-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
DDR400 CAS timings vs. DDR2-800 and above
- Thread starter dawgtuff
- Start date
18 Is Number 1
Member
DDR and DDR2 are not interchangeable. However higher RAM speed generally offsets lower latency. Intel processors really like higher speed (MHz).
alcoholbob
Diamond Member
With integrated memory controllers, the bandwidth is so high that you will want to focus on latency over speed--at least this was true with AMD64.
zerocool84
Lifer
A little behind the times are we??
Anywho, first of all you cannot compare DDR to DDR2. Secondly, faster RAM is better than getting better timing RAM. RAM timing would only be noticeable in benchmarks and not realworld.
Forumpanda
Member
Keep in mind that latency is defined in clock cycles, DDR2 (and 3) RAM is likely to have a higher clockrate so the absolute latency (in seconds) is likely very similar.
And without knowing it I would guess that if there is an increase in latency it comes from the increase in memory per chip, which is still a net benefit (more memory is almost always preferable).
And without knowing it I would guess that if there is an increase in latency it comes from the increase in memory per chip, which is still a net benefit (more memory is almost always preferable).
lopri
Elite Member
I think what OP is asking is:
DDR400 / CL 2.0 -> 10 ns
DDR2-800 / CL 4.0 -> 10 ns
DDR3-1600 / CL 8.0 -> 10 ns
What's the point? is his question. (a well-justified question, IMO)
OP: While the first access time might be the same for different generation of DDR, as the data size gets bigger and sequential higher frequency improves latencies. And each new generation of memory improve density, power/thermal characteristics as well.
DDR400 / CL 2.0 -> 10 ns
DDR2-800 / CL 4.0 -> 10 ns
DDR3-1600 / CL 8.0 -> 10 ns
What's the point? is his question. (a well-justified question, IMO)
OP: While the first access time might be the same for different generation of DDR, as the data size gets bigger and sequential higher frequency improves latencies. And each new generation of memory improve density, power/thermal characteristics as well.
Ok, from the info ya'll have given me, let me surmise this:1) Faster ram overcomes higher latencies.2) More memory is a plus.3) On chip controllers are better than Northbridge.
I posted this question because I thought that DDR2/3 was somewhat of a scam because of the higher latencies involved. I'm planning on a new build in the next few months and I'm leaning toward DDR3.....is it really worth it? Keep in mind my new build won't be the latest in components.
Present system circa 2006:
Opty 144@2.6
ASUS A8N-E (on chip mem controller)
2gb Crucial @2-3-3-8 1T
EVGA 7600GT(could be better,like a 4850/70)
2-Viewsonic 21" full screen@1400x1050
A well cooled Antec 180 case
I posted this question because I thought that DDR2/3 was somewhat of a scam because of the higher latencies involved. I'm planning on a new build in the next few months and I'm leaning toward DDR3.....is it really worth it? Keep in mind my new build won't be the latest in components.
Present system circa 2006:
Opty 144@2.6
ASUS A8N-E (on chip mem controller)
2gb Crucial @2-3-3-8 1T
EVGA 7600GT(could be better,like a 4850/70)
2-Viewsonic 21" full screen@1400x1050
A well cooled Antec 180 case
The main focus of a fresh build shouldn't be what type of memory to use, that comes farther down the list.
1. Budget
2. Tasks to be performed
3. Find best balance of performance & price, on CPU/GPU/MB/HD and memory that meet the above criteria.
4. Before buying, choose the highest quality components for the job, within your price range.
5. If your budget is unlimited, let's talk! :laugh:
Think of it this way...
* A married couple with three kids, living on a modest income, doesn't go down to the Farreri dealership when they're looking for a new car.
* On the otherhand, a single guy with money to burn would probably not start his car shopping at a Buick dealership.
* Likewise, an oil field worker looking for a work vehicle wouldn't buy a Farreri or Buick.
"Are you following me camera guy?" 😉
1. Budget
2. Tasks to be performed
3. Find best balance of performance & price, on CPU/GPU/MB/HD and memory that meet the above criteria.
4. Before buying, choose the highest quality components for the job, within your price range.
5. If your budget is unlimited, let's talk! :laugh:
Think of it this way...
* A married couple with three kids, living on a modest income, doesn't go down to the Farreri dealership when they're looking for a new car.
* On the otherhand, a single guy with money to burn would probably not start his car shopping at a Buick dealership.
* Likewise, an oil field worker looking for a work vehicle wouldn't buy a Farreri or Buick.
"Are you following me camera guy?" 😉
Thanks Blain.....My new system will be budget midrange. I rarely multitask and the hardest thing will be photo editing and games. I'm going to keep my 2 -21" monitors,so I won't be going above 1400x1050.
In a few months, I should be able to get a good, OC'able system for around $800. Thanks again for ya'll's input......Dawgtuff.
In a few months, I should be able to get a good, OC'able system for around $800. Thanks again for ya'll's input......Dawgtuff.
alcoholbob
Diamond Member
Proof?
alcoholbob
Diamond Member
No, the question is, what is zerocool's proof.
Because google (http://techreport.com/articles.x/15967/6) obviously shows that at the very least, tighter timings is as good as high clockspeed.
In fact, it seems to show that typically, tighter timings > RAM speed.
So again, I want to see zerocool provide proof. What's the point of telling me to look up google? Does google have zerocool's opinion written somewhere? If he has a blog, maybe you could help me locate that because it seems like my google skills are lacking.
n7
Elite Member
What?
Look at your own link.
Specifically, the 965 results which are running the same speed multiplier/QPI.
In every single game, the 1600 8-8-8 beats the 1333 7-7-7.
The 920 @ 16x167 vs. 13x200 isn't an accurate comparison, as we have a different QPI & consequently different uncore speed...
There's nothing to argue here.
If you've followed RAM at least slightly in the last few years you'd know that generally speaking, higher speed will beat lower speed with tighter timings, with a few exceptions.
The exceptions:
- nForce chipsets where you can do 1:1 synced, especially with 1T.
- situations in which you have say DDR2-800 4-4-4 vs. DDR2-850 5-5-5....sure, the lower speed tigher timings would win that.
But on Intel chipsets, the difference is rarely that low, & nearly every time, higher speeds will beat lower ones with tighter timings, especially since you can usually gain a lot of speed with the slightly looser timings.
This is true for i7 also.
I've done more than enough testing on my own mobos in the last few years to understand this, & reviews back this up.
Now all this said, there's very little difference in real world scenarios between the lowest end & highest end RAM...so it often really doesn't matter much.
alcoholbob
Diamond Member
Actually, the link is terrible. I've yet to find one where there was direct comparison on equal ground between timings and RAM speed. Obviously 1600 CAS 8 will beat 1333 CAS 7 since there's more work per cycle.
I think the reason why these benchmarks are lacking is due to the dominance of Intel over the last few years, which is why there aren't proper benchmarks regarding timings.
lopri
Elite Member
Here is a slightly better one. I see what you're saying, but it is what it is. With the ever increasing size of cache, higher bandwidth will win over lower latency 9 out of 10 times. I myself prefer efficiency over waste, but chips are getting faster yet the interconnect needs to maintain the standards, so the trend won't change anytime soon.
http://www.anandtech.com/cpuch...howdoc.aspx?i=3551&p=3
Besides, you can trust what n7 says when it comes to memory. He wastes plenty of time on negligible performance gain and as you can see he's proud of it! 😉
Latency might become more important when Intel/AMD start making those CPU+GPU / CPU+PCIe / CPU+everything chips.. we'll see how things will pan out.
n7
Elite Member
Aha!
Now we're getting somewhere.
You're absolutely right on that, though it's certainly easier to get closer to accurate numbers on a system with a built in memory controller.
Fact of the matter is, for Intel FSB-based platforms, it's pretty much impossible to get a perfectly accurate comparison, since even if you keep the FSB the same speed, when ratio is changed from say 1:1 to 5:6 or 4:5, the tRD & other subtimings actually change unless you force them not to, which also affects performance.
And even if you force them the same, there are always internal frequencies/timings @ play the user cannot control.
E.g, on most P45s, running 1:1 (FSB:RAM) at the same speed is actually the worst possible scenario, no matter how tight you an get your timings, as for whatever reasons, tRD is always looser there than 5:6, meaning you can basically always easily beat 1:1 regardless of whether it was CAS 4 vs. even CAS 6.
Now testing aside, one thing that makes it pretty obvious to me that bandwidth (higher speed) is superior to timings is how the DRAM industry keeping pushing for higher speeds, usually at the expense of timings. If tightening latency had as much of an affect, i imagine we'd see more of a push toward super low latency instead of the opposite.
Intel, AMD, nV...they all design around the improved speeds...or perhaps it's the DRAM industry partially designing around Intel/AMD/nV's suggestion for improved speeds?
That's well beyond my understanding, so i could very well be mistaken there.
alcoholbob
Diamond Member
It appears it's fairly typical behavior for you to make interjecting quips into threads or emoticons as the content of your entire post.
bobross419
Golden Member
Here are a pair of AT articles on the subject of DDR v. DDR2. It was kind of sparked by the presence of both memory slots on the AsRock 775Dual-VSTA mobo.
http://www.anandtech.com/mb/showdoc.aspx?i=2810
http://www.anandtech.com/mb/showdoc.aspx?i=2813
Maybe not exactly what you guys are discussing, but I thought it might be pertinent to the topic and help the OP.
http://www.anandtech.com/mb/showdoc.aspx?i=2810
http://www.anandtech.com/mb/showdoc.aspx?i=2813
Maybe not exactly what you guys are discussing, but I thought it might be pertinent to the topic and help the OP.
I thought this post had a lot of good information in it:
http://www.overclock.net/intel...ndwidth-explained.html
The 1:1 ratio myth was a surprise for me.
http://www.overclock.net/intel...ndwidth-explained.html
The 1:1 ratio myth was a surprise for me.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
