Does High-Speed DDR3 Help AMD's FX? Four 8 GB Kits, Reviewed

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

oxford373

Distinguished
Jun 22, 2009
618
0
19,060
high memory bandwidth don't improve any thing, this thing i learned from SB-E 3820( SB-E have quad channel memory controller) which performs a little bit faster than i7 2600, because of extra 2mb cache and 3820 is a little bit higher clocked than 2600.
not because of high memory bandwidth.
 

fastleo63

Distinguished
Sep 23, 2009
1
0
18,510
A question to the testers:
And how about CPU/NB clock? It needs to be raised from default, especially with high RAM speeds.
With my 1090T (clocked at 3.2 GHz) i gained about 1.5 Gbps in memory r/w with CPU/NB raised at 2800 MHz (by default, it is set to 2000 MHz).
 

Crashman

Polypheme
Former Staff
[citation][nom]fastleo63[/nom]A question to the testers:And how about CPU/NB clock? It needs to be raised from default, especially with high RAM speeds.With my 1090T (clocked at 3.2 GHz) i gained about 1.5 Gbps in memory r/w with CPU/NB raised at 2800 MHz (by default, it is set to 2000 MHz).[/citation]When you set the max memory multiplier, you automatically get the max CPU/NB multiplier. We wanted to use a lower NB ratio to allow higher reference clocks, but found that it was only possible after lowering the memory multiplier.
 

webcat62

Honorable
Jun 8, 2012
2
0
10,510
I have DDR 3 3400 8GB Ram with an Intel icore7 CPU and Intel's 1333 ram controller. Very much faster than any AMD CPU. almost 60% faster than AMD. Now for the con: Price of Intel icore7 CPU $799.00 plus tax.

webcat62
 
[citation][nom]webcat62[/nom]I have DDR 3 3400 8GB Ram with an Intel icore7 CPU and Intel's 1333 ram controller. Very much faster than any AMD CPU. almost 60% faster than AMD. Now for the con: Price of Intel icore7 CPU $799.00 plus tax. webcat62[/citation]

DDR3-3400? That's something like a suicide overclcok for DDR3, no way you're running it that high. Did you mean DDR3-2400? Even in that case, I'd bet you a lot of money that I could get an an AMD CPU to close that 60% number that you stated down much lower than 60%, depending on the workload. Also, I'm not aware of an i7 that currently costs $800, at least not in the USA. Different country or something?
 
[citation][nom]oxford373[/nom]high memory bandwidth don't improve any thing, this thing i learned from SB-E 3820 which performs a little bit faster than i7 2600 ,trinity piledriver cores close the gab between intel IPC and amd, but if amd fx piledriver(fx 8350) have 10% more IPC over trinity so intel ivy bridge will be just 25% faster IPC so next gen fx will be real threat for intel and may be if AMD will do that they can win in all benchmarks.and may be steel efficiency crown from intel. so amd isn't far away from intel all what they have to do is to improve trinity piledriver cores IPC by only 10%.[/citation]

Some workloads benefit greatly from fast RAM (not so much for most consumer workloads) and I'm pretty sure that more or less all of those numbers are false. Trinity has more IPC than Bulldozer, Ivy Bridge has something like 55-65% more performance than FX at the same frequency (performance per Hz and IPC are not the same thing) and core count, Ivy would still have a huge lead in performance per Hz per core, Trinity is still far from Sandy Bridge (although it closed the gap between Llano and Sandy Bridge on the mobile side by roughly 50%), and FX Piledriver shouldn't have greater IPC than Trinity, just greater performance per Hz due to the L3 cache.

IPC is an architectural term that refers to the cores themselves. The advantage that Vishera (Piledriver FX) would have over Trinity (Piledriver APU) is not in architecturally superior cores, but in having a high-capacity L3 cache. The cores themselves should be unchanged, thus IPC should be unchanged, they simply have an improved cache/memory system due to the L3.
 

lunatic80

Honorable
Jul 6, 2012
1
0
10,510
I have the same CPU and a pair of Kingston Hyper X T1 at 1866 black edition 2 X 4 Gb kit (KHX1866C9D3T1BK2/8GX) on Crosshair V Formula and Sandra 2012 gives 21,01 Gb/sec @ 1866 and 22,27 Gb/sec @ 2133 (not perfectly tuned voltage and timings yet).
I run the test after reading this review and honestly I cannot understand why my scores are higher than the review.

http://valid.canardpc.com/show_oc.php?id=2431061
 

Crashman

Polypheme
Former Staff
[citation][nom]lunatic80[/nom]I have the same CPU and a pair of Kingston Hyper X T1 at 1866 black edition 2 X 4 Gb kit (KHX1866C9D3T1BK2/8GX) on Crosshair V Formula and Sandra 2012 gives 21,01 Gb/sec @ 1866 and 22,27 Gb/sec @ 2133 (not perfectly tuned voltage and timings yet).I run the test after reading this review and honestly I cannot understand why my scores are higher than the review.[/citation]The "Test Settings" chart may be listing the wrong version. Since the charts say Sandra 2011, can you try Sandra 2011.10.17.80? Your system is set up, mine isn't, and if the scores align with that Sandra version it would make any necessary article editing easier.
 

masmotors

Distinguished
Dec 9, 2011
501
0
19,010
hey i just got my ares gskill 1866 2x8 16gb kit im looking for amd timings and see none for 990fx i set my to 1866 then 10 11 10 24 its running great i saw 27 some where as last # any 1 info
 
[citation][nom]masmotors[/nom]hey i just got my ares gskill 1866 2x8 16gb kit im looking for amd timings and see none for 990fx i set my to 1866 then 10 11 10 24 its running great i saw 27 some where as last # any 1 info[/citation]

Aren't their timings listed by G.Skill where you bought them from and at G.Skill's own website?
 

ronch79

Distinguished
Jan 16, 2010
181
0
18,680
Badly structured article. Instead of each page containing test results using the same sort of memory, the article could have had pages where in each page tests a certain application using different memory speeds. Plus, those model numbers at least could have been captioned with the actual memory speed, not just pure model numbers, which could be confusing. Again, badly structured article. This is one reason why Tom's is no longer my main tech review site.
 
Status
Not open for further replies.