Lets try another memory kit (#4) with Elipida memory chips, this time the memory kit comes from Elpida themselves

This kit uses Elpida BDSE ICs, they are supposed to be decent but nothing spectacular.

.

This (#43) Elpida 10600U (1333MHz) kit is a dual ranked 2x2GB memory kit with Elpida 1Gbit BDSE chips.

The highest JEDEC specification the modules officially can handle is 1333 (666ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V. Just like the 2Gbit BCSE chips on the Kingston (#3) kit tested earlier these 1Gbit BDSE ICs found on this Elpida kit has the potential to be decent, lets hope they are from a better bin!

.

First the 1333MHz SPD auto settings.

SPD @ 1333 CL9-9-9-24-74-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

SPD @ 1333 CL9-9-9-24-74-1. Frybench x86.

SPD @ 1333 CL9-9-9-24-74-1. SuperPI 32M.

SPD @1333 CL9-9-9-24-74-1. Doom III 1024*768 "Ultra".

.

No real surprises this time but I will keep posting the results I get with auto memory settings. There will be some unexpected results sooner or later.

Edit

Improved testing methodology, results updated. The benchmark scores are just as expected.

/Edit

Now for the interesting part.

After testing that awful Kingston kit I did not have very high hopes for these modules. Sure they are from Elpida themselves but it's a bog-standard 10600U kit intended for some random boring office PC. I think we can be sure that this vanilla kit isn't made from a good bin of 1Gbit BDSE, the question is more how bad is the lower quality grades these chips.

Decent bin or not it would be a disappointment if this memory kit didn't overclock better than the Kingston (#3) one.

.

.

The best overclocked settings I could achieve with this memory kit was 2133 MHz with 9-10-9-24-72-1 (TRRD4 TFAW16). It can hardly be called particularly great result but on the other hand it's good enough for some decent performance. To really perform their best these Elpida 1Gbit offerings need a level of voltage I'm not willing to give them.

.

Lets see what kind of performance the overclock resulted in.

OC 2133 MHz CL9-10-9-24-72-1 (TRRD4 TFAW16). Aida64 Cache & Memory Benchmark plus PhotoWorxx.

OC 2133 MHz CL9-10-9-24-72-1 (TRRD4 TFAW16). Frybench x86.

OC 2133 MHz CL9-10-9-24-72-1 (TRRD4 TFAW16). SuperPI 32M.

OC 2133 MHz CL9-10-9-24-72-1 (TRRD4 TFAW16). Doom III 1024*768 "Ultra".

.

The results are nice! Even if the overclock isn't anything really spectacular it still brings a very nice improvement over the stock settings.

Edit

Testing methodology improved, results updated. Bettered a few of the secondary timings as well.

/Edit

Lets continue with the value offerings, this time actual Kingston value ram (#5).

I can not say for certain what kind of 2Gbit IC this kit is built from but I can make a guess. It is known that Kingston used 2Gbit PSC chips for some of their value ram during 2012. This kit is from late 2012 and the ICs do not scale from extra voltage at all, 2Gbit PSC chips are known to not scale from voltage.

.

.

This (#5) Kinston value ram (1333MHz) kit is a single ranked 2x2GB memory kit with what is likely PSC 2Gbit chips.

The highest JEDEC specification the modules officially can handle is 1333 (666ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V.

.

Lets see how they perform with auto settings.

SPD @ 1333 CL9-9-9-24-107-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

SPD @ 1333 CL9-9-9-24-107-1. Frybench x86.

SPD @ 1333 CL9-9-9-24-107-1. SuperPI 32M.

SPD @1333 CL9-9-9-24-107-1. Doom III 1024*768 "Ultra".

.

The stock auto SPD results are as expected.

Edit
This kit has been retested.
/Edit

With luck the overclocked results with these (#5) modules are more exciting!

When dealing with unknown memory ICs you always hope for a nice surprise. No nice surprise this time, the ICs behave like 2Gbit PSC and those are not great, on the other hand it could have been a lot worse.

.

.

The best settings I could squeeze out of these (#5) were 1866 MHz with 9-9-8-20-100-1 (TRRD4 TFAW16) timings. I'm pretty happy with the result as there is zero scaling with extra voltage with these memory modules. The max speed and timings is yet another hint that this very likely could be 2Gbit PSC memory ICs.

.

Lets see if the performance improved.

OC 1866 MHz CL9-9-8-20-100-1 (TRRD4 TFAW16). Aida64 Cache & Memory Benchmark plus PhotoWorxx.

OC 1866 MHz CL9-9-8-20-100-1 (TRRD4 TFAW16). Frybench x86.

OC 1866 MHz CL9-9-8-20-100-1 (TRRD4 TFAW16). SuperPI 32M.

OC 1866 MHz CL9-9-8-20-100-1 (TRRD4 TFAW16). Doom III 1024*768 "Ultra".

.

The performance did in fact improve! 😁

Memory kit #6 is identical to kit number #1.

Let us see if anything differs when it comes to the results!

.

.

Lets repeat the spec. Memory kit #6 is a dual ranked 2x2GB kit built around Samsung 1Gbit E-die memory chips with HCH9 (10600) speed bin.

The best JEDEC specification the modules officially can handle is 1333 (666ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V.

.

Lets see if the SPD @1333 results match the other kit (#1).

SPD @1333 CL9-9-9-24-74-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

SPD @1333 CL9-9-9-24-74-1. Frybench x86.

SPD @1333 CL9-9-9-24-74-1. SuperPI 32M.

SPD @1333 CL9-9-9-24-74-1. Doom III 1024*768 "Ultra".

.

It seems the stock results match the other identical memory kit perfectly.

Now is the question how does memory kit #6 overclock compared to the identical number #1 kit.

It's fully normal that the overclocking results vary a bit going from one kit to another even if they are the same kit model with the same typ of memory ICs.

.

.

This isn't normal, this kit maxes out at the exact same settings as the other identical kit. I lied, not exactly the same settings, this kit can handle a tREF setting of 60 while the other kit could only handle 61... The best setting for this I could find was 1866MHz CL8-10-9-20-60-1 TRRD4 TFAW16 @1.65V.

.

Lets see if the overclocked performance match the other kit.

DDR3 1866 CL8-10-9-20-60-1 TRRD4 TFAW16. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

DDR3 1866 CL8-10-9-20-60-1 TRRD4 TFAW16. Frybench x86.

DDR3 1866 CL8-10-9-20-60-1 TRRD4 TFAW16. SuperPI 32M.

DDR3 1866 CL8-10-9-20-60-1 TRRD4 TFAW16. Doom III 1024*768 "Ultra".

.

Once again the results match the other Samsung E-die kit perfectly! 😀

So far every memory kit has performed as expected running "auto settings" letting the motherboard use the modules SPD profiles.

Let us test a kit that doesn't perform as expected. Memory kit #7, some leftovers Micron found at the bottom of the bin.

.

.

This single ranked Micron 2x2GB 10600U kit (#7) uses Microns own 2Gbit D9LGK chips. The D9LGK was probably Microns first retail revision of 2Gbit memory IC.

The best JEDEC specification the modules officially can handle is 1333 (666ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V.

.

Lets see how they perform using the "auto" SPD settings.

DDR3 1333 CL9-9-9-24-107-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

DDR3 1333 CL9-9-9-24-107-1. Frybench x86.

DDR3 1333 CL9-9-9-24-107-1. SuperPI 32M.

DDR3 1333 CL9-9-9-24-107-1. Doom III 1024*768 "Ultra".

.

The performance is just like similar single ranked 10600U kits.

Edit
This kit has been retested. The odd earlier results could not be reproduced. Unlucky memory training or BIOS hiccup most likely.
/Edit

After a busy week vacation is finally here so naturally time to hide in the basement a bit. 😀

Lets see if overclocking once more can save a hopeless memory kit (#) or if this single ranked Micron 2x2GB kit with 2Gbit D9LGK chips still misbehaves.

.

.

The best settings this Micron memory kit (#7) can handle is 1600 MHz CL10-8-7-22-68-1 TRRD4 TFAW16 @1.65V. Not being able to handle at least CL9 at 1600 MHz at 1.65V is mediocre but the more reasonable tRCD8 and tight tRP7 sort of makes up for it.

.

Lets see if the kit at least gives reasonable results in all benchmarks when overclocked.

OC 1600 MHz CL10-8-7-22-68-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

OC 1600 MHz CL10-8-7-22-68-1. Frybench x86.

OC 1600 MHz CL10-8-7-22-68-1. SuperPI 32M.

OC 1600 MHz CL10-8-7-22-68-1. Doom III 1024*768 "Ultra".

.

While the overclocking results are nothing to brag about the results are at least reasonable.

Edit
This memory kit was retested, and scores updated.
/Edit

Memory kit #8 is the Second Corsair Dominator "DHX Pro" (CMP4GX3M2A1600C8) 2x2GB kit, this time with version number 3.5.

Oddly enough while the first (#2) Dominator kit was a 2x2GB single ranked kit, this kit (#8) with the exact same product number is dual ranked...

.

.

This (#8) 2x2GB Corsair Dominator "DHX Pro" CMP4GX3M2A1600C8 ver3.5 kit uses a dual rank layout with 1Gbit Micron ICs.

The highest JEDEC specification the modules officially can handle is 1333 (666ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V. The modules have an XMP profile @1600 MHz with 8-8-8-24-41*-2 timings @1.65V.

*TRC not tREF.

.

This is the last 2x2GB dual ranked kit with 1333 9-9-9 -24 SPD I will make a separate post with SPD results for unless I find one that Asus memory tables treat notably differently. These results earned a spot by showing the impact the version number has on the otherwise exact same Corsair product. I will still add SPD results for all kits to the result lists in post number 3.

1333MHz CL9-9-9-24-74-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

1333MHz CL9-9-9-24-74-1. Frybench x86.

1333MHz CL9-9-9-24-74-1. SuperPI 32M.

1333MHz CL9-9-9-24-74-1. Doom III 1024*768 "Ultra".

.

As expected the results show that the dual ranked version of this Corsair Dominator kit is clearly superior compared to the single ranked version when using auto settings.

Lets see if the dual ranked version of the Corsair Dominator kit (#8) still is superior compared to the single ranked version (#2) when using the XMP profile.

.

.

The XMP profile for this (#8) dual ranked Corsair Dominator CMP4GX3M2A1600C8 ver3.5 with 1Gbit Micron ICs is @1600 MHz with 8-8-8-24-41*-2 timings @1. 65V.

*TRC

.

XMP 1600 CL8-8-8-24-88*-2. Aida64 Cache & Memory Benchmark plus PhotoWorxx. *tREF

XMP 1600 CL8-8-8-24-88*-2. Frybench x86

XMP 1600 CL8-8-8-24-88*-2. SuperPI 32M.

XMP 1600 CL8-8-8-24-88*-2. Doom III 1024*768 "Ultra".

.

Yes, the dual ranked version of this Corsair Dominator 1600MHz CL8 kit is still superior compared to the single ranked version when using the XMP profile.

In my head selling different performing products with the same exact product name is a bad idea.

At least, Corsair label memory by version number and provide actual chip information on their forums.

The Serpent Rider wrote on 2023-07-11, 13:20:

At least, Corsair label memory by version number and provide actual chip information on their forums.

Yes, it's better than nothing I guess. I bet a few non geeks have wondered why their identical memory kits won't overclock together.

The kits will probably run the advertised 1600 8-8-8-24 XMP profile even if disadvantageously combined as XMPs with command rate 2 will often end up treated as command rate 3 when more than 2 ranks of memory per channel is used...

Asus X79 boards may even slack the command rate of the XMP profile one step if you use 1 identical dual ranked module per channel but sold as two dual channel kits and not a single 4 channel kit. It's rarely a good idea to use an XMP profile... 😀

Lets see if the dual ranked (#8) Dominator kit still comes out on top when overclocked.

.

.

The best overclock I could archive with the (#8) dual ranked 2x2GB Corsair Dominator ver3.5 kit with 1Gbit Micron ICs was 1866 MHz with 9-9-9-22-58-1 TRRD4 TFAW16 timings @1.65V. This strongly indicates that this memory kit uses D9KPT or D9KPV memory chips.

.

Lets see how the overclocked (#8) dual ranked Dominator ver3.5 kit performs.

DDR3 1866 CL9-9-9-22-58-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

DDR3 1866 CL9-9-9-22-58-1. Frybench x86.

DDR3 1866 CL9-9-9-22-58-1. SuperPI 32M

DDR3 1866 CL9-9-9-22-58-1. Doom III 1024*768 "Ultra"

.

Well with the risk of sounding like a broken record... Yes the dual ranked (#8) Dominator ver3.5 Micron kit still beats the single ranked (2) Dominator ver8.14 Nanya kit. This doesn't mean that there are no instances when a single ranked kit will be a better choice than a dual ranked kit when overclocking.

Lets take a look at a promising single ranked 2x2GB memory kit.

This (#9) Samsung 12800E single ranked 2x2GB ECC memory kits uses 2Gbit D-die memory chips.

.

.

These (#9) modules are single ranked with 8x 2Gbit "D-die" K4B2G0846D-HCK0 memory chips per module.

The highest JEDEC specification the modules officially can handle is 1600 MHz with CL 11, tRCD 11, tRP 11 and tRAS 28 timings at 1.5V. The type of memory chip (2Gbit D-die) that this (#9) 12800E single ranked Samsung kit uses is one of the better DDR3 ICs ever made when it comes to more hardcore overclocking. Extreme results takes extreme voltages and preferable a known to be good bin of said memory chips so its nothing we will see here. These chips are from Samsungs 1600 MHz HCK0 sorting. If Samsung sorts/bins the memory as 1333 MHz (HCH9) or 1600 MHz (HCK0) do not seem to matter for overclocking but the bins within bins do. It's nice to know that there is potential for the memory to overclock well.

.

Lets see how the 1600 MHz CL11 SPD profile performs.

DDR3 1600 CL11-11-11-28-128-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

DDR3 1600 CL11-11-11-28-128-1. Frybench x86.

DDR3 1600 CL11-11-11-28-128-1. SuperPI 32M.

DDR3 1600 CL11-11-11-28-128-1. Doom III 1024*768 "Ultra".

.

The stock SPD performance is on pair with the dual ranked 1333 MHz 10600U kits.

I think overclocking might improve this (#9) single ranked Samsung 2Gbit D-die kits position in the pecking order.

Lets see if a single ranked kit with a decent overclock have a chance against the overclocked dual ranked kits.

(#9) Samsung single rank 2x2GB 12800E with D-die HCK0 memory ICs.

.

.

The best settings I could find for this (#9) memory kit was 2400 MHz with 9-11-11-21-96-1 TRRD4 TFAW20 timings @1.65V.

This is probably a decent overclocking result. It's nice with CL9 at 2400 MHz but needing to slack TFAW is not ideal.

.

Is it enough to top the result lists (for now)?

DDR3 2400 CL9-11-11-21-96-1 TRRD4 TFAW20. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

DDR3 2400 CL9-11-11-21-96-1 TRRD4 TFAW20. Frybench x86.

DDR3 2400 CL9-11-11-21-96-1 TRRD4 TFAW20. SuperPI 32M.

DDR3 2400 CL9-11-11-21-96-1 TRRD4 TFAW20. Doom III 1024*768 "Ultra".

.

A slight win in SuperPI 32M and a draw in Frybench compared to the best (for now) overclocked dual ranked kit is good as these benchmarks really prefers dual ranked memory.

Doom III loves both high memory speed and low latency but the performance doesn't hurt as much when using single ranked memory. This (#9) single ranked Samsung 2Gbit D-die kit totally blows the other kits tested up until now out of the water.

This shows how great Micron modules and OEM memory in regard, are and I used these Micron and Crucial all the time on my computers. Not overclocking through.

Keep in mind that Corsair and others are third party and they do not have the access to top grade chips. They were re-binned from best chips after first binning by OEM makers they can get on the open market. Unless there is one that is exclusive deal with OEM maker of chips to get the top grade binnings. I wish there is known brand that does this. The best grade one for overclocking you can get and still original maker is Crucial's Ballistrix line from Micron who owns Crucial.

Cheers,

pentiumspeed wrote on 2023-07-11, 16:27:

This shows how great Micron modules and OEM memory in regard, are and I used these Micron and Crucial all the time on my computers. Not overclocking through.

Keep in mind that Corsair and others are third party and they do not have the access to top grade chips. They were re-binned from best chips after first binning by OEM makers they can get on the open market. Unless there is one that is exclusive deal with OEM maker of chips to get the top grade binnings. I wish there is known brand that does this. The best grade one for overclocking you can get and still original maker is Crucial's Ballistrix line from Micron who owns Crucial.

Cheers,

I actually think Corsair have had access to some great memory bins during the years but you have to pay top money for it. For anything Corsair reasonable priced you get bargain bin bins.

One example is some of the Dominator GT DDR3 line like those that used Elpidas top bins of 1Gbit BASE (MNH-E/MGH-E Hyper). This is still considered to be the best DDR3 overclocking memory for Sandy Bridge and everything older than that. While the first version, the MNH-E has a well deserved reputation of being fragile I still have some Corsair Dominator GT 2000 8-8-8 with BASE MGH-E Hyper and considering that the 2000 8-8-8 wasn't even close to a top bin Corsair must have had access to some really nice BASE bins.

Microns Crucial also released some nice DDR3 memory and they got Microns top memory bins for sure. The next kit I tested was actually a Crucial Ballistix kit, probably not a "top bin" but nice memory none the less.

The next memory kit up for testing (#10) is a single ranked Crucial Ballistix 2x2GB kit.

These Crucial modules are marked "1600MHz 9-9-9-24 1.5V" and I figured that this was the XMP profile spec.

.

.

This (#10) 2x2GB Crucial Ballistix kit uses a single rank layout with 2Gbit Micron ICs, likely D9PFJ.

The highest JEDEC specification the modules officially can handle is 1600 (800ddr) MHz with CL 9, tRCD 9, tRP 9 and tRAS 24 timings at 1.5V. The modules have an XMP profile @1600 MHz with the same timings at the same voltage but with command rate 2... Let us totally ignore the XMP profile...

.

Lets us see what kind of performance this relatively aggressive SPD profile offer.

SPD @1600 CL9-9-9-24-128-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

SPD @1600 CL9-9-9-24-128-1. Frybench x86.

SPD @1600 CL9-9-9-24-128-1. SuperPI 32M.

SPD @1600 CL9-9-9-24-128-1. Doom III 1024*768 "Ultra".

.

This memory is really great for people that do not want to get involved with memory settings at all. Just leave everything at auto and let it sort it self, the performance is still somewhat decent.

For OEM systems upgrading to memory using the tighter of the JEDEC SPD timings is often the only way to improve memory performance. As this kit is single ranked using two kits (4 modules) for a total of two ranks of memory per memory channel is likely ideal when not overclocking / for OEM systems.

Will the nice performance using the "auto" SPD settings translate to good overclocked performance?

Lets see what this (#10) single ranked Crucial Ballistix 2x2GB kit can do with the speed brakes turned off.

.

.

The best overclock this (#10) Crucial kit could handle was 2133MHz with 9-11-10-27-84-1 tRRD4 tFAW16 @1.65V.

This is arguably the best Micron kit so far.

.

Lets see the numbers! (for easy viewing in post number 3 on the first page)

2133MHz 9-11-10-27-84-1 TRRD4 TFAW16. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

2133MHz 9-11-10-27-84-1 TRRD4 TFAW16. Frybench x86.

2133MHz 9-11-10-27-84-1 TRRD4 TFAW16. SuperPI 32M.

2133MHz 9-11-10-27-84-1 TRRD4 TFAW16. Doom III 1024*768 "Ultra".

.

This (#10) memory kits best feature is with out a doubt the tight SPD profile but the overclocked performance isn't bad either.

The secondary timings not shown like "WRITE Recovery Time" and "CAS Write Latency" could be set really tight with this kit.

It's time to take a look at the memory kit that caught my eye and made me make a bid on the memory scrap lot.

This fancy (#11) G.Skill 2x4GB RipjawsX kit.

.

.

This (#11) G.Skill RipjawsX F3-17000CL11D-8GBXL 2x4GB kit is dual rank and is using Samsung 2Gbit memory chips.

The best JEDEC specification this kit can officially handle is 1600 MHz with 11-11-11-28 timings. There is an XMP profile @ 2133 MHz with 11-11-11-30-41(tRC) timings. When I saw this G.Skill memory kit with a Samsung serial number in the scrap lot I was hoping for 2Gbit D-die. After doing some research it seems much more likely that this kit with its September 2014 manufacturing date is using Samsung 2Gbit E-die. Samsung 2Gbit E-die do not overclock as well on average as 2Gbit D-die but it's still a well performing memory IC.

.

Lets us see how this kit behaves using "auto settings".

1600MHz CL11-11-11-28-128-1. Aida64 Cache & Memory Benchmark plus PhotoWorxx.

1600MHz CL11-11-11-28-128-1. Frybench x86.

1600MHz CL11-11-11-28-128-1. SuperPI 32M.

1600MHz CL11-11-11-28-128-1. Doom III 1024*768 "Ultra".

.

This (#11) dual ranked 2x4GB memory kit do well in Frybench and SuperPI using the SPD profile. The Doom III result is also decent.