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
Wreckage - Wednesday, September 23, 2009 - link
Hot, loud, huge power draw and it barely beats a 285.A disappointment for sure.
SiliconDoc - Thursday, September 24, 2009 - link
Thank you Wreckage, now, I was going to say draw up your shields, but it's too late, the attackers have already had at it.--
Thanks for saying what everyone was thinking. You are now "a hated fanboy", "a paid shill" for the "corporate greedy monster rip off machine", according to the real fanboy club, the ones who can't tell the truth, no matter what, and prefer their fantasiacal spins and lies.
Zstream - Wednesday, September 23, 2009 - link
They still allow you to post?yacoub - Wednesday, September 23, 2009 - link
He's right in the first sentence but went all fanboy in the second.Griswold - Wednesday, September 23, 2009 - link
Not really, he's a throughbred fanboy with everything he said. Even on the "loud" claim compared to what previous reference designs vom ATI were like...SiliconDoc - Wednesday, September 30, 2009 - link
So if YOU compare one loud design of ati's fan to another fan and as loud ati card( they're all quieter than 5870* but we'll make believe for you for now),
and they're both loud, anyone complaining about one of them being loud is "an nvidia fanboy" because he isn't aware of the other loud as heck ati cards, which of course, make another loud one "just great" and "not loud". LOL
It's just amazing, and if it was NV:
" This bleepity bleep fan and card are like a leaf blower again, just like the last brute force monster core power hog but this **tard is a hurricane with no eye."
But since it's the red cards that are loud, as YOU pointed out in the plural, not singular like the commenter, according to you HE's the FANBOY, because he doesn't like it. lol
ULTIMATE CONCLUSION: The commenter just told the truth, he was hoping for more, but was disappointed. YOU, the raging red, jumped his case, and pointed out the ati cards are loud "vom" prior.. and so he has no right to be a big green whining fanboy...
ROFLMAO
I bet he's a "racist against reds" every time he validly criticizes their cards, too.
---
the 5870 is THE LOUDEST ATI CARD ON THE CHART,AND THE LOUDEST SINGLE CORE CARD.
--
Next, the clucking rooster will whiplash around and flap the stubby wings at me, claiming at idle it only draws 27 watts and is therefore quiet.
As usual, the sane would them mention it will be nice not playing any 3d games with a 3d gaming card, and enjoying the whispery hush.
--
In any case:
Congratulations, you've just won the simpleton's red rooster raving rager thread contest medal and sarcastic unity award.(It's as real as any points you've made)
Anyhow thanks, you made me notice THE 5870 IS THE LOUDEST CARD ON THE CHARTS. I was too busy pointing out the dozen plus other major fibboes to notice.
It's the loudest ati card, ever.
GourdFreeMan - Wednesday, September 23, 2009 - link
I thought the technical portion of your review was well written. It is clear, concise and written to the level of understanding of your target audience. However, I am less than impressed with your choice of benchmarks. Why is everything run at 4xAA, 16xAF? Speaking for most PC gamers, I would have maxed the settings in Crysis Warhead before adding AA and AF. Also, why so many console ports? Neither I, nor anyone else I personally know have much interest in console ports (excluding RPGs from Bethesda). Where is Stalker: Clear Sky? As you note its sequel will be out soon. Given the short amount of time they had to work with DX11, I imagine it will run similarly to Stalker: Call of Pripyat. Also, where is ArmA II? Other than Crysis and Stalker it is the game most likely to be constrained by the GPU.I don't want to sound conspiratorial, but your choice of games and AA/AF settings closely mirror AMD's leaked marketing material. It is good that you put their claims to the test, as I trust Anandtech as an unbiased review site, but I don't think the games you covered properly cover the interests of PC gamers.
Ryan Smith - Wednesday, September 23, 2009 - link
For the settings we use, we generally try to use the highest settings possible. That's why everything except Crysis is at max quality and 4xAA 16xAF (bear in mind that AF is practically free these days). Crysis is the exception because of its terrible performance; Enthusiast level shaders aren't too expensive and improve the quality more than anything else, without driving performance right off a cliff. As far as playing the game goes, we would rather have AA than the rest of the Enthusiast features.As for our choice of games, I will note that due to IDF and chasing down these crazy AA bugs, we didn't get to run everything we wanted to. GRID and Wolfenstein (our OpenGL title) didn't make the cut. As for Stalker, we've had issues in the past getting repeatable results, so it's not a very reliable benchmark. It also takes quite a bit of time to run, and I would have had to drop (at least) 2 games to run it.
Overall our game selection is based upon several factors. We want to use good games, we want to use popular games so that the results are relevant for the most people, we want to use games that give reliable results, and ideally we want to use games that we can benchmark in a reasonable period of time (which means usually having playback/benchmark tools). We can't cover every last game, so we try to get what we can using the criteria above.
GourdFreeMan - Thursday, September 24, 2009 - link
Popularity and quality are strong arguments for World of Warcraft, Left 4 Dead, Crysis, Far Cry, the newly released Batman game... and *maybe* Resident Evil (though it is has far greater popularity among console gamers). However, HAWX? Battleforge? I would never have even heard of these games had I not looked them up on Wikipedia. In retrospect I can see you using Battleforge due to it being the only DirectX 11 title, but I still don't find your list of games compelling or comprehensive.To me *PC* gaming needs to offer something more than simple action to justify its cost of entry. In the past this included open worlds, multiplayer, greater graphical realism and attempts at balancing realistic simulation with entertaining game play. Console gaming has since offered the first two, but the latter are still lacking.
It's games like Crysis, Stalker and ArmA II along with the potential of modding that attract people to PC gaming in the first place...
dvijaydev46 - Wednesday, September 23, 2009 - link
Good review but it would be good if you could also add Steam and Cuda benchmarks. Now you have a common software Mediashow Espresso right?