Corsair Sets New Overclocking World Record

[John Doe]

What draw? Power consumption? Well they scale better, and P55 was plagued with crap Foxconn sockets anyways.

Yeah, 2000 Mhz is 2010, dude. We're almost @ 2012.

Thuban and BD can easily top 2000, as can Intel chips(Thuban prefers PSC it seems, anyway). Like I said, hypers are good for old hardware, and as a reviewer, I don't really have time to play with old stuff any more.

Hypers do about 2200, which is what PSC also does. And they do it at much lower latencies.

 

[sneekypeet]

yup that same crap socket took PCS chips over 2500 CAS9 with way more volts. I will say again The hypers have a higher draw!

 

[John Doe]

yup that same crap socket took PCS chips over 2500 CAS9 with way more volts. I will say again The hypers have a higher draw!

That's a ridicilious statement. It was a specific fault of the Foxconn socket on P55, nothing else. No other hardware would have blew up. RAM doesn't use significant amount of power, or put sufficient load to damage any modern hardware.

 

[sneekypeet]

so why does the socket blow with 1.7V through Hypers but not with 1.86V through my PSCs?

 

[cadaveca]

@ dave, it tells me manufacturers need to get off their bums and do something about Corsair taking year old stick and making a scene with then, even if limited by capacity of the stick. This is what I meant about laying the ground work, it proves there is still interest in insane memory speeds, even if it does nothing but increase epeen length.

Take a look at all the responses (5, or is it 6 now)that Athlon's benchmark/OC thread has . That's about the same level of interest that ram clocking has right now. Fact foothe matter is that I think guys over at Corsair are bored, and the only interesting things left to play with is the old stuff they have laying around from two years ago, not that interest is there in ram clocking now.

Otherwise, we'd be seeing more 2400 MHz ++ 4GB DIMMs. Heck, Corsair has problems getting 2133 MHZ 4GB DIMMs in stock. My refund is proof of that, because it makes no sense for them to refund me $550 when they could have just sent me a $300 kit if they had it...a kit that WOULD NOT have cost them the same $300. The very nearly paid me more than quadruple what the ram is truly worth, becuase they cannot get high-speed ICs in large quantities, not matter the density.

Why Corsair has this problem currently, I am not sure. G.Skill had no problem getting me a 16 GB 2133 MHz kit.

 

[John Doe]

so why does the socket blow with 1.7V through Hypers but not with 1.86V through my PSCs?

Because that socket on P55 is bork... it was going to blow either way, and it blew. There's no other platform that had that issue.

 

[sneekypeet]

Because that socket on P55 is bork... it was going to blow either way, and it blew. There's no other platform that had that issue.

nice rational cop out there. I guess you had a DMM hooked up to gauge the amp draw of the supplied voltage for both PSC and Hyper memory, and can prove me wrong on my assumption, or you are just going to sit here and call BS with no proof?

 

[John Doe]

nice rational cop out there. I guess you had a DMM hooked up to gauge the amp draw of the supplied voltage for both PSC and Hyper memory, and can prove me wrong on my assumption, or you are just going to sit here and call BS with no proof?

I'm not calling any of that "BS". Hypers pull more power and that's a fact. But the thing is, the amount of power RAM pulls doesn't matter. It was only your P55 with the defective socket that blew up. Show me any other platform with blown sockets, can you? You can't. Because P55 is the only platform with the socket issue.

 

[cadaveca]

I guess you missed the burnt 1155 sockets at launch then.

MSI tried to say that the board sample has been used during an OC competition, and that's why the socket burnt, but hte fact of the matter is that burnt sockets are NOT jsut a 1156 thing, unfortunately.

 

[John Doe]

I guess you missed the burnt 1155 sockets at launch then.

MSI tried to say that the board sample has been used during an OC competition, and that's why the socket burnt, but hte fact of the matter is that burnt sockets are NOT jsut a 1156 thing, unfortunately.

I was referring to everyday usage... if you're going to put a pot on it, you can burn anything. Not just the socket. That's not what the hardware is built to stand for.

Burnt sockets usually only happened to S1156 in the past decade. If any other socket burnt, that would be an extremely rare situation. In the past years of everyday forum reading, coming as far as to failed systems, a burnt socket hasn't been one of the cause of failure.

 

[cadaveca]

Peet's socket died under extreme ram clocking. In fact, I told him to be careful as I knew what would happen...but he wanted to see for himself, and he did. It's not his died during normal operation...2000 MHz CAS6 is hardly "normal".

Under normal usage, no socket should blow, period. Foxconn's problematic socket or not.


Fact of the matter shows that during the launch of SKT1155, many boards were warmped from teh socket retention mechanism. On Z68, this became less of a problem, but this specific issue is one that I tracked quite closely, as if it was a potential problem, I wanted to let my readers know.


That said, P55 sockets only really blew under OC'd conditions, as did the 1155 socket in question. Personally, I see it as no big deal.

 

[John Doe]

No, Foxconn's weak socket is the culprit of what has happened here. Any other socket wouldn't have blew up. It's not running 2000 at CL6 or power consumtion of RAM that's the issue. The Foxconn socket was problematic, so even if you didn't push it just as hard, it still would have failed sooner or later. Or for example, MSI boards have had (and still got) VRM issues on some. That's an another route of cause on them.

No other platform has had such socket issues.

 

[cadaveca]

You can hold any opinion you like. Fact of the matter is that I knew what Peet was doing would cause problems, warned him about it(maybe not forcefully enough), and when I spoke to him the next day, he confirmed my warnings.

At the same time, Peet had been running far higher ram speeds, and votlage, not too long before he had the problem. The fact that the problem with the Foxconn sockets was identified, and later remedied, means there are specific reason why the socket fails, and that OEMs were able to replicate the problem.

In fact, I even know what pins burnt, before ever looking at his pictures. Becuase it's not jsut the Foxxconn socket, but the cobination of the foxconn socket, with Hyper ram, IMHO, that causes the socket to fail.


Anyway, this has nothing to do with Corsair, the GTX sticks used, or the record that was broken. Please feel free to open a new thread to discuss the foxconn socket issue if you would like to continue the dicussion. It's an intersting subject I wouldn't mind talking about now that I am waiting on my CPU to arrive so I can get some work done on my X79 reviews.

 

[John Doe]

You can hold any opinion you like. Fact of the matter is that I knew what Peet was doing would cause problems, warned him about it(maybe not forcefully enough), and when I spoke to him the next day, he confirmed my warnings.

At the same time, Peet had been running far higher ram speeds, and votlage, not too long before he had the problem. The fact that the problem with the Foxconn sockets was identified, and later remedied, means there are specific reason why the socket fails, and that OEMs were able to replicate the problem.

In fact, I even know what pins burnt, before ever looking at his pictures. Becuase it's not jsut the Foxxconn socket, but the cobination of the foxconn socket, with Hyper ram, IMHO, that causes the socket to fail.

Because the socket was bunk and Hypers pulled more power than other memory, that's why it borked the socket. Blown sockets are a rare scenario under regular usage. It's only the Foxconn sockets that really had an issue with it. Those MSI boards failed due other reasons (lack of QC etc.).

 

[Planet]

But dave the picture shows two sticks but cpu-z is showing 1GB? 2 x 512 is what im thinking, do they even make 512 sticks anymore?

CPU-Z bug with Bulldozer

/me is not impressed
call me when they have 4GB of 3Ghz ram at decent timings

2800MHz with 8GB is here.

 

[cadaveca]

So 2x1 GB sticks? And the same GTX6 sticks as last year, right?




I'm way more eager to see 4GB DIMM clocking other than the hard-to-find kits, BTW.

 

[Planet]

So 2x1 GB sticks? And the same GTX6 sticks as last year, right?




I'm way more eager to see 4GB DIMM clocking other than the hard-to-find kits, BTW.

Nope single stick so single channel. CPU-Z just reports dual channel with Bulldozer. I think the new CPU-Z they released today fixes that as it hasnt been updated in 6 months.

Different GTX6 module than the previous blog post. I have an older blog on 16GB(4x4GB) 1600MHz Vengeance kit doing almost 2000MHz.

 

[cadaveca]

Yeah, I pulled that pic from your old blog(obviously).

So I gotta ask..why the pics with two sticks frozen then? Kinda false advertizing, in a way, no?


I've been playing with the Vengence LP Whites. It's not no 4x4 GB, but still pretty good, IMHO. Haven't found a max yet.:

View attachment 44449

 

[OneMoar]

I need no point in overclocking just for the fun of it UNLESS you CAN see some kind of usable gain
by either "stretching the life out" or providing a boost to current software

 

[Planet]

Yeah, I pulled that pic from your old blog(obviously).

So I gotta ask..why the pics with two sticks frozen then? Kinda false advertizing, in a way, no?


I've been playing with the Vengence LP Whites. It's not no 4x4 GB, but still pretty good, IMHO. Haven't found a max yet.:

http://www.techpowerup.com/forums/attachment.php?attachmentid=44449&d=1321830487

I think you are getting confused.

Here is a 16GB kit (4x4GB) @ 1920MHz these are Vengeance 1600MHz modules.
http://www.corsair.com/blog/corsair-vengeance-high-density-module-overclock-results

This is 8GB (2x4GB) of CMGTX7 running @ 2800MHz and they are cooled using liquid nitrogen.
http://www.corsair.com/blog/overclocking-the-new-dominator-gt-cmgtx7-ddr3-memory/

The World Record was set with a single 1GB module of CMGTX6 @ 3467MHz using liquid nitrogen.
http://www.corsair.com/blog/corsair-breaks-memory-frequency-world-record/

 

[cadaveca]

I think you are getting confused.

Not really. But yes.


This looks like two modules, and hence the confusion. Of course, it's just ice buildup.

Seperately,

This post of yours is something many people complain about, because near noone has been able to get anything even remotely close:

http://www.corsair.com/blog/corsair-vengeance-high-density-module-overclock-results

However, I'm getting pretty close with these LP Whites. Hence my pic/post.




Now. Don't get me wrong, I was asked specifically today why you had a pic with two modules in the GTX6 34xx MHz post. I'm just clearing the air now that you're here to answer directly. Personally, there's no confusion. I read all of your blog posts pretty much as soon as you post them. Checking the Corsair blog is something I do very regularily, and that's why I went right there to get that pic of the GTX6 I posted on the first page. I thought it was the same module, even.

 

[Planet]

Oh ok yea that is only a single module just a ton of ice built up around the Dominator heatspreader. As for the Vengeance all of them should overclock pretty well. I saw that same kit doing over 2133MHz on another forum. Awesome glad to hear you read the blogs

 

[cadaveca]

Yeah, gotta check store in the store anyway, so the blogs make the visit worthwhile when no modules I want are in stock.

Anyway, added with the CPU-Z saying dual channel, and you can understand what's going on regarding the GTX6.

 

[badtaylorx]

it can be easily broken by g skill ripjaws

what was that???

http://www.neoseeker.com/Articles/Hardware/Reviews/gskill_ripjawsz_ddr3_1866_16gb/4.html

not even close:shadedshu

 

[LAN_deRf_HA]

Unfortunately the platforms that allow ram tweaking aren't worth owning anymore. I could get cas 6 at 1600 out of these sector 5s on 1366. Supposedly they could run 2000 cas 6 with a top 1156 chip and board. On 1155 the best I can do is 1866 cas 8. Sure the platform gives you so much bandwidth you don't need it, but it feels like a pretty big waste.