DDR DIMMs sold already overclocked?

[Boris Gjenero]

I have two PC3200/DDR400 DIMMs which according to the SPD should work
at CL2.5-3-3-8 at 200 MHz and CL2-3-3-7 at 166 MHz. The DIMMs have
Samsung K4H560838D-TCC4 chips. I found the datasheets at:
http://www.samsung.com/Products/Sem...AMcomponent/256Mbit/K4H560838D/K4H560838D.htm
They specify that the chips work at CL3-4-4 at 200 MHz. This means
the SPD data on the DIMM causes the chips to operate above their rated
speed. Is this sort of thing normal nowdays?

 

[George Macdonald]

I have two PC3200/DDR400 DIMMs which according to the SPD should work
at CL2.5-3-3-8 at 200 MHz and CL2-3-3-7 at 166 MHz. The DIMMs have
Samsung K4H560838D-TCC4 chips. I found the datasheets at:
http://www.samsung.com/Products/Sem...AMcomponent/256Mbit/K4H560838D/K4H560838D.htm
They specify that the chips work at CL3-4-4 at 200 MHz. This means
the SPD data on the DIMM causes the chips to operate above their rated
speed. Is this sort of thing normal nowdays?

Answer: I think so. Who makes the DIMMs? What happens is that based on
some sampling frequency, Samsung bins the chips. A smallish DIMM mfr then
tests every single chip and pushes the operating margin to see what they
can get out of them at the limit. They get some which exceed Samsung's
specs so they stick them on a DIMM and, if you're lucky check every DIMM
for integrity in a number of different mbrds with different chipsets/memory
controllers.

In the case above, they got some chips from a batch which Samsung found to
be just slightly outside the TCCC spec so they got binned as TCC4; the DIMM
mfr found them good enough for TCCC, possibly by pushing the voltage to
2.75V instead of Samsung's quoted 2.6V, and pushed the CL to see what they
could get there. Do they work OK?

 

[Franc Zabkar]

I have two PC3200/DDR400 DIMMs which according to the SPD should work
at CL2.5-3-3-8 at 200 MHz and CL2-3-3-7 at 166 MHz. The DIMMs have
Samsung K4H560838D-TCC4 chips. I found the datasheets at:
http://www.samsung.com/Products/Sem...AMcomponent/256Mbit/K4H560838D/K4H560838D.htm
They specify that the chips work at CL3-4-4 at 200 MHz. This means
the SPD data on the DIMM causes the chips to operate above their rated
speed. Is this sort of thing normal nowdays?

The PC industry is awash with fakes of all sorts. There are, or have
been, remarked CPUs, plastic "placebo" cache SRAMs (silicon-less
fakes), fake power transistors, etc. I have a "400W" ATX PSU which is
really only good for about 170W, according to the ratings of its
diodes. Then there is that other kind of fake where an up-market
vendor/manufacturer rebadges a down-market product and extracts a
price premium by leveraging his reputation rather than his technical
expertise.

/end of rant


- Franc Zabkar

 

[Boris Gjenero]

Answer: I think so. Who makes the DIMMs? What happens is that based on
some sampling frequency, Samsung bins the chips. A smallish DIMM mfr then
tests every single chip and pushes the operating margin to see what they
can get out of them at the limit. They get some which exceed Samsung's
specs so they stick them on a DIMM and, if you're lucky check every DIMM
for integrity in a number of different mbrds with different chipsets/memory
controllers.

In the case above, they got some chips from a batch which Samsung found to
be just slightly outside the TCCC spec so they got binned as TCC4; the DIMM
mfr found them good enough for TCCC, possibly by pushing the voltage to
2.75V instead of Samsung's quoted 2.6V, and pushed the CL to see what they
could get there. Do they work OK?

I don't know who makes the DIMMs. They do however use the supposedly
really good BrainPower 808 PCB.

No, they don't work okay. Prime95 fails after a few hours at the SPD
timings.

I guess I should have known the price was too good to be true :|
I found a way to get them for even less now, $50 Canadian until
April 5th:
http://www.ncix.com/products/index.php?sku=9545&promoid=1048&affiliateid=472595

 

[George Macdonald]

I don't know who makes the DIMMs. They do however use the supposedly
really good BrainPower 808 PCB.

From the labeling, Samsung's name but no logo, the normal assumption would
be that the DIMM was produced by Samsung for the OEM market; then again
that might be what some shady outfit wants the buyer to think.:-(

No, they don't work okay. Prime95 fails after a few hours at the SPD
timings.

Repeatable? If you back off to TCC4 timings, does it work better? I don't
think I've run Prime95 for that long myself.