DDR600 TCCD vs. DDR500 UTT (Results Inside)

[Sucka]

So is it possible to be a lil unstable in test #8 and still be windows stable? If that's the case, I may just live with #8 being stupid.

Well it's certainly worth a try. Memtest will only tell you that the memory is Memtest stable, and nothing else. As i've said in the past, running 1 program tells for any period of time only shows that's it's stable in the given program for that time and nothing else. What i would suggest is running the settings you're having problems at in Windows doing whatever your main tasks are, and if they can do all those tasks, who cares what Memtest says? Don't sacrafice clocks on account of a few errors. I take Memtest for what it is, a general guide to what's stable. If that was the end all program no one would ever have problems in Windows after passing Memtest, and of course we know this isn't the case. Good luck, i'm sure you'll get it squared away

 

[larva]

Nice thead, one suggestion, though.

Why not judge the memory by the overall PCMark scores, rather than the memory component? CPU, memory, and chipset interact in such complex and variable ways that knowing which memory scores best on the memory section is really just a curiosity. The point is to make the fastest system, not the fastest memory bench. And if PCMark's overall result isn't valid in a purely comparitive sense such as this, I don't really trust the memory component, either.

I vastly prefer things like the SuperPi, overall PCMark scores, 3DMark2001 overall scores, and Q3A fps for guaging the impact of memory performance. And conversly, I just couldn't care less what makes the highest sandra numbers, they are at best a means to an end, and we can just measure the end itself.

 

[Sucka]

Nice thead, one suggestion, though.

Why not judge the memory by the overall PCMark scores, rather than the memory component? CPU, memory, and chipset interact in such complex and variable ways that knowing which memory scores best on the memory section is really just a curiosity. The point is to make the fastest system, not the fastest memory bench. And if PCMark's overall result isn't valid in a purely comparitive sense such as this, I don't really trust the memory component, either.

I vastly prefer things like the SuperPi, overall PCMark scores, 3DMark2001 overall scores, and Q3A fps for guaging the impact of memory performance. And conversly, I just couldn't care less what makes the highest sandra numbers, they are at best a means to an end, and we can just measure the end itself.

I chose not to include overall scores in PCMark because in the end they showed the same end result of TCCD being faster at the given speeds. Had i run the system suite, the same trend of result would have been identical, therefore i chose to leave it out. As you stated running the whole suite would test every component, which is exactly what i was trying to avoid as much as possible in these tests. I have read many reviews in the past, and most sites use PCMark memory as a way to test, so i think it's a very valid test to take into account here. But since it was brought up, i'll try to get them both run in the near future. Thanks for the input, i will take that into consideration for future tests i run.

As for Sandra, we all know TCCD beats UTT in bandwidth alone, but why would i exclude one of the most popular tests everyone loves to run

 

[larva]

As you stated running the whole suite would test every component, which is exactly what i was trying to avoid as much as possible in these tests.

I just feel that the most valid way to judge memory performance is to guage its impact on the overall system performance. Synthetic tests are so prone to uselessness that in the end the only to make them even remotely valid is to calibrate them against the overall system performance. I am glad the overall PCMark numbers correlate well with the memory scores, but if I had to act on one or the other, it would have to be the overall.

In the end though, I agree, both results are the most illustrative. I just feel that you can't really understand the meaning of the memory score without the overall, where the overall needs no such explanation. SETI classic was just such a case. It did an excellent job at isolating memory performance, too good. The winner in SETI was only the winner in SETI, so unless this one application was your interest knowing which memory in isolation was fastest was of little use. I feel synthetic memory benchies are the same situation-their results are only valid for that benchmark, and therefore cannot be used to advance system design in most cases.

Each user has their own view, of course, and as I spent 10 years building systems, mine comes as little surprise. I focused on creating the best application performance for a given sum of money, and found that although what I could learn about memory performance through its impact on the overall performance could be used to improve my systems, what I learned by testing the memory in isolation was purely academic in the overwhelming majority of instances.

 

[g0dM@n]

Well, all I really care about is 3dmarks... gaming is my #1 concern... but yes, you're right. A lot of people love the bandwidth test(s).

 

[Sucka]

I just feel that the most valid way to judge memory performance is to guage its impact on the overall system performance. Synthetic tests are so prone to uselessness that in the end the only to make them even remotely valid is to calibrate them against the overall system performance. I am glad the overall PCMark numbers correlate well with the memory scores, but if I had to act on one or the other, it would have to be the overall.

In the end though, I agree, both results are the most illustrative. I just feel that you can't really understand the meaning of the memory score without the overall, where the overall needs no such explanation. SETI classic was just such a case. It did an excellent job at isolating memory performance, too good. The winner in SETI was only the winner in SETI, so unless this one application was your interest knowing which memory in isolation was fastest was of little use. I feel synthetic memory benchies are the same situation-their results are only valid for that benchmark, and therefore cannot be used to advance system design in most cases.

Each user has their own view, of course, and as I spent 10 years building systems, mine comes as little surprise. I focused on creating the best application performance for a given sum of money, and found that although what I could learn about memory performance through its impact on the overall performance could be used to improve my systems, what I learned by testing the memory in isolation was purely academic in the overwhelming majority of instances.

You certainly carry valid points. What i tried to do was give as unbiased a look as possible at both sets of memory at their given speeds. But like i said, and you are saying is true, you can't take a few tests and stamp a winner on one set or the other. I'm sure i could sit here for days and find certain applications that would sway opinions one way or the other, so what i did was take several very common benchmarks and use those for comparison sake. These tests are quite time consuming, so i needed to end it at some point. I think the overall tests used can give a good general look into what to expect from each. While they are not the end all results many may be looking for, i think they did exactly what i aimed to do, and that was put some hard results behind DDR600 vs. DDR500. This gives us at least some hard evidence for the next time this comes up, which from what i've seen is daily. But again i do appreciate the feedback. I understand no tests can be perfect, and i always make it a point to mention that in my threads. I only offer my results, i'm not a professional reviewer, so you need to take it for what it's worth. Not a whole lot of people sit down to do these types of comparisions, so i think it at least adds some insight to the end users decision on which memory to go for.

 

[g0dM@n]

Not a whole lot of people sit down to do these types of comparisions

You can say that again!

 

[larva]

Not a whole lot of people sit down to do these types of comparisions, so i think it at least adds some insight to the end users decision on which memory to go for.

No doubt it does, and it is actually more valid than most "professional" reviews. Professional means you are getting paid, not that you are worth it.

And this is a minor point. Fortunately there is no memory component of 3DMark or SuperPi, so there it was mostly application or overall system performance based, anyway. But take it from somebody who has spent untold hours doing these comparisons, overall PCMark can replace the memory score, but the memory score cannot replace the overall result. We are in agreement that publishing both numbers is the best course, and in this case enough agreeement exists that the memory component alone might well suffice, but this is a result one cannot take for granted (not that you were).

 

[Sucka]

Sucka,
Thanks for the comparo Bro!
I could not agree more, it is all dependant upon your individual system.
In my case, the CPU would not even boot past 291 HTT no matter what setting I used above the 8x multi. So, my TCCD ram became effectively useless. After finding the Max stable OC of my CPU which was 262x11, I figured UTT\New BH5 would be far better, as long as it could run @262. My Mushkin Redline does that, though just barely. Still I know that it is far faster than TCCD @ the same speeds, therefore it is the best for my rig.

DREAD

P.S. I tried the settings you suggested and gained maybe a small but hardly noticable increase in speed. But, I had to bump up the voltage to 3.6v + .03 v to get it stable. So, I just put it back for now.

Thanks dread. The memory you choose is highly dependent on your system as you found out. I can't stress that enough. I have the luxery of having both sets here so i can test on both. It's a lot harder for someone buying a complete system as they don't know the max CPU/HTT speeds prior to buying

And hey, at least you tried the timings, always worth a shot for a performance gain

 

[g0dM@n]

All of my UTT ram tops out best at 3.6+0.3v pretty much. Anything higher will not help, even with insane active cooling that I have (3x60mm 22cfm fans), and any less voltage will only lower its end mhz.

Anyone else notice 3.6+0.3v as the sweetspot?

 

[Eldonko]

Sucka, I would be interested to see same test with trcd of the TCCD @ 4 instead of 3, all other settings the same. UTT at 250Mhz vs TCCD @ 300Mhz and 2.5-4-3 are both pretty easily achievable by average sets of RAM, and both have room for improvement. Although your comparison is great, it is not really fair to compare top end OC on TCCD and a meager OC on UTT.

 

[Sucka]

Sucka, I would be interested to see same test with trcd of the TCCD @ 4 instead of 3, all other settings the same. UTT at 250Mhz vs TCCD @ 300Mhz and 2.5-4-3 are both pretty easily achievable by average sets of RAM, and both have room for improvement. Although your comparison is great, it is not really fair to compare top end OC on TCCD and a meager OC on UTT.

I won't be able to do it today, but i'll run some tests with that and see what i come up with. Personaly i don't see there being a huge difference, but who knows, might supprise myself

 

[g0dM@n]

Try 2.5-4-4-8, and 3-4-4-8 as well IF you can. You da mang!

 

[Sucka]

Try 2.5-4-4-8, and 3-4-4-8 as well IF you can. You da mang!

Sure thing. I'm not going to run every test here, so which tests would you like to see. If i run every test i did here another 3 times i'll be at this for another full day

 

[Eldonko]

I won't be able to do it today, but i'll run some tests with that and see what i come up with. Personaly i don't see there being a huge difference, but who knows, might supprise myself

No hurry, I can run the same comparison myself and we can compare our comparisons heh. I think you may be quite surprised to see how much of a difference trp and trcd make. My TCC5 @ 313Mhz and 2.5-4-4 can't touch my UTT @ 260Mhz.

 

[Eldonko]

Try 2.5-4-4-8, and 3-4-4-8 as well IF you can. You da mang!

Dunno if he will be able to run cas 3, I can't even boot using cas 3 with either set of RAM. Changing cas doesn't do much, it's everything else.

 

[ianocean]

My TCC5 @ 313Mhz and 2.5-4-4 can't touch my UTT @ 260Mhz.

Same here I loved UTT myself and i've got also TCCD which hits 300+Mhz

 

[Sucka]

No hurry, I can run the same comparison myself and we can compare our comparisons heh. I think you may be quite surprised to see how much of a difference trp and trcd make. My TCC5 @ 313Mhz and 2.5-4-4 can't touch my UTT @ 260Mhz.

Interesting, i'll certainly take a look at that. Here is something interesting to look at. I ran SPI32m on both TCCD and UTT and the TCCD was "underclocked" compared to its max, and the UTT is just about maxed, and the TCCD beat it by 2 seconds. While there is just a slight difference in CPU speed, it is pretty much dead even.

Take a look here for TCCD.

And look here for UTT results.

I've done my fair share of testing (granted mainly with TCCD) and i still contest that TCCD can beat UTT at almost any clocks only if it is good memory and does tight timings. I'm sure it will be a lot closer the looser the timings get, i'm pretty anxious to find out for myself now

Thanks for the ideas. Not something i really think about much since the sticks i do have clock so well....lol

 

[Eldonko]

It's all buttah at 2.5-3-3, but when you have to run 2.5-4-4 it is a different story. My TCC5 sits on my bookcase collecting dust even though it can do good speeds.

 

[Sucka]

It's all buttah at 2.5-3-3, but when you have to run 2.5-4-4 it is a different story. My TCC5 sits on my bookcase collecting dust even though it can do good speeds.

Yeah, these results obviously don't apply to everyone nor every set of TCCD/UTT. And yes, 2.5-3-3 is very nice