Settling The DDR4 Vs DDR3 Debate: ASRock Fatal1ty Z170 Gaming K4/D3 ATX Motherboard Review

[alidan]

honestly, what i would love to see is ddr2 compared to ddr4, synthetic benchmarks will show 4 kicking the crap out of 2, but real world i think quantity matters far more so you aren't paying crap.

 

[ramon zarat]

Brilliant idea to bench the 1600Mhz memory at CL11... My G Skill 1600Mhz 16GB kit (4X4GB) has been running flawlessly for years at 8-8-8-21-1T 1.5V for christ sake...

 

[Crashman]

Brilliant idea to bench the 1600Mhz memory at CL11... My G Skill 1600Mhz 16GB kit (4X4GB) has been running flawlessly for years at 8-8-8-21-1T 1.5V for christ sake...

JEDEC standard timings were used for both baselines, and C11 is the remaining DDR3-1600 standard since memory manufacturers abandoned JEDEC DDR3-1600 C9 a couple years ago. Most memory PR doesn't even remember that JEDEC DDR3-1600 C9 even exists.

The JEDEC standard program for your kit probably isn't even at DDR3-1600. It's probably 1066 or something.

I don't see you crying over the DDR4-2133 C15 baseline. Oh yeh, because that's JEDEC standard. You would have us use nonstandard memory settings for the DDR3 baseline, but wouldn't concern yourself with the baseline of DDR4? Yours is the kind of bias that was intentionally removed from baseline settings of this article.

 

[Mindaugas_3]

What about performance of integrated GPU, does ddr4 is better than ddr3?

 

[Crashman]

What about performance of integrated GPU, does ddr4 is better than ddr3?

Find me an integrated GPU that's both worth testing AND compatible with both.

Lacking that, I can still go back to the latency math...

 

[blazorthon]

What about performance of integrated GPU, does ddr4 is better than ddr3?

Like Crashman said, the problem is that Skylake doesn't have any of the good Iris Pro IGP models out yet (GT4 looks awesome on paper, but it isn't off the paper yet) and AMD doesn't have any "high" end APUs that support DDR4, so we can't make such a comparison. We can bet on DDR4 making a huge difference, especially on AMD where there isn't a large and fast eDRAM cache, but there just aren't any processors to test.

 

[rantoc]

This makes me feel better about my choice of upgrading from DDR3 1866 to DDR4 3000 on my Sky Lake build. Those higher latencies did put me off a bit.

Many seem forget to calculate the real world latency, since DDR4 often works faster even with higher rated latency in cycles the real world latency is lower.

Say a DDR memory works in 1500 and another in 3000. If the 1500 is rated to 10 cycles in a latency value that could be doubled ti 20 in the 3000 version and it would break even.

 

[Nintendork]

You really need to test minimum DDR4 3000 to actually talk about skylake and DDR4.

 

[ammaross]

We know DDR4 is better, its just not worth upgrading simply for DDR4. Upgrade for other reasons, such as features that Skylake has.

People are still satisfied by sandy bridge, but upgrading now will bring a lot to the table. M.2, DDR4, more native usb 3.1, faster CPUs, lower power usage, etc. Upgrading for one reason is silly. Upgrade for the platform.

Reason to use DDR4 over DDR3: voltage. If you are chasing speed with 1.65v DDR3 in an attempt to keep up with DDR4, you're out of spec on Skylake. If you fry your mem controller, that's because you thought you were smarter and could "cheat" the system. Yes, you may get a configuration that is temporarily faster for today, but will it still work properly in 6mo? A year? Is shaving a few seconds off a long encode or a couple FPS out of 100 really important enough you're willing to deal with instabilities and repairs down the road?

 

[Jefftb]

No minimum FPS recorded during a CPU memory benchmark? Come on, Toms..... Get with the program. We demand some kind of 1% low or min fps on this kind of article.

 

[Sam Bittermann]

Since when have any of us that read these articles are serious gamer's or do it yourself builders used JEDEC standards when setting up their gaming pc's? And yes I understand the reasoning behind using JEDEC but that seems to be a weak argument.

PS: That withstanding great article as the differences should correlate no matter what timings you use!

 

[Crashman]

Since when have any of us that read these articles are serious gamer's or do it yourself builders used JEDEC standards when setting up their gaming pc's? And yes I understand the reasoning behind using JEDEC but that seems to be a weak argument.

PS: That withstanding great article as the differences should correlate no matter what timings you use!

JEDEC is an easy starting point. High O/C was the target. I could just as easily say "if you're a performance PC builder you probably won't use any DDR3-1600, even at CAS 7". But testing only the max performance doesn't show the spread between "standard" and "what you can get".

 

[IInuyasha74]

Nice article Thomas, been wondering how the two memory standards would fare against each other.

For those upset that DDR3 1600 MHz CAS 11 RAM was used, like Crashman already said, that is the official JEDEC standard. It is true that DDR3 1600 MHz CAS 9 RAM is popular, but RAM at those settings came later and is already running above and beyond the standard DDR3.

If anything, we should be looking closer at what DDR4 will eventually be. DDR3 2400 CAS 11 is close to the maximum performance you are going to get out of DDR3, but we have already seen DDR4 pushing past 4000 MHz. Those kits aren't common yet, and we don't know where DDR4 will ultimately land, but it seems likely that we will see this slim performance gap grow as the standard matures. As that is all still to come, this article does a good job of representing the current state of high performance DDR3 vs. decent DDR4 that is above the JEDEC standard, but still far below what DDR4 has already been seen capable of reaching.

 

[Onus]

^I agree (although the editor in me insisted on a word change). If DDR4 is even a little bit better now, that difference will only grow.

 

[ramon zarat]

Brilliant idea to bench the 1600Mhz memory at CL11... My G Skill 1600Mhz 16GB kit (4X4GB) has been running flawlessly for years at 8-8-8-21-1T 1.5V for christ sake...

JEDEC standard timings were used for both baselines, and C11 is the remaining DDR3-1600 standard since memory manufacturers abandoned JEDEC DDR3-1600 C9 a couple years ago. Most memory PR doesn't even remember that JEDEC DDR3-1600 C9 even exists.

The JEDEC standard program for your kit probably isn't even at DDR3-1600. It's probably 1066 or something.

I don't see you crying over the DDR4-2133 C15 baseline. Oh yeh, because that's JEDEC standard. You would have us use nonstandard memory settings for the DDR3 baseline, but wouldn't concern yourself with the baseline of DDR4? Yours is the kind of bias that was intentionally removed from baseline settings of this article.

"Crying"??? WTF, dude? Just went to NewEgg U.S. for a quick reality check and guess what? Out of the 1938 DDR3 module available, 1433 are of CL10 or lower!!! 74%!!! CL9 by itself in fact represent a whopping 49% of all DDR3 modules listed on NewEgg!!! CLL11 and higher are for all intent and purpose limited to ECC, low voltage and ultra cheap memory.

JEDEC is a strictly followed standard like lava is cold as the vast majority of modules DO NOT CONSTRAINT THEMSELVES WITHIN THE BOUNDARIES OF THAT SO CALLED "STANDARD". What is the point and credibility of a "standard" if everyone ignore crucial part of it, like voltage, speed AND timing???

Unlike you, I'm not concerned with "standard". I'm concerned with actual ***REALITY***. And the reality is:

1- The vast majority of DDR3 module available are below CL11.
2- Memory speed beyond 1600MHz DO NOT SCALE with newer CPU technology, except for 2 specific scenarios: Adobe and 7-zip as this article perfectly demonstrate DESPITE using DDR3 CL11 that NO ONE USE OR BUY!!!
3- Buying "fast RAM" at this point in time is completely idiotic as DDR4 3.8GHz often perform within the margin of error compared to memory 60% slower!!! There is also no justification for DDR4 beyond better power consumption.

The bottom line is this: Benchmark should use hardware component that are REPRESENTATIVE OF WHAT USERS ACTUALLY USE and hiding behind "standard" to justify weak argumentation incompatible with reality is inadmissible...

 

[clonazepam]

You clearly wanted a different article. You should go write it.

 

[Outlander_04]

Most 1600MHz is sold as CL 9
It will run at that speed once you apply the appropriate XMP profile.

So why run this test after crippling the performance of the DDR3?

 

[logainofhades]

Maybe some should realize that us enthusiasts are a minority, and most systems out there are cheap OEM's and are probably using JDEC standard C11 ram.

 

[Outlander_04]

Maybe some should realise that this is an enthusiast site for computer enthusiasts

 

[Onus]

Indeed; people who know how to interpolate and extrapolate using a variety of available data points.

 

[logainofhades]

Maybe some should realise that this is an enthusiast site for computer enthusiasts

Not only enthusiasts use this site. Many OEM users come here looking for answers as well.

 

[IInuyasha74]

Maybe some should realise that this is an enthusiast site for computer enthusiasts

This is a technology website for PC enthusiasts, PC beginners and everything in between. If you don't believe that is true, you should probably listen closer to Onus (a TH motherboard reviewer), Crashman (a reviewer of several different components who has been here for numerous years) and myself (contributing writer for various aspects of the site), all of whom acknowledge that we have a diverse community and that using JEDEC's DDR3 specification was the most practical solution. If you think that we are wrong, then clearly you know more about testing products and this website than the people that work on it (which is simply not true).

 

[Outlander_04]

Maybe some should realise that this is an enthusiast site for computer enthusiasts

This is a technology website for PC enthusiasts, PC beginners and everything in between. If you don't believe that is true, you should probably listen closer to Onus (a TH motherboard reviewer), Crashman (a reviewer of several different components who has been here for numerous years) and myself (contributing writer for various aspects of the site), all of whom acknowledge that we have a diverse community and that using JEDEC's DDR3 specification was the most practical solution. If you think that we are wrong, then clearly you know more about testing products and this website than the people that work on it (which is simply not true).

I would say , still , that testing RAM at its advertised speed and timing settings would be the most appropriate option . This is after all what a consumer would expect .
Another might be to run it at its maximum stable overclock as a comparison [ but accept that is outside the scope of this article]

 

[f-14]

If you really want to compare the memory performance in games, you have to pick games where the CPU is easily a bottleneck. GTA V for example. Not BF4... And, using a weak CPU instead also helps. So, an i3 rather than an i7.

i3.lmfao i3 doesn't support ddr beyond 1600 and it's fsb is limited to 1333 in all i-3's pointless to even bother with i-3 it's work station cpu.

"Consider that CAS 11 is the most-common latency standard for DDR3-1600"
Simply not true, the most common latency for DDR3 1600 is CAS 9, I don't think I've ever even seen a 1600MHz stick with a CAS latency of 11 (other than cheap value RAM), CAS 10 is more common than CAS 11 even, but really it should be CAS 9. In fact CAS 9 can and has been used on DDR3 modules up to 2400MHz, further making your use of CAS 11 on 1600MHz bunk...

idk where the cas11 came from 8-8-8-24 was the standard for ddr3 1600 aside from 9-9-9-24 which was considered the slow stuff CAS wise. this cas 11 stuff only came out when you were trying to buy 1600 for $20 for 8gb sticks, seemed like a purposely built handicapped ram just to make ddr4 look attractive, much like the 64bit ddr5 video cards vs the same model 128bit with ddr3. 128bit ddr5 models refute everything said about 64bit even if it had ddr12 soon as you bumb up that bit width back to 512 like the ati3850 and 3870's they destroyed anything 256bit or less.

We saw this same thing when DDR3 came out. "Is Core 2 better with DDR3?" We saw DDR3 come out at 1066, when DDR2 was just reaching 1066 and even 1200. And no, DDR3 wasn't any faster. As a matter of fact, it was slightly slower. However, DDR3 was able to stretch on and get faster than DDR2 would ever have been able to do.

It doesn't really matter. DDR4 is needed for the future. We need a point where people start adopting it. That point is now. The platform is this one.

So, stop griping about it.

ya it's a repeat of ddr1 vs ddr2 at the 400mhz level and ddr2 vs ddr3 at the 1066 level

ddr3 at 3200 stomps this pathetic ddr4

CORSAIR Dominator Platinum 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 3200 (PC3 25600) Desktop Memory Model CMD8GX3M2A3200C13
Cas Latency: 13
Voltage: 1.65V
Timing: 13-15-15-39

CORSAIR Vengeance Pro 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 3000 (PC3 24000) Desktop Memory Model CMY8GX3M2A3000C12R
Cas Latency: 12
Voltage: 1.65V
Multi-channel Kit: Dual Channel Kit
Timing: 12-14-14-36

 

[turkey3_scratch]

The thing is, the high-frequency DDR4 latency really is not much higher than the 2133Mhz DDR4. The high-frequency DDR4 IMO is a great buy. But as these charts show, all these make little difference, though the difference indeed is visible.