AMD starts selling Radeon-branded DDR3 memory

[FU1G0R]

What the hell is grong with thos nasty latency numbers Radeon ? Really? No need to relate that loved name with 11 11 11 disgustin latency times WHAT THE HELL IS WRONG WHITH THEM SHEEESH

[>HyperlogiK<]

I wonder if GloFo can manufacture Flash chips (or perhaps already does)?

They did some years ago, but it was spun off as Spansion during one of AMD's many periods of financial trouble. I'm always amused at peoples' suggestions that AMD enter 'new' markets be it Flash, RISC processors, ultra mobile/embedded SOCs, or DSPs, because they used to make all of that stuff years ago but were forced to sell it off.

[Hans de Vries]

http://www.tcmagazine.com/tcm/news/h...ed-ddr3-memory



Attachment 118741


edit: origin here:
http://akiba-pc.watch.impress.co.jp/...6/etc_amd.html

DDR3 1333-9-9-9 and 1600-11-11-11 are exactly the standard speeds of all
the DRAM manufacturers under standard conditions and standard voltage.

http://www.samsung.com/global/busine...1&xFmly_id=690
http://www.elpida.com/en/products/ddr3module.html
http://www.micron.com/partscatalog.h...RAM/ddr3_SDRAM
http://www.hynix.com/inc/pdfDownload...M(Rev.0.3).pdf

AMD has a GDDR5 brand as well.

http://www.amd.com/us/products/techn...cs-memory.aspx

Seifert used to be COO and CFO of Qimonda in Dresden. They made some real good GDDR5:

http://www.pclaunches.com/other_stuf..._available.php

Hundreds of Qimonda engineers went to Globalfoundries but I expect that they are
now just reselling DRAMs from the big DRAM manufaturers



Regards, Hans

[jimbo75]

AMD entered a non-competition agreement with fujitsu and spansion in 2005 http://www.hexus.net/content/item.php?item=2583, preventing AMD from making standalone flash memory devices, so we wont see SSD's until that is over at least.

[verndewd]

I see no one has a thinking cap on today :P
wheres the corporate announcement stating GF was making these or tsmc?
Who is really making these? there should be an investors news brief about it but there isnt.

[verndewd]

DDR3 1333-9-9-9 and 1600-11-11-11 are exactly the standard speeds of all
the DRAM manufacturers under standard conditions and standard voltage.

http://www.samsung.com/global/busine...1&xFmly_id=690
http://www.elpida.com/en/products/ddr3module.html
http://www.micron.com/partscatalog.h...RAM/ddr3_SDRAM
http://www.hynix.com/inc/pdfDownload...M(Rev.0.3).pdf

AMD has a GDDR5 brand as well.

http://www.amd.com/us/products/techn...cs-memory.aspx

Seifert used to be COO and CFO of Qimonda in Dresden. They made some real good GDDR5:

http://www.pclaunches.com/other_stuf..._available.php

Hundreds of Qimonda engineers went to Globalfoundries but I expect that they are
now just reselling DRAMs from the big DRAM manufaturers



Regards, Hans

i would say thats spot on.
They can do a lot for thier product in assembling the pcb and chips but seeing no corp news on it , its unlikely that they are etching the chips or designing them. Not that they couldnt because node transitions depend on successful dram circuits being made first for cache memory.

[jstutman]

No one has got ahold of the chips and figured out who actually makes the chips? I almost believe its AMD as crappy as they are. lol

[verndewd]

No one has got ahold of the chips and figured out who actually makes the chips? I almost believe its AMD as crappy as they are. lol

yup so crappy they are the number 2 chip maker in the world. seriously dewd. i dont know what youre submerged in but it cant be good.

[G0ldBr1ck]

I think the point you guys are missing is when AMD claims this memory to do what it does it works 100% in all situations with there APU. Thus the tag "radeon memory"

Try a few sets of ram on there APU before you pretend to know what the deal is.

Certain IC's are failing at stock rated in 3dmark at even stock clocks.........much less games, oh and the failure is an artifact to name a couple I know of Micron D9KPT, Elpida Hypers.

System wise no failures are noticeable, APU wise however is a diff story.

I think everyone needs to see this post again ^. It seems some folks still dont understand the need for this ram.

[verndewd]

http://chipdesignmag.com/display.php?articleId=2488

heres a clue

[Philip_J_Fry]

so is this real or not? i mean its spreading around most tech news sites

[verndewd]

http://www.synopsys.com/Solutions/In...s/default.aspx

looks like synopsis bought virage

so heres my take. AMD gets synopsis to design a memory with the consideration of APU through put.
take some DDR3 adust a few things here and there and slap a brand on it becuase at some level AMD had to make design proposals and changes. They didnt have to reinvent the wheel they just went to a wheel maker with some design modifications.
either that or its quimonda Its very hard to find tsmc memory partners.

http://www.bing.com/search?q=tsmc%20...B&PC=BB07&QS=n

http://www.simmtester.com/page/news/...deal&num=12621
qualcomm partners with GF for memory at smaller nodes.

http://www.globalfoundries.com/design/memory_ip.aspx

virage in both camps.

[HKPolice]

Wow you guys are all over analyzing this.

AMD has extra DDR3 chips lying around from the production of low end video cards that still use DDR3. Especially with the current low price of memory, AMD killed 2 birds with one stone: make some quick cash by producing desktop dimms as well as expanding their product line at the same time.

[verndewd]

http://www.tomshardware.com/forum/248493-30-gddr
uhm no
http://forums.hardwarelogic.com/showthread.php?t=13860
and no

[HKPolice]

http://www.tomshardware.com/forum/248493-30-gddr
uhm no
http://forums.hardwarelogic.com/showthread.php?t=13860
and no

ATi employee told me himself, analyze all you want, believe what you want, it doesn't change the truth.

BTW did you even check Video card specs on AMD's website? The low end 6000 series radeons all use DDR3 OR GDDR5, NO mention of GDDR3 anywhere.

http://www.amd.com/us/products/deskt...verview.aspx#2

[STEvil]

how about both of you bother reading each others posts a bit closer so I dont have to toss you both out of news for calling each other tards?

[Poci]

Me want!!!!

[Mungri]

Ok, for people who think that ram timings dont make any difference for gaming, I did this test for you all:



Lets say I'm not putting my Corsair Vengeance sticks back into my PC. I did multiple runs, and my average frame rate is consistently up by 4-5 FPS with the faster 2 Gb modules.

[zalbard]

Why do you have a lot lower max framerate, then? The benchmark simply appears to be inconsistent.

[Mungri]

Why do you have a lot lower max framerate, then? The benchmark simply appears to be inconsistent.

The max framerate is always a 'blip (or blimp?) on the radar' in this benchmark. In triple runs with the slower ram, the max framerate would be anywhere between 110-150ish. Its normally just a freak frame where the performance spikes up or something, and doesnt make as much difference as average FPS does. Average framerate always remains consistent across multiple tests though, and never varies by more than 1 FPS.

I was actually running it for a Vram rest not for the ram modules, and I noticed it being so much higher now (I was already planning on selling my Vengeance modules because they didnt even manage to run at cas 8 and swapped them out yesterday).

[zalbard]

The max framerate is always a 'blip (or blimp?) on the radar' in this benchmark. In triple runs with the slower ram, the max framerate would be anywhere between 110-150ish. Its normally just a freak frame where the performance spikes up or something, and doesnt make as much difference as average FPS does. Average framerate always remains consistent across multiple tests though, and never varies by more than 1 FPS.

That makes sense.
It'd be also interesting to see Crysis 2 DX11 test of this. Its benchmarking software is quite good. Getting 15% performance increase just due of timings seems to be a situational thing (I'd expect 5% on average, perhaps up to 10%).

[eXa]

Corsair Vengeance can do thoose timings on AMD, on intel they should do much better than that...

[Mungri]

Corsair Vengeance can do thoose timings on AMD, on intel they should do much better than that...

There was a thread in the ram section where I posted about this ram -

There are two versions of it, Ver 5.11, and Ver 2.12.

Early sticks that were released and also sent to reviews were ver 5.11, these consistently managed Cas 8 1600 Mhz, and even Cas 10 2000 Mhz. After a while however corsair sneakily replaced these modules with Ver 2.12, which is what I and a few other people got. This version only goes up to 1500 Mhz Cas 8, hardly much over 1600 Mhz cas 9, and only up to around 1800 Mhz on Cas 10.

The older Ver 5.11 sticks have been rebranded and are being sold for an inflated price in the new Cas 10 2000 Mhz kit:

http://www.overclockers.co.uk/showpr...=8&subcat=1392

I bought the 1600 Mhz sticks based on reviews of the PC12800 kit using ver 5.11 sticks, and I got ver 2.12 sticks instead.

Finding good ram is too difficult to do. I already have 12 Gb of perfect 2 Gb sticks though, so I shouldnt bother with trying slower 4 Gb sticks again.

That makes sense.
It'd be also interesting to see Crysis 2 DX11 test of this. Its benchmarking software is quite good. Getting 15% performance increase just due of timings seems to be a situational thing (I'd expect 5% on average, perhaps up to 10%).

I think its because this benchmark uses up well over 1 Gb of Vram at maximum settings. Extra Vram info is going into the shared memory, and lower timings are going to reduce the latency on shared ram. But I've been comparing to a result on a pair of 2 Gb GTX 560s and my results are always higher due to the overclocks. Exceeding 1 Gb Vram in small amounts doesnt appear to degrade performance (apparantly this benchmark uses up 1.3 Gb vram on 2 Gb cards).

If you get 2 Gb GTX 560s and overclock them and the CPU to same amount, then Vram timings probably wont make any difference. But I'm not 100% certain about this, just guessing based on what I can think would make sense.

[Darxide]

M'eh. I've had Radeon branded RAM for years :-)

[BoredByLife]

Ok, for people who think that ram timings dont make any difference for gaming, I did this test for you all:



Lets say I'm not putting my Corsair Vengeance sticks back into my PC. I did multiple runs, and my average frame rate is consistently up by 4-5 FPS with the faster 2 Gb modules.

By changing sticks you ruined the comparison. Just up the timings of your first 6*2GB set for a fair comparison. And make the axis equal between the 2 settings.