AMD's Radeon HD 5870: Bringing About the Next Generation Of GPUs
by Ryan Smith on September 23, 2009 9:00 AM EST- Posted in
- GPUs
More GDDR5 Technologies: Memory Error Detection & Temperature Compensation
As we previously mentioned, for Cypress AMD’s memory controllers have implemented a greater part of the GDDR5 specification. Beyond gaining the ability to use GDDR5’s power saving abilities, AMD has also been working on implementing features to allow their cards to reach higher memory clock speeds. Chief among these is support for GDDR5’s error detection capabilities.
One of the biggest problems in using a high-speed memory device like GDDR5 is that it requires a bus that’s both fast and fairly wide - properties that generally run counter to each other in designing a device bus. A single GDDR5 memory chip on the 5870 needs to connect to a bus that’s 32 bits wide and runs at base speed of 1.2GHz, which requires a bus that can meeting exceedingly precise tolerances. Adding to the challenge is that for a card like the 5870 with a 256-bit total memory bus, eight of these buses will be required, leading to more noise from adjoining buses and less room to work in.
Because of the difficulty in building such a bus, the memory bus has become the weak point for video cards using GDDR5. The GPU’s memory controller can do more and the memory chips themselves can do more, but the bus can’t keep up.
To combat this, GDDR5 memory controllers can perform basic error detection on both reads and writes by implementing a CRC-8 hash function. With this feature enabled, for each 64-bit data burst an 8-bit cyclic redundancy check hash (CRC-8) is transmitted via a set of four dedicated EDC pins. This CRC is then used to check the contents of the data burst, to determine whether any errors were introduced into the data burst during transmission.
The specific CRC function used in GDDR5 can detect 1-bit and 2-bit errors with 100% accuracy, with that accuracy falling with additional erroneous bits. This is due to the fact that the CRC function used can generate collisions, which means that the CRC of an erroneous data burst could match the proper CRC in an unlikely situation. But as the odds decrease for additional errors, the vast majority of errors should be limited to 1-bit and 2-bit errors.
Should an error be found, the GDDR5 controller will request a retransmission of the faulty data burst, and it will keep doing this until the data burst finally goes through correctly. A retransmission request is also used to re-train the GDDR5 link (once again taking advantage of fast link re-training) to correct any potential link problems brought about by changing environmental conditions. Note that this does not involve changing the clock speed of the GDDR5 (i.e. it does not step down in speed); rather it’s merely reinitializing the link. If the errors are due the bus being outright unable to perfectly handle the requested clock speed, errors will continue to happen and be caught. Keep this in mind as it will be important when we get to overclocking.
Finally, we should also note that this error detection scheme is only for detecting bus errors. Errors in the GDDR5 memory modules or errors in the memory controller will not be detected, so it’s still possible to end up with bad data should either of those two devices malfunction. By the same token this is solely a detection scheme, so there are no error correction abilities. The only way to correct a transmission error is to keep trying until the bus gets it right.
Now in spite of the difficulties in building and operating such a high speed bus, error detection is not necessary for its operation. As AMD was quick to point out to us, cards still need to ship defect-free and not produce any errors. Or in other words, the error detection mechanism is a failsafe mechanism rather than a tool specifically to attain higher memory speeds. Memory supplier Qimonda’s own whitepaper on GDDR5 pitches error correction as a necessary precaution due to the increasing amount of code stored in graphics memory, where a failure can lead to a crash rather than just a bad pixel.
In any case, for normal use the ramifications of using GDDR5’s error detection capabilities should be non-existent. In practice, this is going to lead to more stable cards since memory bus errors have been eliminated, but we don’t know to what degree. The full use of the system to retransmit a data burst would itself be a catch-22 after all – it means an error has occurred when it shouldn’t have.
Like the changes to VRM monitoring, the significant ramifications of this will be felt with overclocking. Overclocking attempts that previously would push the bus too hard and lead to errors now will no longer do so, making higher overclocks possible. However this is a bit of an illusion as retransmissions reduce performance. The scenario laid out to us by AMD is that overclockers who have reached the limits of their card’s memory bus will now see the impact of this as a drop in performance due to retransmissions, rather than crashing or graphical corruption. This means assessing an overclock will require monitoring the performance of a card, along with continuing to look for traditional signs as those will still indicate problems in memory chips and the memory controller itself.
Ideally there would be a more absolute and expedient way to check for errors than looking at overall performance, but at this time AMD doesn’t have a way to deliver error notices. Maybe in the future they will?
Wrapping things up, we have previously discussed fast link re-training as a tool to allow AMD to clock down GDDR5 during idle periods, and as part of a failsafe method to be used with error detection. However it also serves as a tool to enable higher memory speeds through its use in temperature compensation.
Once again due to the high speeds of GDDR5, it’s more sensitive to memory chip temperatures than previous memory technologies were. Under normal circumstances this sensitivity would limit memory speeds, as temperature swings would change the performance of the memory chips enough to make it difficult to maintain a stable link with the memory controller. By monitoring the temperature of the chips and re-training the link when there are significant shifts in temperature, higher memory speeds are made possible by preventing link failures.
And while temperature compensation may not sound complex, that doesn’t mean it’s not important. As we have mentioned a few times now, the biggest bottleneck in memory performance is the bus. The memory chips can go faster; it’s the bus that can’t. So anything that can help maintain a link along these fragile buses becomes an important tool in achieving higher memory speeds.
Facebook
Twitter
RSS
327 Comments
View All Comments
SiliconDoc - Thursday, September 24, 2009 - link
Oh really ? Now wait a minute, spin master. When the site here whined about "paper launch" it was Derek who brought up a two or three year old nvidia card, and cried and whined about it. Then speculated the GTX275 was paper, and then "a phantom card".Well, that didn't happen.... no apologies about it ever either.
---
The PAPER launches of late are ATI ATI ATI ! ! !
We have the 4770, and now this one !
----
Gee, when ATI BLOWS IT, we suddenly talk in vague terms about "the companies" having "papery launches" as " the general rule of thumb of how it's done.." - and that makes us "not a fan boy!??!"
R0FLMAO !!!!
Yes, of course, since the red ati is bleeding paper launches and the last one from nvidia one can actually cite is YEARS AND YEARS ago, yes, of course, you're correct, it's "unnamed companies in the multiple" that "do it"....
---
I swear to god, I cannot even believe the massive brainwashing that is a gigantic pall all over the place.
---
If I'm WRONG, please be kind, and tell me what nvidia paper launches I missed.... PLEASE LET ME KNOW.
Genx87 - Wednesday, September 23, 2009 - link
Is\was a good idea to shoot for as this is most certainly what Nvidia is going to attempt to achieve. But I am a bit disappointed this care rarely achieved it.I do like angle independent AF though. Should be interesting to see what Nvidia brings to the table. But kind of like the CPU situation(i5) I am kind of meh. But will say this has more potential compared to its predecessor than the i5 series does compared to Core 2 Duo.
SiliconDoc - Wednesday, September 23, 2009 - link
I thought that was just great, those pretty pictures, and then I get to reading. I see the 4890 and SQUARES. I see the GTX285, with CIRCLES and an outer rounded octagon.Then the 5870- and it's "perfectly round" angle independent algorithm, but I still see some distortions.
--
So I get to reading and am told "the 4890 and 285 are virtually the same". I guess the wheel was first made square, and rolled as well as when it became round. No chance the reviewer could tell the truth and remark that NVidia has the best, until today.. NOPE can't do that!
---
Then, of course, the celebration for the "perfection" of the 5870 and ATI's superb success in the "round" category...
EXCEPT:
We get to the actual implementations and NO PERCIEVED DIFFERENCE IS VISUALLY THERE. It cannot be seen. The article even states they searched in vain for some game to show the difference. LOL
All that extra effort to for pete sakes show that ATI superiority...all WASTED EFFORT, but for red roosters I'm certain it was a very exciting quest, titillating, gee a change to take down big green...
---
So bottom line is IT'S A BIG FAT ZERO, even the older, worse ati implementation is apparently "non distinguishable".
It is remarked that NVidia doesn't "officially" support this method in game, and of course, after much red rooster effort, one finds out why.
THERE IS NO DIFFERENCE in visual quality. Another phantom red "win".
Another reason NVidia makes money (why waste it on worthless crap in developement that makes no difference), while ATI does not.
Yeah, that was so cool.
So happy the "mental ideation of perfection in the card for ati fans" was furthered. ROFL
Dante80 - Wednesday, September 23, 2009 - link
A quick question. Why is there no 5850 review available atm?1> Was there a separate NDA for the 2 cards?
2> Were there no sample cards given by AMD to reviewers?
3> Did AMD ask reviewers to postpone said reviews due to market supply problems/glitches?
4> Was this a strategy decision by AMD, for marketing or other reasons?
Ryan Smith - Wednesday, September 23, 2009 - link
AMD only provided us with 2 5870s, the 5850 was not sampled. 5800 series cards are in short supply, even for reviewers.Dante80 - Thursday, September 24, 2009 - link
Thank you for the prompt answer, that was what I was guessing too. Cheers...^^Spoelie - Wednesday, September 23, 2009 - link
To get enough 5870 cards in the channel for a hard launch, they used every possible die.There are probably not enough harvested dies to create the 5850 line just yet. And they're not gonna use fully functional ones that can go in a 5870 when supply for them is tight already.
Once the 5850 is launched, demand for them is up and yields matured, they'll have to use fully functional dies to keep supply up, but now they're building up inventory for a hard launch during the coming weeks.
SiliconDoc - Wednesday, September 23, 2009 - link
Uhh, just a minute there feller. The SOFT or PAPER LAUNCH has already hit, the big LAUNCH DATE is today....Newegg is a big ZERO available... (one Powercolor was there 30 mins ago, the other 3 listed are NOT avaailable, I watched them appear last night).
---
So, when Ati has a "hard launch" they get "many weeks after the launch date" to "ramp up production" and "fill the need".
ROFLMAO
I was here when this site and the red roosters whined about Nvidia and appear launches, and I believe it was the GTX275 that was predicted to be PAPER (not very long ago in fact) here, and the article EVEN SPECULATED IT WAS A PHANTOM CARD.
All the red roosters piled on, but.... the card was available on launch, it wasn't a PHANTOM, and all that bs was quickly forgotten and shoved into the memory hole like it never happened...
---
Oh, but when it's ATI and not Nvidia, the 4770 can remain almost pure paper near forever, and this one, golly it can be 95% paper and it's just " getting ready for a hard launch" WEEKS BEYONDS the launch date!
ROFLMAO
--
No bias here?!? "Where's da' bias?!?!" said the red rooster (to the green goblin)...
Give me a break.
chrnochime - Friday, September 25, 2009 - link
Just because you can't find it in the states doesn't mean it's a fake launch. And fake launch? What are you a 12 year old or something? You're like the nvidia version of snakeoil. Just go play with your nvidia part m'kay ?SiliconDoc - Sunday, September 27, 2009 - link
Well, since you insulted, and mischaracterized, I came across the reminder about the 4870 paper launch.Yes, that's correct, this is how ATI rolls, a big fat lying launch date, a piddle of a few cards, then wait a couple weeks or a month.
--
" The cards are fast, but as many pointed out HD 5870 is not faster than Geforce GTX 295, which is something that many have expected. Radeon 5850 will also start selling in October time, but remember, last summer when ATI launched 4870, the card was almost impossible to buy and weeks if not months after, the availability finally improved. "
http://www.fudzilla.com/content/view/15643/1/">http://www.fudzilla.com/content/view/15643/1/
--
Like I've kept saying, the bias is so bad... I keep discovering more big fat ati blunderous moves, that are instead ascribed to imaginarily to Nvida.
Thanks for the incorrect whining, anger, and standard PC e-mindless chatroom repeated, non original, heard ten thousand times, brainless insult, it actually helped me.
I learned ATI blew their 4870 launch with paper lies as well.
You're a great help friend.