DDR600 TCCD vs. DDR500 UTT (Results Inside)

Intro: This argument comes up time and time again on every forum you visit, but which one is really "better"? Well why don't we run a few tests and see if we can't come up with an answer shall we? What we'll do here is to compare a set of G.Skill TCCD @ DDR600 to a set of UTT @ DDR500 and see who comes out on top. By running a set of very common benchmarks hopfully this will give us some idea of which set of memory will fit your system best. Although this won't be the end of the arguement, perhaps this will makes your decision a little less cloudy.

Method: So what i did here was take a set of G.Skill PC-4400 LE TCCD 440 memory and put it up against a set of OCZ PC-4000 Gold VX UTT running in the exact same system. The only difference with these tests was the memory sticks used. I took a series of common benchmarks and ran them each a total of 3 times, and took the score that fit into the middle as the end result. Although there is no real good way to make a comparison for everyone, this should give a good idea of how the two sets stack up.

Test Setup:

DFI NF4 SLI-DR 6/23-3 Beta
AMD A64 FX-57 @ 3000MHz 1.4v
Sapphire x850XT Cat 5.6 (520/540)
OCZ Powerstream 600x Adj
2x74GB Raptors RAID 0
Windows XP SP1
Mach II Cooled

Memory:

2x512mb G.Skill PC-4400 TCCD LE @ 2.7v



2x512mb OCZ PC-4000 Gold VX @ 3.2v



Results:

Super PI 1m:

TCCD



UTT



Here in our first test you can see TCCD clearly beats the UTT at these settings. This test has always been very memory dependent. Through all my testing i have found it takes UTT at least 265HTT or better to take down a good set of TCCD @ DDR600.

Futuremark PCMark04 Memory Test Suite:

TCCD



UTT

Futuremark PCMark05 Memory Test Suite:

TCCD



UTT



So here we see two difference results from similar benchmarks. PCMark05 takes much better advantage of latency than PCMark04 does. Although the clear winner here is the TCCD, the UTT does very well in PCMark05 and could probably be far more competitive given more speed. Note: The TCCD also beats out the UTT in overall score at these settings.

Sisoft Sandra:

TCCD

Bandwidth


UTT

Bandwidth


TCCD

Arithmetic


UTT

Arithmetic

TCCD

Multimedia


UTT

Multimedia


No supprise here on the bandwidth results, the TCCD is the clear winner in this benchmark. Both the CPU tests show slight gains while using low latency UTT. Both Arithmetic and Multimedia tests appear to be almost tottaly CPU dependent, but can take advantage of UTT memory latency. Useing Sandra we have a tossup. Both sets show they are extremely capable in this benchmark.

Everest 2.02.291 Beta:

TCCD



UTT



Testing in Everest shows mixed results. I think it is pretty clear the low latency UTT does very well here. Given more MHz the UTT would easily pull ahead in Latency and Memory Write. Although the TCCD beats it in every test, this is not a conclusive test that TCCD is faster overall than UTT based modules here.

Futuremark 3DMark03:

TCCD



UTT



This benchmark is very GPU intensive, but memory and CPU also play a role here. Given these results there is no winner overall, however once again if the UTT were clocked higher we would likely see it pull ahead here in 3DMark03.

Futuremark 3DMark2001SE:

TCCD



UTT



Here once again we see very similar results between the two. This benchmark is probably the least conclusive of all as there are several varibles that need to be taken into account here. Although the TCCD showed better today, 3DMark2001 clearly eats up low latency. With 50MHz less on the memory it is still neck and neck with DDR600. Low latency gets the nod here.

Conclusion: Well there you have it, some simple tests between Low Latency UTT and High Bandwidth TCCD. So what does this prove? Well not a whole lot really. Although TCCD clearly beat the UTT here at these speeds this will not put to rest the argument of which is "better". There was no way to test both sets at their max and still maintain the same clocks on the CPU, this makes the results slightly bias to the TCCD @ DDR600.

I have tested UTT/TCCD on several platforms and always have gone back to TCCD as the superior RAM. Why you ask? Well it's simple, the TCCD allows for a much larger range of clocking options for me. I've had a host of problems using UTT at high HTT + dividers. Without the right multi the UTT just does not allow for that range in clocking options that many (myself included) need. TCCD always shows strong from 280HTT-300+ while UTT only shows strong at speeds over ~250HTT making it much less adaptable to all systems. What does this mean to you? Well it's quite simple really, you need to match your system to your memory. In many cases the UTT can and will be better. You need to look at the max HTT your CPU can obtain, and match the speed to the memory of your choice. If you're going to be limited to anything less than 250HTT 1:1 you might want to consider a set of TCCD as you can always drop the multi and run 1:1 at almost any memory frequency. If you have a system with a low multi you run the risk of haveing similar problems that i have experienced over the past few months in regards to high HTT + Dividers. Although many report no issues, there are many varibles to take into account. The CPU Memory Controller, Motherboard, Motherboard Options/Divider options, and so forth. While these tests do not show what memory is superior, i think it should make your choice a little less cloudy.

Hope you enjoyed the read. I really wish i could max both sets out and still have the same CPU clock, but given my system that is not possible. I will try and add in some BH5 results asap and see if we can get a more clear picture.

Great stuff Sucka..... I've been wanting to do this with all my ram but I don't have enough memory controller to make it even. (subscribed)

HousERaT said:

Great stuff Sucka..... I've been wanting to do this with all my ram but I don't have enough memory controller to make it even. (subscribed)

Click to expand...

Thanks.

This was really the only "even" way to test them head to head. Short of shaving some serious MHz off the CPU we can't get a apples to apples max clock comparison. The TCCD is pretty well maxed out here, but the UTT has another 15HTT left it in without any problems. Once i go beyond 265HTT i need to really lax timings.

It will still be interesting to see what will happen with BH-5 at 250..... some of those timings should be even lower.

How fast could you run the UTT with 2-2-2-5 timings with the same volts?

HousERaT said:

It will still be interesting to see what will happen with BH-5 at 250..... some of those timings should be even lower.

Click to expand...

How are you going to beat 1.5-2-2-5 timings?

Quailane said:

How fast could you run the UTT with 2-2-2-5 timings with the same volts?

Click to expand...

Well there is no real difference between 1.5CAS and 2CAS. Kinnda like the placebo effect. So there would be no difference in results between CAS1.5 and 2.

At 3.2v these sticks max out at around 255HTT, and do up to around 262HTT with 3.3v

Great thread

The PI results are the main interest for me, it’s the only bench I really like. I have been saying that DDR600 @ 2.5-3-3 kicks butt around here for a while now but everyone just argues about it.

You could gain some more time with the TCCD if you used 16 ICL like you do for the UTT as well.

Eldonko said:

You could gain some more time with the TCCD if you used 16 ICL like you do for the UTT as well.

Click to expand...

I'll give it a go, but last time i used 16 on the TCCD's it showed no gains. I'll test more.

And yes, i've been saying for quite some time now that tight TCCD is very comparable to tight UTT's. It's when you loosen the TCCD's that they really fall behind.

And thanks for the feedback. Nice to know i'm not wasting my time

I always use 0 ICL for my BH, but it is tougher to get TCCD to use low ICL. That is a good set of TCCD you have there, 2.7v is craziness :O

You are not wasting your time, I love threads like this.

Eldonko said:

I always use 0 ICL for my BH, but it is tougher to get TCCD to use low ICL. That is a good set of TCCD you have there, 2.7v is craziness :O

You are not wasting your time, I love threads like this.

Click to expand...

It does 295HTT memtest stable with only 2.6v



Now that's impressive for San Diego/TCCD.

Thanks for the tests Sucka
BTW those LE's are being well taken care of

Ok good comparison, but I need to know about my particular situation. How does 265HTT @2-2-2-5 compare to whatever timings at 300HTT? I happen to have the same RAM as you so I know you can do it. I may give up the PC4000 VX if the comparison goes in favor of TCCD.

/edit: God TCCD at 320HTT is freakin insane. I know this comparison is a no-holds-barred type thing but how much did the G.Skill cost?

/edit pt. 2 Ok I see in your first post yo say it takes 265Htt for the VX to surpass TCCD at 300HTT. What is a safe voltage to run these VX's, as I can get mine to run at 270HTT at 3.7v, but I wuss out after my suicide runs.

Skeen said:

Ok good comparison, but I need to know about my particular situation. How does 265HTT @2-2-2-5 compare to whatever timings at 300HTT? I happen to have the same RAM as you so I know you can do it. I may give up the PC4000 VX if the comparison goes in favor of TCCD.

/edit: God TCCD at 320HTT is freakin insane. I know this comparison is a no-holds-barred type thing but how much did the G.Skill cost?

/edit pt. 2 Ok I see in your first post yo say it takes 265Htt for the VX to surpass TCCD at 300HTT. What is a safe voltage to run these VX's, as I can get mine to run at 270HTT at 3.7v, but I wuss out after my suicide runs.

Click to expand...

VX at 265HTT would give TCCD at any speed a good run for its money, although i'm not 100% sure it could beat it out fully. I can test VX at 265HTT but i can't provide an apples to apples comparison of TCCD at the same clocks here. 3.7v on UTT is a tad high for my taste, i haven't taken mine over 3.6v, i also saw no gains over 3.5v either. Most discussion is based around 250 low latency vs. 300 loose timings, so that's what i tried to show here. These tests are VERY far from max scores obtained using this hardware.

How does TCC5 compare to TCCD performance wise? Clock for clock

ziggo0 said:

How does TCC5 compare to TCCD performance wise? Clock for clock

Click to expand...

afaik it would put up the exact same results out clock for clock. I don't see how it would differ in anyway at the same settings. I don't have a set to compare with however.

Long Live Tccd!!

Sucka said:

Most discussion is based around 250 low latency vs. 300 loose timings, so that's what i tried to show here. These tests are VERY far from max scores obtained using this hardware.

Click to expand...

So you're saying both the TCCD and the UTT can go beyond the limits of this showdown? What it looks like to me is that UTT clocked over 270HTT and 320HTT with TCCD is comparable. And this scales downward. It seems that UTT can keep up within 50MHZ assuming it can keep its tight timings? I guess then it breakes down to a price : performance ratio. For some with a NF4 chipset the UTT can be a better bang per buck type thing, as opposed to TCCD. At least G.Skill, not incluing some Golden Sample from some other manufacturer.