Please Recommend 3x2GB DDR3 Kit

[bofors]

Dear XtremeSystems forum members,

I am looking for a little advice or comments on a high value, performance 3x2GB DDR3 Kit. After surveying the market, it looks like I am in the $200 ballpark.

This RAM will going into an Intel DX58SO revision 504 (and perhaps DX58SO2 if and when released). I only use Intel motherboards because that is the standard for the "hackintosh" Mac OS X development project I work on.

Otherwise, I am planning to use this memory with an Intel Core i7-980X and a Noctua NH-D14 (so I may have to deal with a heatspeader versus CPU heatsink clearance issue on one DIMM, but I am prepared to Dremel!).

Here is what I am naively looking at:

$220 after rebate:
Mushkin Enhanced Ridgeback 998826
DDR3 1600 (PC3 12800) 6-8-6-24 1.65V
http://www.newegg.com/Product/Produc...82E16820226124

$220 after rebate:
Mushkin Enhanced Redline 998805
DDR3 1600 (PC3 12800) 6-8-6-24 1.65V
http://www.newegg.com/Product/Produc...82E16820226121

Is the any real difference between Ridgeback or Redline (besides the free 2GB USB false drive)? Same IC's?


$190:
G.SKILL PI Series F3-12800CL7T-6GBPI
DDR3 1600 (PC3 12800) 7-8-7-24-2N 1.5V
http://www.newegg.com/Product/Produc...82E16820231335


I am inclined to go with the Mushkin Enhanced Ridgeback 998826 kit.

What other 3x2GB DDR3 Kit kits should I be looking at?

Thanks for your input!

[bofors]

I just noticed and read this thread on the Mushkin Enhanced Ridgeback 998826:
http://www.xtremesystems.org/forums/...d.php?t=250809

(By the way, I searched for "998826" with the forum search tool before posting and it did not come up.)


I am wondering about this comment in that thread:

It's PSC, it's the same chip that's used on G.Skill Eco and PIS, and probably Corsair GTX1 as well. My guess is this will be the next best thing when Hyper is gone, it just can't run tight tRCD, hence the spec of tRCD = tCL + 2.



Most likely the same chip but different binning. ( not sure about this. )

I am not particularly concerned with running a tight rRCD, but are any of these Hyper ICs in the $200 price range?

[Musho]

First of all, this is wrong section of the forum. This is for posting high overclocks on RAM. For these kind of question please use the memory forums under "General Hardware". Having said that, generally it's a lot easier to run a fast kit at slower frequency with tighter timings, than to run a slower, but tighter kit at a higher frequency and looser timings. So a high frequency should give you the most options, which is always good (And fun if you like to tinker and tweak ). I would look at the following two sets:
- http://www.newegg.com/Product/Produc...82E16820231306
- http://www.newegg.com/Product/Produc...82E16820231254

They should use the same ICs (BBSE) and as far as I know, the only difference is the slightly better heatspreader on the Tridents, which are also slightly more expensive. Enjoy!

[bofors]

First of all, this is wrong section of the forum. This is for posting high overclocks on RAM. For these kind of question please use the memory forums under "General Hardware".

Sorry about that, I have asked a moderator to move it.

There used to be a sub-forum under Xtreme Bandwidth for these types of threads that seems to be gone now, so I was not sure where to put it.

Thanks for pointing out the correct place for it.

Having said that, generally it's a lot easier to run a fast kit at slower frequency with tighter timings, than to run a slower, but tighter kit at a higher frequency and looser timings. So a high frequency should give you the most options, which is always good (And fun if you like to tinker and tweak ).

I see your point, but was not able to equate DDR3-1600 CL 6 with DDR3-2000 CL 9 due to my inexperience with DDR3 shopping.

I would look at the following two sets:
- http://www.newegg.com/Product/Produc...82E16820231306
- http://www.newegg.com/Product/Produc...82E16820231254

They should use the same ICs (BBSE) and as far as I know, the only difference is the slightly better heatspreader on the Tridents, which are also slightly more expensive. Enjoy!

Same ICs? So I take that I should completely ignore the minor 24 versus 27 difference in the rated timings:

G.SKILL Ripjaws F3-16000CL9T-6GBRH:
9-9-9-27-2N

G.SKILL Trident F3-16000CL9T-6GBTD
9-9-9-24-2N



Since I am dealing with a potential heatspreader clearance issue because of the size of the Noctua NH-D14, I wondering if the top of these Trident heatspeaders are designed to be taken off easily? It looks like the Trident tops might be detachable but I am not sure yet.

[XII]

Looking at your Noctua,

Max hight is Perfect Storm / OCZ Blade size.

Trident are to high. Taking the heatspreaders off, isn`t that difficult,
but take your time.

[Musho]

Sorry about that, I have asked a moderator to move it.

There used to be a sub-forum under Xtreme Bandwidth for these types of threads that seems to be gone now, so I was not sure where to put it.

Thanks for pointing out the correct place for it.



I see your point, but was not able to equate DDR3-1600 CL 6 with DDR3-2000 CL 9 due to my inexperience with DDR3 shopping.



Same ICs? So I take that I should completely ignore the minor 24 versus 27 difference in the rated timings:

G.SKILL Ripjaws F3-16000CL9T-6GBRH:
9-9-9-27-2N

G.SKILL Trident F3-16000CL9T-6GBTD
9-9-9-24-2N



Since I am dealing with a potential heatspreader clearance issue because of the size of the Noctua NH-D14, I wondering if the top of these Trident heatspeaders are designed to be taken off easily? It looks like the Trident tops might be detachable but I am not sure yet.

They are said to use the same ICs, but perhaps the Trident ones are slightly better binned. For the small difference, I'd personally go with the Tridents. I'm not sure on any clearance issues, but a dremel might do the job. Not sure if you'd want to remove the heatspreaders completely. They are there for a reason, although they might be forgiving if you blow some cold air over the sticks.

[bofors]

Looking at your Noctua,

Max hight is Perfect Storm / OCZ Blade size.

Trident are to high.

Thanks for the first-hand experience with the Noctua NH-D14 DIMM clearance.

Too bad the only 6GB PerfectStorm kit on the market now is far outside of the $200 price range:
http://www.extremerigs.com.au/gskill...nel-p-247.html

I am happy to buy RAM from G.Skill, Mushkin, Crucial or Corsair, but keep hearing problems about OCZ memory.

Taking the heatspreaders off, isn`t that difficult,
but take your time.

Right, I actually found some Trident heatspreader removal instructions here:

http://www.overclock.net/intel-memor...l-trident.html

http://www.ocinside.de/go_e.html?htt...tspreader.html

[bofors]

They are said to use the same ICs, but perhaps the Trident ones are slightly better binned. For the small difference, I'd personally go with the Tridents.

At this point, my plan is to get the Tridents.

Thanks again for your advice.

I'm not sure on any clearance issues, but a dremel might do the job. Not sure if you'd want to remove the heatspreaders completely. They are there for a reason, although they might be forgiving if you blow some cold air over the sticks.

I might get lucky with the DX58SO and get the third Trident DIMM to fit next to the Noctua NH-D14, but I am not counting on it:



http://www.legitreviews.com/article/1212/4/

However, I will be putting fans on the RAM whether I have to remove and/or Dremel a heatspreader or not.

[bofors]

Right, I actually found some Trident heatspreader removal instructions here:

http://www.overclock.net/intel-memor...l-trident.html

http://www.ocinside.de/go_e.html?htt...tspreader.html

I was just checking a review on the G.Skill Trident F3-16000CL9 3x2GB kit at OverClock3D.net:
http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/6

They posted a good cross section picture, so I can what I am getting into with heatspreader removal and/or dremeling:

http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/2



That does not look bad at all, but I am much less enthused about this G.Skill Trident F3-16000CL9 3x2GB kit after reading:

As you can see from the shot above, all I could manage was a miserly 26MHz over stock. While I didn't fully determine stability it passed SuperPI and 3DMark06 at this speed. However when the bandwidth was pushed any higher sporadic BSOD began to appear. Lowering the latency had an even worse effect with anything lower than CAS 9 resulting in the Gigabyte motherboard failing to POST, even lowering the bandwidth down to 1866MHz and raising the Vdimm to 1.7v had no effect and the setup still refused to come to life. While I am disappointed with the overclocking experience of the GSkill Trident kit, one has to remember that when a kit is preset to it's absolute limit, anything extra is a bonus. Sadly this is one area where I would not expect the GSkill to thrill the end user.

http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/3



However, this review lends some weight to Musho statement above it being better to buy higher frequency albeit higher latency RAM than vice versa. I need to read some more reviews but the comments at NewEgg are a mixed bag. Some can't get this RAM to run at spec., other are reporting that it runs at DDR3-2000 CL8.

[XII]

As for the Noctua, I`ve modded mine with a scissor,
so it fits on a Classy.

Doesn`t look pretty, , but works

[bofors]

I am also taking a another look at the Mushkin Ridgeback 998826 kit.

This photo posted by Fitseries3 in his review thread on this kit implies that it has no clearance problems with the Noctua NH-D14:

http://www.xtremesystems.org/forums/...d.php?t=250809




Now that I have looked at some results posted for the G.SKILL Trident F3-16000CL9T-6GBTD kit, which were at best DDR3-3000, CL 8 at 1.65 volts, these results by Fitseries3 with Mushkin Ridgeback 998826 looks better than what I think I could expect with the Tridents:

12gb @ 1950mhz 7-9-7-27 1t 1.68v vdimm

However, I still need to look at a few more reviews.

[Musho]

I was just checking a review on the G.Skill Trident F3-16000CL9 3x2GB kit at OverClock3D.net:
http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/6

They posted a good cross section picture, so I can what I am getting into with heatspreader removal and/or dremeling:

http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/2



That does not look bad at all, but I am much less enthused about this G.Skill Trident F3-16000CL9 3x2GB kit after reading:

As you can see from the shot above, all I could manage was a miserly 26MHz over stock. While I didn't fully determine stability it passed SuperPI and 3DMark06 at this speed. However when the bandwidth was pushed any higher sporadic BSOD began to appear. Lowering the latency had an even worse effect with anything lower than CAS 9 resulting in the Gigabyte motherboard failing to POST, even lowering the bandwidth down to 1866MHz and raising the Vdimm to 1.7v had no effect and the setup still refused to come to life. While I am disappointed with the overclocking experience of the GSkill Trident kit, one has to remember that when a kit is preset to it's absolute limit, anything extra is a bonus. Sadly this is one area where I would not expect the GSkill to thrill the end user.

http://www.overclock3d.net/reviews/m...cl9_ddr3_kit/3



However, this review lends some weight to Musho statement above it being better to buy higher frequency albeit higher latency RAM than vice versa. I need to read some more reviews but the comments at NewEgg are a mixed bag. Some can't get this RAM to run at spec., other are reporting that it runs at DDR3-2000 CL8.

Worst thing you can do is to read newegg reviews on high frequency RAM A lot of people will suffer uncore instability, rather than RAM instability, because the uncore multiplier needs to 2x the RAM multiplier. It's going to be a lot easier with your 980X as that one only requires and uncore multiplier of 1,5x the RAM multiplier, making is easier to run higher frequency RAM. Trust me, all brands on newegg have bad reviews on their high frequency kits of people unable to run them at rated speeds, just because their uncore is failing them, rather than the RAM (In most cases atleast).

As for overclocking the kit, I've seen a mixed bag of results as well. Could be due to the fact there are just some bad kits out there, could be due to motherboard not playing nice with the RAM, a bad overclocker, etc
But then again, ask yourself this: Do you really need even higher performance than this? Application speed already barely scales with better RAM, and this is already far faster than the average DDR3 RAM out there. Just take some overclocking headroom as a bonus, rather than to require it. Unless the rig is for benching and HWbot points, you won't need headroom. It's nice to have, but not a necessity. Just enjoy the good performance/price ratio, and as long as it runs atleast the rated timings, I wouldn't worry too much

Edit: I have a W3530 and X58A-UD5 sitting next to me here, but I can't use them yet because I don't have any RAM. As a matter of fact, I ordered the Trident kit. Should arrive here wednesday or so. Will post some results

[bofors]

Worst thing you can do is to read newegg reviews on high frequency RAM
...
Trust me, all brands on newegg have bad reviews on their high frequency kits of people unable to run them at rated speeds, just because their uncore is failing them, rather than the RAM (In most cases atleast).

Yes, I know what NewEgg reviews are good for.... and what they are not good for. That is why I come to XtremeSystems for advice before I buy performance RAM!

A lot of people will suffer uncore instability, rather than RAM instability, because the uncore multiplier needs to 2x the RAM multiplier. It's going to be a lot easier with your 980X as that one only requires and uncore multiplier of 1,5x the RAM multiplier, making is easier to run higher frequency RAM.

EDIT: After doing my homework, I now know that the the QPI and memory multipliers are not locked on non-Extreme Editions Bloomfields, and that Xeon W3500 series chips have proven to be as good or better overclockers than Core i7 Extreme Editions.

Yes, I am learning that there are a lot of reasons to go with the Core i7-980X like this 1.5x versus 2x multiplier requirement versus other versions of the Gulftown (which at this point are only available as engineering samples). Previously I have avoided paying for Extreme Edition CPUs and been happen with lower priced Xeons.

The additional dimensions of flexibility with the x58 architecture seem much more important to overclocking. For example, I have read that the ability to lower the QPI multiplier can be an important in achieving a safe QPI voltage, high bandwidth overclock too. So, it appears that this not just about unlocking a single CPU multiplier anymore.

As for overclocking the kit, I've seen a mixed bag of results as well. Could be due to the fact there are just some bad kits out there, could be due to motherboard not playing nice with the RAM, a bad overclocker, etc

Motherboard issues are actually something I am increasingly concerned about with the DX58SO, and I am not getting the latest and presumedly best version of it either. Although Intel officially rates it for DDR3-1600 and even certifies some DDR3-1866 kits for it:
http://www.intel.com/products/deskto...O-overview.htm
http://www.intel.com/en_US/Assets/PD...i7_ww18_09.pdf

... I know someone who had to get DDR3-1066 last week to make a problematic revision of the DX58SO to work. Other people with DX58SO experience are suggesting that I stay away from DDR3-2000 for it.

So, for these reasons and the others I posted above, I am probably going to get the Mushkin Ridgeback DDR3-1600, CL 6 kit, although I suspect that I would be perfectly happy with the G.Skill Tridents (except for dremeling one if necessary).

But then again, ask yourself this: Do you really need even higher performance than this? Application speed already barely scales with better RAM, and this is already far faster than the average DDR3 RAM out there. Just take some overclocking headroom as a bonus, rather than to require it. Unless the rig is for benching and HWbot points, you won't need headroom. It's nice to have, but not a necessity. Just enjoy the good performance/price ratio, and as long as it runs atleast the rated timings, I wouldn't worry too much

Right, this is almost certainly an exercise in overkill. I have almost almost no interest in benchmarking other than posting my results for comparison. As far application speed goes though, which is what I am most concerned about, the custom software I use has not been bandwidth intensive. So, I am more concerned with latency (although I have noticed that higher speed DDR3 RAM has lower latency).

However, for my own intensive professional use and satisfaction, I found that higher end performance RAM is worth the extra money, even though I am not Xtreme. I would have rather RAM that is a little better than what I actually need than vice versa.

Thanks again for your valuable input!

[Musho]

RAM is taking ages. I just got my Track & Trace number. Will hopefully get it tomorrow. I will post some results here, but I will probably focus on clocking my CPU before messing with my RAM. I'll let you know soon (hopefully!). How are you liking your Mushkins?

[bofors]

RAM is taking ages. I just got my Track & Trace number. Will hopefully get it tomorrow. I will post some results here, but I will probably focus on clocking my CPU before messing with my RAM. I'll let you know soon (hopefully!). How are you liking your Mushkins?

As you might have gathered, I had a change of heart here regarding spending the $$$ for a Core i7-980X. The used market for them is too strong now. I did not really want to spend more than about $750 or so. I also know that discounted Gulftowns (e.g. like the Core i7-970) will be coming out in a matter of weeks or months.

Instead, I have more carefully studied the situation with Bloomfield overclocking and narrowed it down to either ordering a new W3520 or a new W3530 (which I know is what is you ordered). It was a tough decision, but I thought the W3520 was a slightly better bet given some of the less than stellar Core i7-930 results and bad batchs being reported, and that any W3520 was likely to be from at least a good D0 batch. In short, I did not think the 1x multiplier difference was significant compare to potential for significant batch variation. It was also hard to find W3530 overclocking results posted: http://www.xtremesystems.org/forums/...d.php?t=252295

I have had my Mushkins Ridgebacks, Noctua NH-D14, Indigo Xtreme and DX58SO for days, but the rest of the parts I needed for this build just showed up: Lian Li PC-9 case and Corsair HX750 PSU. Hence, I now actually have the hardware I need to test the Mushkin Ridgebacks (I have been sticking with Intel D975XBX2 boards for years because they are good for running Mac OS X).

However, I got lucky last night and found two "gem batch" (3849A) W3520's for sale last night. I bought them both for $309 shipped each:
http://forums.overclockers.com.au/sh...0&postcount=33

I am now planning on sending back (declining receipt) of the new W3520 I had ordered which was just dropped off. If I do, until the two "gem batch" (3849A) W3520's early next week, I will not really be in a position to test the Mushkins Ridgebacks.

(Bye the way, I have flagged this thread twice and requested that it be move the proper forum section now.)

[BeepBeep2]

You could have gotten these:
http://www.ewiz.com/detail.php?name=...nel-Memory-Kit

Wow I'm on a super talent RAM page today...

[bofors]

You could have gotten these:
http://www.ewiz.com/detail.php?name=...nel-Memory-Kit

You really think that SuperTalent CL 8, DDR3-1600 is better Mushkin CL 6, DDR3-1600?

EDIT: As BeepBeep2 point out below, I misread the spec. on this SuperTalent RAM. It is actually PC3-16000, DDR3-2000.

Wow I'm on a super talent RAM page today...

Perhaps a virtual rampage?

[BeepBeep2]

You really think that SuperTalent CL 8, DDR3-2000 is better Mushkin CL 6, DDR3-1600?



Perhaps a virtual rampage?

Fixed it. I suppose that SuperTalent CL8 would do about 1700 6-6-6-24, 1900 7-7-7. Maybe more around 1.7v...

[Shiranui Gen-An]

You could have gotten these:
http://www.ewiz.com/detail.php?name=...nel-Memory-Kit

Wow I'm on a super talent RAM page today...

That's probably one of the best DDR3 deals around right now. 6GB of Elpida Hyper for $222?! I'd better grab a kit before they all get snatched up.

[BeepBeep2]

That's probably one of the best DDR3 deals around right now. 6GB of Elpida Hyper for $222?! I'd better grab a kit before they all get snatched up.

The 4GB kit is $150 and 4GB 2200 C8 is $200.

[bofors]

Fixed it. I suppose that SuperTalent CL8 would do about 1700 6-6-6-24, 1900 7-7-7. Maybe more around 1.7v...

I see, that sounds like some good RAM, but the Muskin Ridgesback I have were reported to do 1950 MHz 7-9-7-27 1t at 1.68v for about the same price:
http://www.xtremesystems.org/forums/...d.php?t=250809

Also, I mentioned above, since I am using a DX58SO that may be picky about RAM SPD being at most DDR3-16000, and since I have no first hand experience with that board or DDR3 RAM for that matter, I decided to avoid DDR3-2000 RAM for the moment.

However, it is clear to me, that DDR3-2000 RAM is where most of the good deals in performance RAM are right now. So, I may be picking up a DDR3-2000 RAM kits, and perhaps the Super Talent CL 8 kit that is being recommended here, after I feel out the Intel DX58SO.

Thanks for your input and correcting my mistake.

[Shiranui Gen-An]

The Super Talent kit will be able to do tighter timings than the Ridgebacks due to better ICs. DDR3-2000 7-8-7 or maybe even DDR3-2100 8-9-8 is likely possible with more voltage. And for nice 24/7 settings they should be able to do DDR3-1600 6-7-6 easily.

[Praz]

Fixed it. I suppose that SuperTalent CL8 would do about 1700 6-6-6-24

Does 1730/6-6-6-20 stable on the Crosshair IV Formula.

[BeepBeep2]

Theres also a 2GB 1600 6-6-6 stick for $65, three of them would have been ~$200.

EDIT:

All of SuperTalent's memory is made in the USA also...

[bofors]

The Mushkin Ridgeback RAM has very nice, solid, heavy heatspreaders. These are not merely taped on like traditional heatspreaders, they use screws.







They clear the Noctua NH-D14 by about 3mm. Is going to be a little tricky to get some fans on the DIMMs because with the DX58SO they are near the top of the case.





I have yet to run do anything with these Mushkin Ridgebacks beyound posting the machine at spec: DDR3-1600, CL 6. Otherwise, I saw no sign of complaint from the DX58SO about the memory which I was somewhat concerned about happening. The XMS settings were an option in BIOS that immediately worked.