Corsair Intros 3000MHz Vengeance Extreme DDR3 Kit

Status
Not open for further replies.

guru_urug

Distinguished
Aug 6, 2010
160
0
18,680
^^
Even if it did, it wouldnt be economical. You can build a whole better system at the cost of the RAM itself. Still, it would be interesting to see the gains of such fast memory on an APU.
 

crisan_tiberiu

Distinguished
Nov 22, 2010
1,185
0
19,660
[citation][nom]wanderer11[/nom]I'd like to see a benchmark with an APU. Let's see how much faster memory really helps.[/citation]
Probably the gains are massive but @ this price they are pointless....
 

shloader

Distinguished
Dec 24, 2001
231
0
18,690
Ball players need compz, too. This is for the same crowd that takes a pair of Samsung 512GB 840 Pros and throws them in Raid 0.

"Coarsair said these super-fast memory chips are hand-built"... uh huh. With inhuman soldering skills.
 
[citation][nom]wanderer11[/nom]I'd like to see a benchmark with an APU. Let's see how much faster memory really helps.[/citation]


they already have on here and it does help a little bit but anything over 1866 there was only a slight boost in fps. its the lower latency settings that really give you a bigger boost
 

JOSHSKORN

Distinguished
Oct 26, 2009
2,394
19
19,795
[citation][nom]wanderer11[/nom]I'd like to see a benchmark with an APU. Let's see how much faster memory really helps.[/citation]
Remember ? I'm gonna go with "kinda, not really, who cares, save your money and put it towards GPU".
 

Nebby

Distinguished
May 20, 2004
175
0
18,690
[citation][nom]06yfz450ridr[/nom]they already have on here and it does help a little bit but anything over 1866 there was only a slight boost in fps. its the lower latency settings that really give you a bigger boost[/citation]


Maybe its Super Human solders =) with computer precision
 

chicofehr

Distinguished
Jan 29, 2012
538
0
18,990
Reduce the latencies first!!!! 12 is way too high. I want 1600MHz with a latency of 2 or 3, and once we get there, then we can start talking about more speed. They concentrate too much on speed when its the latencies that are making the speed increase somewhat less impressive in real life performance.
 

InvalidError

Titan
Moderator
[citation][nom]wanderer11[/nom]I'd like to see a benchmark with an APU. Let's see how much faster memory really helps.[/citation]
AMD's APUs got benchmarked with RAM at least up to 2400MT/s and there was almost no improvement with going over 2133MT/s. It seems unlikely there would be much if any gains to be made from bumping rates to 3000MT/s on current APUs... certainly not any that could justify the huge price premium on those DIMMs.
 
[citation][nom]06yfz450ridr[/nom]they already have on here and it does help a little bit but anything over 1866 there was only a slight boost in fps. its the lower latency settings that really give you a bigger boost[/citation]

The "latency settings" generally do not give that much of a boost. That's why GPU memory is always focused on bandwidth over latency. Latency does matter, but it is almost always less important than bandwidth.

[citation][nom]06yfz450ridr[/nom]they already have on here and it does help a little bit but anything over 1866 there was only a slight boost in fps. its the lower latency settings that really give you a bigger boost[/citation]

According to Tom's, it's not until after DDR3-2133 where there is a serious cut-off in performance for the top Trinity desktop APUs gain and even then, it might have been a different issue than the GPU not being able to take advantage of greater memory bandwidth.
 

InvalidError

Titan
Moderator
[citation][nom]blazorthon[/nom]No, the "latency settings" do not give that much of a boost. That's why GPU memory is always focused on bandwidth over latency.[/citation]
The importance of latency varies depending on workload-specific access patterns.

Most heavy computational workloads like rendering, transcoding, physics simulations, etc. lend themselves pretty well to long read/write stripes on RAM and will heavily favor bandwidth over latency like GPUs do.

Heavily conditional (branchy) code like compilers, control code for interactive applications, algorithms that rely on sparse arrays, trees and other branchy structures, cache misses are far more common and these scenarios will heavily favor low latency.
 


Those are timings, not latency. The timings are not too high whatsoever. They're actually very low considering the frequency. At that frequency, those timings still mean lower latency than standard DDR3-1600 memory such as 9-9-9-24 and 8-8-8-24 timings. At almost double the frequency and considerably less than double timings, latency is lower.

Latency is found as a relationship between the timings and the frequency and timings must increase as the technology gets higher in frequency and complexity in the hardware. They probably can't make DDR3 CAS 2 or CAS 3 with a frequency as high as 800MHz such as what DDR3-1600 has, at least not with modern technology. So, no, the memory in the article would not at all be held back in latency because it is in fact lower latency than most lower bandwidth modules.

No offense, but you don't seem like you understand the topic enough to make claims such as yours.
 


When we're talking about GPUs, which we were, that some CPU workloads can favor below-average latency over above-average bandwidth seems irrelevant to me.
 

mynith

Distinguished
Jan 3, 2012
133
0
18,680
Remember that a CAS latency of 12 at 3000 MHz is still a shorter amount of time than CL7 at 1600 MHz. And at the end of the day, you have more bandwidth as well. 3 GHz is so ludicrously fast, however, that you don't need it. Want more bandwidth? Then maybe investing in a quad-channel platform like LGA2011 is a better idea. I believe some AMD processors have quad-channel as wel, but I've forgotten. Probably the Opterons.
 


As I recall, AMD's Opterons with quad channel memory are their newer MCM models such as Magny-Cours, Interlagos, and Abu Dhabi.
 
Status
Not open for further replies.