Super Talent 2200C8 @DDR3-2392Mhz Elpida Hyper ICs FTW

[eva2000]

3x 2GB Super Talent Speed 2200C8 (made up of 2x2GB dual channel kits WS220UX4G8)

@DDR3-2058Mhz 7-7-7-21 at 1.66v​

@DDR3-2214Mhz 8-8-8-24 1T at 1.70v​

@DDR3-2384Mhz 8-8-8-24 1T at 1.70v​

Max screen shot

@DDR3-2392Mhz 8-8-8-24 1T at 1.70v​

Max CPUZ Validation





System:

• Intel Core i7 980X 3951A763 B1 Step 2 ES
• EK Supreme LT H20 + TFC480 + MCP655
• Asus P6T6 WS Revolution 0609 beta bios
• Sapphire HD5870 10.4 Official
• 3x 2GB Super Talent Speed 2200C8 (made up of 2x2GB dual channel kits WS220UX4G8)
• 750GB Samsung SATAII (OS)
• 120GB OCZ Vertex SSD (DATA)
• Pioneer DVR216 Burner
• Corsair HX1000
• Windows 7 Ultimate 64bit RTM (fully updated)

 

[VirtualLarry]

750GB Samsung SATAII (OS)
120GB OCZ Vertex SSD (DATA)

Why the mechanical HD for OS, and the SSD for DATA? Isn't the usual arrangement the opposite of that?

 

[eva2000]

Just the 750GB SATA had the OSes I needed on it dual boot winxp and win7 - needed access to xp for testing. Original system config has 256GB Crucial C300 SSD for OS with just Win7.

 

[AdamK47]

Do they only come in dual channel kits? Why not triple channel kits? Seems kinda silly when they appear to work well with the 980X on the X58.

 

[eva2000]

2200C8 only come in dual channel kits as P55 platform and dual channel is much easier to hit 2200C8 compared to X58 triple channel. But i always like a challenge i.e. 6x2GB A-Data 2000X (3 dual channel kits) for 12GB ram clocking http://i4memory.com/f83/12gb-data-2000x-ddr3-testing-20350/

 

[eva2000]

12GB Super Talent 2200C8 @6-6-6-20 1t

Quick runs seeing as 2 modules out of 6 can do 900mhz 6-6-6-20 1T at 1.7v memtest86+ v4.10 and the other 4 modules only do 894mhz 6-6-6-20 1T at 1.7v memtest86+ v4.10, then for just single 32M Pi test, 6x2GB Super Talent 2200C8 falling in around:

• 879Mhz 6-6-6-20 1T at 1.7v with CPU VTT 1.4125v
• 885mhz 6-6-6-20 1T at 1.7v with CPU VTT 1.425v
• 889mhz 6-6-6-20 1T at 1.72v with CPU VTT 1.4375v
• 894mhz 6-6-6-20 1T at 1.72v with CPU VTT 1.44375v

Bootable at 879Mhz 6-6-6-20 1T at 1.66v with CPU VTT 1.3875v​

879Mhz 6-6-6-20 1T at 1.7v with CPU VTT 1.4125v​

885mhz 6-6-6-20 1T at 1.7v with CPU VTT 1.425v​

889mhz 6-6-6-20 1T at 1.72v with CPU VTT 1.4375v​

894mhz 6-6-6-20 1T at 1.72v with CPU VTT 1.44375v​

900mhz 6-6-6-20 1T at 1.70v with CPU VTT 1.425v​

System:

• Intel Core i7 980X 3951A763 B1 Step 2 ES
• EK Supreme LT H20 + TFC480 + MCP655
• Asus P6T6 WS Revolution 0609 beta bios
• Sapphire HD5870 10.4 Official
• 6x 2GB Super Talent Speed 2200C8 (made up of 2x2GB dual channel kits WS220UX4G8)
• 750GB Samsung SATAII (OS)
• 120GB OCZ Vertex SSD (DATA)
• Pioneer DVR216 Burner
• Corsair HX1000
• Windows 7 Ultimate 64bit RTM (fully updated)

 

[n7]

More lovely results, as usual.

I wish i got to play with the toys you do.

 

[Rubycon]

Can you get those into fiveland with more down to earth speeds? Latency / throughput is nearly the same without need for 4GHz+ uncores. That's the nice thing about the 980X is you can run uncore at 1.5X (instead of 2X) mem!

 

[AdamK47]

Can you get those into fiveland with more down to earth speeds? Latency / throughput is nearly the same without need for 4GHz+ uncores. That's the nice thing about the 980X is you can run uncore at 1.5X (instead of 2X) mem!

Yeah, that's one of the good suprises I had with my 980X. Currently running my memory at 2040 DDR with the uncore at 3400MHz. I wouldn't be able to run my memory this high on any other i7.

 

[solofly]

I wish i got to play with the toys you do.

Don't we all...

 

[eva2000]

More lovely results, as usual.

I wish i got to play with the toys you do.

yeah expensive addiction though

Can you get those into fiveland with more down to earth speeds? Latency / throughput is nearly the same without need for 4GHz+ uncores. That's the nice thing about the 980X is you can run uncore at 1.5X (instead of 2X) mem!

Still mem frequency is king as cas5 won't get anywhere near the mem frequency needed to beat the mem frequency that cas 7,8 and 9 can do

Although you can push uncore frequency to regain some boost to bandwidth

next up up @DDR3-2462.8Mhz 8-8-8-24 1T at 1.74v with just 1.50v CPU VTT with water cooled i7 980X !

@DDR3-2000Mhz 8-8-8-24 1T at 1.64v​

First safe voltage for 8-8-8-24 1T clocks, how about 8x32M Pi @DDR3-2000Mhz 8-8-8-24 1T at 1.64v and 1.35v CPU VTT

@DDR3-2206Mhz 8-8-8-24 1T at 1.68v​

Now still under 1.4v CPU VTT, 8x32M Pi @DDR3-2206Mhz 8-8-8-24 1T at 1.68v and 1.39375v CPU VTT

8x32M PI complete

Now max screenshot and cpuz validations @DDR3-2436Mhz 8-8-8-24 1T and @DDR3-2462Mhz 8-8-8-24 1T both at 1.74v vdimm with 1.50v CPU VTT

@DDR3-2436Mhz 8-8-8-24 1T at 1.74v​

@DDR3-2462Mhz 8-8-8-24 1T both at 1.74v​

System:

• Intel Core i7 980X 3951A763 B1 Step 2 ES
• EK Supreme LT H20 + TFC480 + MCP655
• Asus P6T6 WS Revolution 0609 beta bios
• Sapphire HD5870 10.4 Official
• 3x 2GB Super Talent Speed 2200C8 (made up of 2x2GB dual channel kits WS220UX4G8)
• 750GB Samsung SATAII (OS)
• 120GB OCZ Vertex SSD (DATA)
• Pioneer DVR216 Burner
• Corsair HX1000
• Windows 7 Ultimate 64bit RTM (fully updated)

 

[Rubycon]

How about some real stability tests - 300+ minutes of LinX with updated Linpack binaries for Gulftown - using 8192MB size?

Running at lower CAS is nice IF you cannot push frequency too far. If it's erroneous it's of little use. Like a big bull with titties.

 

[eva2000]

huh. what's this thing called real stability ? bencher here - otherwise not much point in grabing 2200C8 rated memory

Knowing how much stress linx has on a system beyond everyday stress/loads and doing the above, is like asking a drag racer/formula one racer to test out the car's safety features/air bags by crashing his car into a wall at 300KM/hr

 

[mutz]

nvm .

 

[Rubycon]

huh. what's this thing called real stability ? bencher here - otherwise not much point in grabing 2200C8 rated memory

Knowing how much stress linx has on a system beyond everyday stress/loads and doing the above, is like asking a drag racer/formula one racer to test out the car's safety features/air bags by crashing his car into a wall at 300KM/hr

Some of us like to have our cake and eat it too!
I guess I should not complain about systems running 95&#37; speed-wise of the "bencher crowd" [sic] AND quite stable, i.e. <1 noticeable correctable error in 1TB rendered data. (essentially indistinguishable from a stock system!)

For comparing to drag racing well if the car runs to the end without blowing up they aren't pushing it hard enough!

 

[aigomorla]

How about some real stability tests - 300+ minutes of LinX with updated Linpack binaries for Gulftown - using 8192MB size?

Running at lower CAS is nice IF you cannot push frequency too far. If it's erroneous it's of little use. Like a big bull with titties.

Mah ruby he's a bencher... as long as it finishes, they are happy..

huh. what's this thing called real stability ? bencher here - otherwise not much point in grabing 2200C8 rated memory

Sick results!

Thanks for the preview eva.
This is Naekuh BTW.

 

[Rubycon]

Mah ruby he's a bencher... as long as it finishes, they are happy..



Sick results!

Thanks for the preview eva.
This is Naekuh BTW.

I know because the guys that handpick MY stuff hit crazy high numbers because they can cool things down to temperatures on Pluto. Bastards!

 

[AdamK47]

This thread is pretty much pointless then since the results are not on a stable system. Bah!

 

[aigomorla]

This thread is pretty much pointless then since the results are not on a stable system. Bah!

lulz...

sarcasm?

 

[AdamK47]

lulz...

sarcasm?

No. If the results were stable then it would be worth posting about.

 

[n7]

No. If the results were stable then it would be worth posting about.

Opinions are fun
But that doesn't make them right.

 

[mutz]

No. If the results were stable then it would be worth posting about.

yeah, that's kind of true, what can you do with memory which is running at 2400 non-stable :\?
i think it's been actually asked also earlier.

second thing is you cannot compare the results when thee CPU is running at different clocks,
they seem inconsistent anyhow, at 885 Super PI is running 8.36 at 4.248Ghz, at 889 it is running 8.37 at 4.269...
the actual effect on the results seems to come more from the clock speed then the memory timings, at the 2000+ you can't compare it too, the clock difference might be causing the difference.

if going for OC at least it should have a meaning, testing the system at the same clock with different timing and bandwidth for several times then averaging,
unless this is being done, there is no meaning for the session, showing numbers only has no real value.. -even for OCing actually-.

 

[extra]

No. If the results were stable then it would be worth posting about.

Whatever happened to doing things just for fun? I enjoyed reading this thread.

Don't you ever see what you can get windows to boot with on your cpu, even though it's not stable, just for fun? I've gotten my phenom 1090 to boot at 4.5ghz (on air lol, but it's in NO WAY stable at that..but it was fun to see it!)

 

[Rubycon]

Of course it's fun that's why we tinker, right?
Problem is it's like a highly addictive drug. Once you see it run really fast and tear through benchmarks you want it - really, really want it - to run at the speed every single day! Often you will wind up to go to great lengths to achieve this speed with highest stability. Just when you think you are out of the woods you hit a roadblock and have a setback. Talk about a tease and a half. Make that ten and a half teases. Anyone that's been there knows exactly what I am talking about.