About ram.....mostly DDR ram

Pearl Jammzz

Established Member
-What are all the different kinds of DDR ram?

-What do the numbers mean like the PC blah blah blah stuff? I believe that only certain mobos can support certain numbers but what do the numbers mean?

-What are sum good types of ram?

-What are the differences in the heatsinks that are around the ram?

-Do the heatsinks do all that much?

Any additional knowledge would be great too :). Thanks for sharin yer brains guys, lata.
 
The different kinds of DDR RAM you'll probably be interested in are DDR PC 2700 and DDR PC 3200. DDR 3200 is a little bit faster and better than 2700 and for an extra ten bucks or so I'd say it was worth it.

The numbers thing like i said before just mean the ram is a little bit better. You'll have to check on your mobo to see what kind and how much of a certain type of ram it supports.

You cant really go wrong with the brand of ram you buy as long as it isn't some generic no name brand. The one's i've heard are good are Geil (what I have), Kingston, Corsair and some others. Just make sure you dont get generic ram.

The heatspreaders are all pretty much the same and if your ram doesnt come with one you can buy one for like 1.50.

They spread the heat so it doesnt get hot and fry. I mean you might as well have one to be on the safe side.
 
I don't know I doubt they actually use gold but copper seems to be the best for the transfer of heat. It doesnt really matter.
 
Remember that PC3200 might be "faster" but that doesn't mean his board is going to take any advantage of it. This is why manuals are your friend. But a chipset or motherboard brand+model would be enough to find out what it can use. With VIA-based boards, its better to use the fastest memory your board supports. For Nforce boards, you generally are going to want to run your memory synchronously with the FSB, but you can always clock PC3200 at PC2700 speeds if you (for instance) planned to later buy a CPU supporting a faster FSB.

Ramsinks are not important at all if you are not going to overclock. A hardcore overclocker (especially one with money) would probably aim for memory with a copper heatsink and very agressive (faster than normal) memory timings. For most people, your standard CL 2.5 memory is fine, as long as you don't buy a noname module, something with a warranty is a must.

I might also note that in the case of an Nforce2 board, you could always raise your FSB speed to match your faster memory anyway, without having to worry about PCI/AGP speeds being driven dangerously high. But again, that is overclocking territory... on a side note, the Nforce3 is a disappointment. I was expecting excellent performance, but for Athlon64 chips, the VIA chip is currently a better performer.
 
-What are all the different kinds of DDR ram?

There aren't really different kinds of DDR so much as there are different configurations of memory in general - buffered vs. unbuffered, ECC vs. non-ECC, and so on. Most consumer platforms take non-ECC unbuffered memory. "ECC" means that single-bit errors in the memory can be fixed on-the-fly by the chipset (this isn't used in most consumer systems because it's slower and performance is more highly valued than reliability). Buffered vs. unbuffered is an electrical issue that I'm not prepared to explain.

-What do the numbers mean like the PC blah blah blah stuff? I believe that only certain mobos can support certain numbers but what do the numbers mean?

The numbers are a sort of twisted marketing version of a clockspeed rating (the actual number is theoretical maximum bandwidth in MB/second at rated clockspeed, not taking things like addressing overhead/refresh/precharge timing into account). This is really just like your processor - the memory runs at whatever speed the chipset flips the control signals at, and the number just tells you how fast the part is supposed to be able to reliably run:

PC1600: 100MHz (DDR200)

PC2100: 133MHz (DDR266)

PC2400: 150MHz (DDR300)

PC2700: 166MHz (DDR333)

PC3000: 183MHz (DDR366) (I don't know if this ever caught on)

PC3200: 200MHz (DDR400)

Vendors tend to double the clock speed (i.e. use the DDR data rate) when talking about products, even though it's not really accurate to say that the memory interface runs at twice the speed of the clock. Different chipsets have different options for memory speeds, but most are optimized for a particular speed; most of VIA's chipsets for example are named based on the optimum speed.
 
PC3500: 216MHz (DDR433)

PC3700: 233MHz (DDR466)

PC4000: 250MHz (DDR500)

PC4200: 266MHz (DDR533)

Also available :D
 
And PC4500: 300MHZ (DDR600)

The thing to keep in mind is that when you got some memory that is for example PC3200, it will be able to reliably run UP TO that speed. So using this kind of memory for slower FSB speeds has no negative effects. You just have "wasted" potiential speed sitting in the memory not doing anything. Generally you shouldn't get anything lower than PC3200 since that will provide the best value if you plan on upgrading the CPU to a faster FSB and won't need to buy new ram for it.

BTW, I'm running some ultra high quality PC3700 ram at PC3200 speed. With this ram I'm managing (according to SiSoft Sandra) to get bandwidth speeds of about 2900MB/s for integers which is about 90% efficiency (of the potential 3200MB/s theoretical maximum, also where they get the 3200 rating from). Getting anything above this much is generally going to cost alot of money and be very difficult to achieve.
 
GB is right as usual (except that DDR600 would be PC4800, not PC4500). I'd like to add that at the moment, anything over PC3200 is probably only of interest to overclockers, and anything over PC4000 is probably an overzealous vendor making shit up. Don't count on future-proofing your memory at this stage of the game either - DRAM manufacturers generally want to move to a new DDR architecture (tentatively dubbed "DDR II") that has a deeper pipeline and a faster interface, designed to allow for higher transfer rates with slower (i.e. easier/cheaper to make, lower power consumption, more reliable operation etc.) DRAM cells, rather than grapple with the unfriendly physical issues of running a wide 300+MHz bus with multiple socketed devices.

Edit: reading this again, it occurs to me that with DDR II, that's exactly what they propose to do. What I meant to say is that they don't want to deal with making the actual memory keep up with that kind of interface.
 
Originally posted by gameboy900@Oct 1, 2003 @ 02:03 AM

The thing to keep in mind is that when you got some memory that is for example PC3200, it will be able to reliably run UP TO that speed. So using this kind of memory for slower FSB speeds has no negative effects.

I mentioned that, but nobody read it apparently :(

Gallstaff: Sure. You can OC that fine. But an Nforce board is going to let you really push things, again, without having to worry about PCI/AGP ratios. They are locked at 33/66, respectively.
 
Yup I got my Asus A7N8X Deluxe (nForce 2) running my Barton 2500+ at 200MHz FSB instead of it's normal 166MHz. This has effectively turned it into a 3200+ chip and with hardly any extra cooling effort too. If I had a rev2.0 board I could probably push it by another 5-10MHz but my 1.04 board gets flaky after 200MHz. Then again I'm not complaining.
 
No, FSB goes up to 200MHz. DDR ram is clocked at twice that. So for 166MHz FSB you get 333MHz DDR ram speed. 200MHz FSB would give you 400 MHz DDR ram speed.

Of course if you're overclocking then the FSB would be other values and then the ram frequency would be double that.
 
Originally posted by gameboy900@Oct 3, 2003 @ 02:43 AM

No, FSB goes up to 200MHz. DDR ram is clocked at twice that. So for 166MHz FSB you get 333MHz DDR ram speed. 200MHz FSB would give you 400 MHz DDR ram speed.

Of course if you're overclocking then the FSB would be other values and then the ram frequency would be double that.

No, its not clocked at twice the speed. DDR 333 and a FSB of 333 are STILL clocked at 166Mhz. That is their clock speed. However you are correct that their effective speed is double that. You can thank marketing for labeling FSBs at 333, 400, 533, 800Mhz, etc. I guess on the surface it makes things easier, but when you talk about actual clock, like what you would see in the BIOS, it makes things confusing.

I agree its really sweet to be able to just raise the FSB speed and keep the multiplier the same. This guy in my CSC class has a dual athlon xp system (modded to MP setting), with a watercooled system that has dual radiators running in parallel, one for each CPU. He runs the two 1800+ chips at around 2.3 Ghz (200 FSB), which leads to a pretty sick performer. Especially for encoding. Of course he has money, and I don't, hahaha...
 
Umm....AMD uses the real FSB speed (100, 133, 166 or 200) for it's CPU's. Intel on the other hand does that weird multiply by 4 crap.

Mainly because 400x2=800 (see HyperThreading makes it two CPU's *cough*) and 400 = 200x2 but the 400 is what the ram runs at since it uses DDR. So the real FSB speed is still just 200.
 
What?? Hyperthreading has nothing to do with Intel's quad-pumped FSB. All P4s have used that technology, the original "400Mhz FSB" P4s had a 100Mhz FSB clock. What I was saying was simply that when you said that DDR was "clocked at twice that" is that it ISN'T clocked at twice that. It works much like AMDs FSB, "DDR400" and a "400Mhz" AMD FSB are both still clocked at 200Mhz. That is the clock speed they run at. Like I said, blame marketing.

Just so you know, AMD markets their CPUs by their effective rate just like Intel does. That's why you hear about 333/400Mhz Bartons, etc. The *reason* AMD multiplies by 2 and Intel by 4 is because they are using different technology. AMD's FSB (MUCH LIKE DDR) transfers data twice per *clock*. So it has an effective speed of double its actual clock speed. But the clock speed is still going to be 166/200/whatever Mhz. Intel's solution transfers 4 times per clock, so a 200Mhz FSB gives an effective speed of 800Mhz... it is still clocked at 200Mhz.
 
Back
Top