Tom's Ultimate RAM Speed Tests - whats wrong?

[dragonsprayer]

Did i miss something or i did you guys just compate ddr1600 at 1333mhz to super expensive ddr2 and crank down the latencies too 3-3-3-8 at 800mhz and compare that to ddr3 at 1333?

where is the 1600fsb test?
where is the 1700mhz?
or the 1800?

everyone knows you must crank ddr3 up to 1600-1800mhz to get the memory bandwidth boost

did i miss something?


added 4/12 quote from andantech on the 790i chipset article was released last week:

"DDR3 prices have also made the idea of moving to either X48 or 790i a whole lot less painful than it would have been just a few short months ago. Great clocking and relatively inexpensive 2GB kits of DDR3 are starting to show up at all our favorite online retailers. We were able to find 2GB kits like the Patriot Viper rated at DDR3-1600 with 7-7-7 timings for under $200 without too much trouble. To date, all of our experiences with DDR3 memory have shown the DDR3-1600 kits to be incredible performers when it comes to overclocking"

this quote is directly the opposite of THG results - this quote is also the real results

Anandtech: "If you're still buying DDR2 and you're serious about overclocking, you're buying yesterday's technology. That might be fine for the price-conscious enthusiasts, but if you're considering the purchase of a $300+ motherboard and $1000 CPU we have to strongly recommend making the move to DDR3. "

ddr1600 pushed to 1700-1800 results in a 15% increase in 3dmark06 scores with a quad crossfire 3870, 3.6ghz q6600 system from 19000 to 22000, corsponding increases in apps should be expected.

 

[dragonsprayer]

Ok everyone, this is very misleading article lets make sure we say:

1st) THG for this test used very low end ddr3, you used $250 ddr3 (i own this ram and good supertalent and better patriot) and compared it to $350 ddr2!

corsair ddr2 is $519 http://www.newegg.com/Product/Product.aspx?Item=N82E16820145044
patriot ddr3 is $219 http://www.newegg.com/Product/Product.aspx?Item=N82E16820220259&Tpk=Patriot%2bPDC32G1600LLK

I just tested the patriot ddr3 you used and got 15% lower scores in 3dmark06 compared to better ddr3 from supertalent and patriot, your testing and test conclusions are false and misleading.

2nd) you did not even run the ram at 1600 7-7-7 timings???????

3rd) you state that ddr2 is 1/5 the cost, but you use ddr2 that is almost twice the cost of the ddr3

this is one the worst articles on THG - worse then the old amd bias

4th) the article states that you can get ddr2 for 1/5th the cost of ddr3 but you do not make it clear that you comparing the most expensive and high end ddr2 to mid to low level ddr3 1600mhz!

super talent ( http://www.newegg.com/Product/Product.aspx?Item=N82E16820609237) runs 7-7-7 easy at 100% stable at fsb1700 for around half the price as the super expensive corsair ddr3:




whats the point of comparing $500 ddr2 to $219 ddr3? i see only one, to provide misleading conclusions?

THG tests on ddr3 show you need fsb1600+ benfit from ddr3 bandwidth, but you guys only posted 1333mhz tests?

did i miss something?

 

[computertech82]

key word you left out of the ddr2 memory.....Out of Stock . REAL price is around $169

 

[Shadow703793]

Imo the point of this article was comparing the effect of DDR2/DDR3 RAM on a P4 and a C2D.

 

[cisco]

I came away from the article feeling pretty good about my DDR2 1066mhz 5.5.5.18, especially since it was 4gigs for $99.00. This seems to be the issue every time a new standard of DDR is released. When DDR went to DDR2 we were seeing the same benchmarks. AMD really likes tight timings because of the integrated memory controller, even Intel likes tight timings to an extent. Once Nehalem is released Intel will be in the same boat as AMD, tight timings will generally give better performance over higher clock speeds. I remember when DDR2 was released, they said it was going to be a short stop over before DDR3. Lets hope DDR3 sticks around a little longer. I see that Nvidia might be going with DDR5 in the very near future which makes me think DDR4 might go by as quickly as DDR2. I think even running the highest clock speeds available in DDR3 rarely boost performance enough to notice outside of a benchmark. DDR3's real potential will be seen much like DDR2's at the end of its cycle.

 

[skittle]

I skipped over this article... toms has been... lower quality lately.

 

[bydesign]

I think the point was DDR 2 800 is 98% as fast as the fastest DDR3 in most applications. The fact is that Intel chips don't need the extra bandwidth. I doubt that AMD results would be all that different.

To the OP they tested cheap DDR2 as well. Do you use DDR3 or something because a 200MHz boost isn't going to change the results very much?

 

[SpinachEater]

I could care less about that DDR3 money furnace at the time being. I did enjoy seeing the CL4 vs CL5 for DDR2 however.

I agree though, drawing conclusions from tests on one RAM sample does not validate statements regarding the entire DDR3 population. On top of that, their RAM selection isn't even from the same companies....huh? I can think of a better way to organize the test of DDR2 vs DDR3 too but the article is what it is. I think the most informative and accurate info comes from the timing comparisons performed with each RAM sample like the latency changes in DDR2.

The conclusions that are drawn between DDR2 and DDR3 are only relative to Patriot PDC32G1600LLK DDR3 and Corsair Dominator CM2X1024-8888C4D DDR2, not the entire DDR2 and DDR3 population. So the actual conclusion should say "In our tests, this Patriot DDR3 is comparable to this Corsair DDR2" not "DDR3-1066 and -1333 memory do not yet result in better performance (than DDR2 800)" That is a very sloppy conclusion.

They should have compared these two in order to draw real conclusions.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820145198
http://www.newegg.com/Product/Product.aspx?Item=N82E16820145194

 

[rockbyter]

i think the point thats been made is that it doesnt matter. small differences exist between a huge range of frequencies and speeds on the core 2. Something else - namely chipset and processor limits everything. We will see some difference between them on amd stuff or intels next gen but amd doesnt have a ddr3 solution so this article serves as a discrete Intel platform specific statement that it doesnt matter

 

[dragonsprayer]

i am being very specific that ram is not $159 - if you find that 3-3-3 ddr800 ram its really expensive - that is what they used

then they used low end 1600, the better 1600 runs 1700-1800 which is the sweet spot, at 7-7-7 with 2v

this test totally ignors the best settings and uses mid leverl ddr3 and compares it to probably the single fastest ddr2 made - not fair!