4:04am |
|
SunfighterLC
DS Veteran Joined: 18 Feb 2008 Online Status: Offline Posts: 1527 |
Quote Reply Posted: 11 Oct 2008 at 10:31pm |
I dont see the use in those charts honestly, 3d mark and vantage are more for measuring the GPU and to a less extent the CPU. There are no memory performance tests.
|
|
E8500@ 4.03Ghz
XFX 790i Ultra 1000W Corsair HX 2 280 GTX EVGA FTW 4GB OCZ Reaper 1800Mhz 250-80-300GB VR HD Logitech Z-2300 2.1 Speakers Asus Xonar 7.1 Hanns-G HG 281D 28" HDMI Monitor |
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 11 Oct 2008 at 10:39pm |
fair enough, how about these from the same link? hope they help, these are actuall apps and games.
again these are from TomD's link. Edited by DST4ME - 11 Oct 2008 at 11:47pm |
|
Nomak
Senior Member Joined: 15 Aug 2008 Online Status: Offline Posts: 442 |
Quote Reply Posted: 12 Oct 2008 at 8:19am |
Wheres 1800 and 2000 models?
Bleh!
|
|
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 12 Oct 2008 at 5:01pm |
this shows you how little differnce there is between 1600 and 1333.
I think DS will run the 2000 on monday. but to behonest this is all pointless, once nehalem comes out, this whole fsb business is out the window, cause it won't have one to limit it. |
|
TomD
Groupie Joined: 22 Jun 2008 Online Status: Offline Posts: 289 |
Quote Reply Posted: 12 Oct 2008 at 7:38pm |
DST4ME,
I don't see how the article supports your opinion that Syncing the RAM 1:1 is an important thing to do.
It demonstrates exactly the opposite - that there is a performance improvement with faster RAM regardless if it is set 1:1 or not.
The "math" you present would seem to make sense on the surface, but you aren't thinking about the impact of the CPU cache where the CPU gets most of its data from, and a host of other factors that come into play. The data doesn't neatly flow from RAM to CPU in a 1:1 ratio with FSB speed, its a much more complicated process.
I agree with you its a very minor improvement, but the topic of the thread is syncing RAM 1:1, not cost-benefit.
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 12 Oct 2008 at 7:50pm |
I'm talking about matching the ram speed to FSB.
in short if the FSB is 1333MHz, its max transfer rate is 10666MB/s. if your ram is 1600MHz, it will have a 12800MB/s transfer rate. you can't transfer data at 12800MB/s when the FSB can only to a max of 10666MB/s. yes the ram is running at 1600MHz but FSB will bottleneck it, the only place the extra 2134MB/s comes into play is in latency. which I'm saying is not worth the money. all I'm saying is that if you have a FSB of 1333MHz the ram you choose should be 1333MHz. the discussing here was that some DS staff are telling customers to go with a 1600MHz ram when they have a FSB of 1333MHz, I was saying that the customer should not do that, and instead should go with 1333MHz. thats all. I was having this discussion in another thread, that was closed this one was opened. what I'm saying is that, that article also says to go with a 1600MHz ram with a 1333MHz FSB, is not really cost effective and the gains are very minimal, which is what I have been saying. |
|
TomD
Groupie Joined: 22 Jun 2008 Online Status: Offline Posts: 289 |
Quote Reply Posted: 12 Oct 2008 at 8:05pm |
DST4ME wrote:
"all I'm saying is that if you have a FSB of 1333MHz the ram you choose should be 1333MHz."
Why wouldn't I choose 1066 and save the money? The performance reduction is very small and the 1333 isn't worth the extra money.
You are confusing Syncing at 1:1 with cost effectiveness. It is a different issue.
That is all I am saying.
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 12 Oct 2008 at 8:13pm |
like I said, the original discussion was/is about DS tech telling to customers to go with 1600 vs the 1333.
that is the discussion, I'm just trying to explain to come down from 1600 to 1333. I'm not saying there is anything wrong with 1066, me and you are both on the same page and saying the same thing. you missed the original discussion, that is why you think I'm confused I'm not we are talking about recommending customers to get what speed ram, that involves price/performance. Just like you I'm trying to save customers money by saying not to go with 1600, some DS techs are saying no 1600 is better. so we are waiting on DS's charts which I don't see how they are gonna be any different then the ones you linked to. your link/charts, prove both of us are right, money vs performance, and a customer does not need to go 1600MHz ram if their fsb is 1333, so we are both on the same page even tho the topic has that name, the real discussion here is, if fsb is 1333 should customer go 1600 or lower ram say 1333 is fine. me and you are saying 1333 or lower, some others were saying 1600. Edited by DST4ME - 12 Oct 2008 at 8:19pm |
|
TomD
Groupie Joined: 22 Jun 2008 Online Status: Offline Posts: 289 |
Quote Reply Posted: 12 Oct 2008 at 8:52pm |
Agreed, we are saying the same thing. I tend to split hairs.
I just get annoyed by the common [mis]conception that syncing RAM at 1:1 has some magic benefit, when the benchmarks I have seen show it doesn't really matter.
Now how do I explain why I bought the 1600 RAM in my sig?
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 12 Oct 2008 at 9:03pm |
LOL agreed, well maybe you have the money and you wanted the most you could get, nothing wrong with that.
|
|
Alex
Admin Group Digital Storm Supervisor Joined: 04 Jun 2012 Online Status: Offline Posts: 16314 |
Quote Reply Posted: 12 Oct 2008 at 10:17pm |
DST4ME,
Let me ask, if you were spending $5,000 on a system, would you get 2GB of 1066MHz memory or 4GB of 2000MHz memory?
From the looks of the charts, it shows no performance difference...
Yet, I honestly seen different results first-hand on the systems we build here.
We will post some real benchmarks soon enough.
Cheers,
Alex
|
|
SunfighterLC
DS Veteran Joined: 18 Feb 2008 Online Status: Offline Posts: 1527 |
Quote Reply Posted: 12 Oct 2008 at 11:12pm |
Alex-- I think the real problem is there isnt any real good memory performance testing programs. So people use tests and ways of testing that arnt tuned for testing memory performance.
I honestly cant even say myself whether faster ram is better. I went from 2GB of 1333 to 4GB of 1600, so the speed and increase of performance could of just been from going from 2 to 4GB. But I have noticed a moderate to high increase in responce from my computer in a lot of actions.
|
|
E8500@ 4.03Ghz
XFX 790i Ultra 1000W Corsair HX 2 280 GTX EVGA FTW 4GB OCZ Reaper 1800Mhz 250-80-300GB VR HD Logitech Z-2300 2.1 Speakers Asus Xonar 7.1 Hanns-G HG 281D 28" HDMI Monitor |
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 12 Oct 2008 at 11:17pm |
Alex let me answer that question like this: if I was going to purchage a system today this would be it: here you go Ticket# 194182 --- Price: $5575 (To see this build click here) Copy of Specifications: Chassis: Cooler Master HAF 932 Exterior Finish: - Standard Factory Finish Interior Finish: - Standard Factory Finish Power Supply: 1200W Thermaltake (Triple and Quad SLI Compatible) (Toughpower Edition) Processor: Intel Core 2 Quad Extreme QX9770 3.20GHz (1600MHz FSB) (12MB Cache) Motherboard: nVidia 790i Ultra Core 2 Quad (nForce 790i Ultra SLI) (DDR3 Only) System Memory: 4GB DDR3 1600MHz OCZ Platinum Edition Card Reader: Internal Digital Media Card Reader (Black) Hard Drive 1: 300GB Western Digital VelociRaptor (10K RPM) (16MB Cache) (Extreme Speed) Hard Drive 2: 500GB Western Digital (16MB Cache) (7200 RPM) (SATA) Raid Option: - No Thanks Hard Drive 3: - No Thanks Optical Drive 1: DVD±R/RW/CD-R/RW (DVD Writer 20x / CD-Writer 48x) (LightScribe Edition) Optical Drive 2: - No Thanks Wireless Access: High Speed Network Port (Supports High-Speed Cable / DSL / Network Connections) Video Card: 3x SLI Triple (NVIDIA GeForce GTX 280 1GB) (Includes PhysX Technology) TV Tuner: - No Thanks Sound Card: Razer Barracuda AC-1 (Recommended) Extreme Cooling: Liquid Chilled FrostBite CPU Only Internal Lighting: Blizzard Internal Lighting (Don't know what color yet Edition) (Cold Cathode Tubes) Modifications: - No Thanks Boost Processor: Yes, Overclock the processor as much as possible with complete stability Boost Video Card: - No Thanks, Please do not overclock my video card(s) Boost Memory: Yes, Overclock memory timings (Includes Memory Fan Kit Installed) Tweak Windows: - No Thanks, Please do not tweak the services on the operating system Windows OS: Microsoft Windows Vista Home Premium (64-Bit Edition) (For Enthusiasts) Restore Kit: Digital Storm Specialized Recovery System (DVD Image Based) Protection: - No Thanks Office: - No Thanks Benchmarking: - No Thanks Install/Test Game: - No Thanks LCD Display: - No Thanks Surge Shield: - No Thanks Speakers: - No Thanks Keyboard: - No Thanks Mouse: - No Thanks External Storage: - No Thanks Warranty: 5 Year Platinum Care Extended Parts & Labor Warranty (Highly Recommended) it would be that or skulltrail with 2 x qx9770 and the rest would look like the above. Before anybody says,"you spend that much on the qx9770 (price vs performance)", I like to say that is a matter of personal preference, I have had better experiences with the extreme chips and I don't use anything but extreme chips, so that decision is not so much performance vs cost, its more of a personal preference, if I was on a budget, it would be the q9650. I can boost my ram to 2000 if I need it, why else would I oc my ram for? still contemplating LC all 3 cards or not. I can't justify the 2000MHz in this case, I can bleed that right out easily myself. I have to agree with SunfighterLC, he didn't see no speed increase, as I never have noticed it going up in ram speed either, instead since he added 2 more GB now his PC can multitask much better. I personally have never noticed a difference in PC performance when I moved up in ram speed, did the charts show better numbers? yes did I notice a difference? just like SLC, no. in very few cases higher ram was need to have a stable OC, but those are isolated and the ram was low to begin with, lol sometime I try to bleed out things that are not there to me looking at the charts and numbers, it looks like once you go over the fsb limit, ram speed is more of a preference then anything else. At the same time if you say you saw different results, then I know you are not lying, but is it more of a charts say there is gain but in real life it does not amount to much scenario? cause just like SLC, when I went up in ram speed, numbers were higher but my experience didn't change a bit, I mean I can't honestly say one thing was running any faster, folding took the same amount of time, encoding, raring, you name it, I checked it, you should have seem me, I was like a nut, testing every software I had on the PC to see if there was any difference. Unless ofcourse you are saying you saw real life difference, eg., FAH folded faster, or encoding or any apps/games. then I'm really interested to find out what it was that I have missed looking forward to seeing you guy's test. Edited by DST4ME - 13 Oct 2008 at 2:18am |
|
Chris W.
Newbie Digital Storm Tech Support Joined: 05 Jan 2009 Online Status: Offline Posts: 149 |
Quote Reply Posted: 13 Oct 2008 at 12:52am |
TMI Harley T M I. lol. And while I typically advise when asked for a budget build that syncing the memory 1:1 is better bang for your buck, if its in their budget its good to bump it up to that faster memory. |
|
<img src="http://www2.razerzone.com/tag-generator/label_images/1222492484_2122770270.jpg" border="0" />
|
|
Nomak
Senior Member Joined: 15 Aug 2008 Online Status: Offline Posts: 442 |
Quote Reply Posted: 13 Oct 2008 at 8:12am |
Im waiting on Venoms data, cant wait. |
|
|
|
DST4ME
DS ELITE Joined: 14 Apr 2008 Online Status: Offline Posts: 36758 |
Quote Reply Posted: 13 Oct 2008 at 8:49am |
I'm waiting too, I like to see what it says. I think we are all in agreement that higher ram gain is not worth the price, now we have to see if the DS's charts say the samething. |
|
Post Reply | Page 12> |
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You can vote in polls in this forum |