pc3200 tops out at 190

[logos]

got some of that 'best kick ass deal you'll ever see' pc 3200 and it is pc3200 NOT! searched around and looked like people were having luck with it but i guess i got shorted. supposed to be cas 2.5 but only does 2.5 up to 146 then it pooped completely at 190. have to start looking for something better. cpu tops out at 196 anyway but eventually should like to get over 200.

[dustybyrd]

what brand is the pc3200, logos?

i got some hyperx pc3000 from newegg that seems fine cas 2 up to 158 fsb (can't test it higher)....but sisoft sandra reports it as pc2700 cas 2...so maybe kingston shorted me a bit...or maybe it's pc2700 that can do pc3000 or 180mhz fsb

[larva]

There is no PC3000 spec to report, and for the longest time there was no PC3200 either. All a memory module is is the semiconductor and their supporting circuitry, plus the SPD EEPROM. So when a (generally bogus) program like Sandra reports something about the ram, all it can do is base it on the chips on the module. Anything can be programmed into the SPD, so in reality basing decisons on the chips present is a more valid approach.

Since there was never a PC3000 spec, there obviously aren't any chips that correspond to this designation. The chips on the PC3000 HyperX are most assuradly intended for PC2700 operation, but as we know this isn't cast in stone. Likely they are BH6 Winbond, PC2700 chips that are renowned for their ability to run far in excess of 333MHz.

These are the same chips used in the Corsair PC3200C2 ram, which was the best avaiable for a long time. Just like it isn't fair to call the Corsair 3200 PC2700 because it bears chips that technically are, it isn't fair to blame Kingston because their PC3000 ram carries the same chip and therefore get reported by Sandra as a PC2700 module.

Oh, and BTW, all JEDEC compliant PC2700 ram is CAS 2.5 (or higher), not 2. The standard excuse for programs like Sandra saying PC2700 for your PC3000 module is that there is no PC3000 standard, but of course there is no PC2700 cas 2 standard either...

[dustybyrd]

Oh, and BTW, all JEDEC compliant PC2700 ram is CAS 2.5 (or higher), not 2. The standard excuse for programs like Sandra saying PC2700 for your PC3000 module is that there is no PC3000 standard, but of course there is no PC2700 cas 2 standard either... [/B]


what do you mean? that my hyperx pc3000 is not cas 2? that is what newegg advertises it as...and on my dual board below it does 158 cas 2...(but i have limited cas settings on this board...so it might be cas 2.5?)

also...what chips are likely on the hyperx pc2700?

[larva]

No, what I mean is that official standards like the PC2700 spec do not allow for cas2 operation. That does not mean your ram won't do it. My whole point was it is useless to focus on what Kingston calls the ram, what Sandra calls the ram, or what anyone or anything else calls the ram. It does what it does, and as you have seen cas2 is workable.

I don't know what the PC2700 HyperX has, but I don't think its BH6 winbond. The reports I have seen for this ram are poor, well beneath the level of any quality BH6 bearing module.

My value ram is a BH6 variant, and it does 470+MHz with ease. Suffice it to say the limits for your setup are the motherboards, there is no point to scratching your head ram-wise.

[timmyqwest]

I'm going to follow all of your posts with this line...

I dont know why your not a Sr. member yet

[Splinter]

It must be the curlers! :) Just kidding. Larva, you definitely know your stuff.

[josi]

Originally posted by larva
There is no PC3000 spec to report, and for the longest time there was no PC3200 either. All a memory module is is the semiconductor and their supporting circuitry, plus the SPD EEPROM. So when a (generally bogus) program like Sandra reports something about the ram, all it can do is base it on the chips on the module. Anything can be programmed into the SPD, so in reality basing decisons on the chips present is a more valid approach.

There is no way for Sandra to read out something off the chips. All decisions are based only on info stored in the SPD.

Oh, and BTW, all JEDEC compliant PC2700 ram is CAS 2.5 (or higher), not 2. The standard excuse for programs like Sandra saying PC2700 for your PC3000 module is that there is no PC3000 standard, but of course there is no PC2700 cas 2 standard either...

Are you sure? Latest JEDEC standard for DDR SDRAM - JESD79C states that:
"For DDR200, DDR266, and DDR333, the latency can be set to 2 or 2.5 clocks (latencies of 1.5 or 3 are optional, and one or both of these optional latencies might be supported by some vendors). For DDR400, the latency can be set to 3 clocks (latencies of 2 or 2.5 are optional, and one or both of these optional latencies might be supported by some vendors".

[larva]

I think you are correct that any information pulled comes from the SPD. I have seen Sandra report a specific chip type, apparantly this information was programmed into the SPD in the module description.

I rarely see a SPD programmed correctly, or at least rarely see a motherboard that can read the SPD correctly. I tend to distrust anything in the SPD. I gave the impression that Sandra might read the chip type from the chips themselves, and this is misleading at best. I don't really care what the manufacturer programs in the SPD, being that it is often wrong or at least read wrong by the system. I find it far more revealing to know which chip is being used, and I think it would be prudent to ascertain this with one's eyes rather than place any stock in what Sandra might report.

I reviewed the JEDEC specification, and you are correct in that CL2 is defined for PC2700 devices. I was told by Micron that it was not, but I see now that this was just an excuse for the sorry performance of their PC2700 ram. That's what you get for believing what you are told, expecially when the means exist to ascertain the truth for yourself.

Interesting to note that there are multiple levels of the JEDEC specs. It makes it clear that DDR266A and B refer to CL2 and 2.5 operation at 266MHz (at least rev1 of the spec does), but nowhere does it specifically define what DDR333B would be. The current consumption values listed for DDR333 are tested as DDR333B, with CL2.5. As not all functional parameters for DDR333 are tested or listed at CL2, it is fairly clear CL2.5 is the common interpretation of the PC2700 spec.

Also noted in the document is that the optional CL of 1.5 is only defined for DDR200 devices. CL2 and 2.5 are listed as optional modes for DDR400, but as with DDR333 testing and several functional parameters for DDR400 are CL3 values. Again it is clear that CL3 is the focus of the DDR400 spec, although there are DDR400A, B, and C sub-levels to the spec that might clear up confustion of this nature should memory module manufacturers elect to use them. CL2 for DDR 400 falls into the realm of "optional" secifications, and is not covered by the A, B, or C sub-levels of the specification.

[josi]

Originally posted by larva
I rarely see a SPD programmed correctly, or at least rarely see a motherboard that can read the SPD correctly. I tend to distrust anything in the SPD. I gave the impression that Sandra might read the chip type from the chips themselves, and this is misleading at best.

It's a pleasure to read your posts, larva. To be exact, almost every SPD is programmed correctly by the manufacturer, only exception I know is Corsair low-latency memories.
There are at least three ways how any system information reporting program can acquire the actual memory information:
1) via BIOS data (DMI configuration, memory information mostly based on the SPD data);
2) - via direct reading from SPD through the SMBus (again SPD data)
3) - via chipset specific register data (mostly based on the BIOS memory settings)
As you see, none of the methods has anything common with the physical memory chips.

I don't really care what the manufacturer programs in the SPD, being that it is often wrong or at least read wrong by the system. I find it far more revealing to know which chip is being used, and I think it would be prudent to ascertain this with one's eyes rather than place any stock in what Sandra might report.

Yes, but poor PCB quality can change excellent memory chip timings to nothing.

[larva]

Originally posted by josi


As you see, none of the methods has anything common with the physical memory chips.


Agreed, that is the primary reason that I wouldn't get worried about what Sandra might report.


Originally posted by josi


Yes, but poor PCB quality can change excellent memory chip timings to nothing.

I agree the PCB is a functional part of the device, and therefore can affect the results. It is rare to find a PCB that does not perform up to snuff though. Those makers that charge 50-100% above market value (*cough* Corsair, Mushkin, OCZ, et al) have a financial interest in playing up this factor to justify their outrageous prices.

Buying from good makers like Crucial or Kingston does insure a competent PCB. The only problem with Crucial is the chips they use (Micron) are junk, and the sanely-priced Kingston modules can carry any chip. Kingston has wisened up to the fact that they can slap a heat spreader on the modules that contain the better performing chips and mark them up to Corsair prices.

In the end testing a particular module in your equipment is the only true way to tell exactly what results will result. But of the visible factors we can use to select likely candidates the chips carried are the overwhelmingly most important. Yes, PCB quality is also a concern, but this is not nearly the problem that some would have you believe.

It is hard to tell how good the programming of the SPD is, as motherboards so often mis-read the information. Often they will set the timings to full-fast or full-slow across the board when you know that is not the information contained in the SPD. It presently escapes where the article was published, but a recent ram test by one of the better sites had to swap back and forth between Asus and Intel motherboards to read the SPD data as no one board would both offer the needed features and read the SPD data correctly.

[floben]

That sucks. I can run at over 190 mhz with my crucial PC-2700 @ 2.7 volts...