How AMD's Istanbul might close the gap with Nehalem EP
by Johan De Gelas on February 25, 2009 12:00 AM EST- Posted in
- IT Computing general
The Istanbul cores are the same as those that can be found in the AMD's latest Shanghai CPU. But the "uncore" part of Istanbul is more interesting. By now, you have probably heard about AMD's "HT-assist" technology, a probe or snoop filter. Every time a new cacheline is brought into the L3-cache of for example CPU 1 on the current Shanghai Platform, a broadcast message is sent to all L3-caches of all CPUs, and CPU 1 has to wait until those CPUs answer.
In the case of Istanbul, the CPU will simply check it's snoop filter in it's own L3-cache, and if none of the other CPUs have that certain cacheline, it can go ahead. This lowers the latency of bringing in a new cacheline and raises the effective bandwidth.
To better understand this, we combined our own stream benchmarking with the one that AMD presented. All AMD systems are using DDR-2 800.
As each Stream thread works on its own data, there is no reason to send out coherency synchronization requests. These requests slow the process of getting new cachelines in the L3 and hence lower effective memory bandwidth. What is interesting is that this will not only benefit the applications that use the HT interconnects a lot for coherency traffic, but also applications like stream which do not need the HT interconnects. Also notice that HT 3.0 does not improve memory bandwidth, as Stream will try to keep its thread data local. Our testing used SUSE SLES 10 SP2 and AMD used Windows 2008. Both OSs are well optimized and NUMA aware.
This means that especially HPC applications, with many threads all working on their own data, will benefit from the higher effective bandwidth. Besides HT assist, AMD has now confirmed to us that the memory controller has been tuned quite a bit. This higher amount of bandwidth will allow the quad Istanbul to stay out of the reach of the dual Nehalem EP Xeons in many HPC applications.
HT assist might also improve the SAP and OLTP scores quite a bit, but for a different reason. SAP and OLTP applications perform a lot of cache coherency syncronization requests, so the snoop filter will substantially lower the average latency of such requests as in some cases:
- the CPU will only wait on one other CPU (instead of waiting for all responses to come back)
- the CPU won't have to wait at all, as the other CPUs don't have this line.
Secondly, this will also lower memory latency, which is a bonus for almost every multi-threaded application.
Lower memory latency, higher bandwidth, lower "cache coherency" latency and more interconnect bandwidth: the improved "uncore" of Istanbul will be vital to close the gap with Nehalem. Much will depend on how quickly Intel introduces its own hexacore 32 nm Xeons, but that probably won't happen before 2010. Istanbul is shaping up to be a really good alternative for Intel's quadcore Nehalem. We might see a good fight after all...
Don't forget to check it.anandtech.com (IT portal) often, as many of our blogposts (for example the VMworld 2009 coverage) are not published on the frontpage of Anandtech.com.
Facebook
Twitter
RSS
40 Comments
View All Comments
tshen83 - Friday, February 27, 2009 - link
I am relieved that there are still sensible people in this world.Now we know how Johan wants to argue with this:
"How AMD's Istanbul might close the gap with Nehalem EP?" By complete removal of reference to Nehalem-EP's performance in the charts and only show AMD's turds.
Stupid^2. Sometimes I wonder how such incompetency can exist.
hellopeach - Tuesday, June 2, 2009 - link
"turds"LOL, that has revealed that you are BIASED and you ARE the turds here.
By showing yourself is BIASED, you have completely invalidated everything in your posts yourself.
tshen83 - Wednesday, February 25, 2009 - link
I know a lot of the AMD employees read Anandtech, and possibly pay anandtech for outrageously favorable reviews. I am taking off. If I was AMD right now, I would do the following:1. Dual Socket market is dead. All of it will be canibalized by Westmere by 2010.
2. AMD would have no choice but to pitch the quad socket solutions against Intel's dual socket Westmere, which means that for the only metric that matters to datacenters: performance per watt per dollar, AMD would have no choice but to cut quad socket CPUs to less than $500 dollars per CPU. So consumers would have to make a choice
between a dual socket (2x$1000) solution from Intel or a quad socket (4*$500) solution from AMD. Even this doesn't solve the fact that quad Socket will draw at least twice as much power vs dual socket Intel. Only markets left for AMD are virtualization and database. Sad but true.
I am glad that Hector Ruiz the moron is out. But watch him ruin the manufacturing spinoff.
I am off.
Zak - Wednesday, February 25, 2009 - link
Jesus Christ... we really need to make up our minds and decide once and for good whether AnandTech is pro-Intel or pro-AMD... This is getting confusing, ya know?Z.
TA152H - Wednesday, February 25, 2009 - link
You know, I get tired of reading about how this site is biased one way or another, and particularly when people say it is so for Intel, and now you for AMD. The fact is, Intel is much better right now than AMD, and if the site didn't point that out, they would be remiss.In your case, I'm fully in agreement, but I don't think it's everyone on the site. I've noticed that Johan is very AMD biased, and even when he ostensibly is talking about how good an Intel product is, he always puts a mitigating remark to offset it. Now he's comparing a dual with a quad, and doesn't really draw too much attention to the facts you mention. Clearly he's smart enough to know them, and know how important they are, but, he kind of let's it slip and doesn't mention that a quad core might actually take more power than a dual core, and forgets how often he mentions how important power consumption is. He's right to mention it, of course, but in this article, strangely, it's not brought up.
Europeans broadly seem to favor AMD, because, somehow, they consider it more European than Intel. I guess the Dresden plants may have something to do with that, or maybe they like the underdog, or maybe AMD does a lot more advertising there, or something like that. I don't know, but Europeans tend to favor AMD, and he just hasn't gotten past his bias, and thinks he's hiding it well.
Face it, the Nehalem is best x86 processor out there. PERIOD. If you have to compare a quad to a dual, then you're proving that point while ostensibly trying to prove the opposite. The performance/power of the Nehalem is untouchable. AMD is fine for the mainstream surfer, and I'd definitely consider a Phenom II for home, but show me someone that buys it over a Nehalem, and I'll show you an idiot in most cases. Nehalem is just a better processor, as the P8 was, but now it has proper technology around it to eliminate the problems that its predecessor had.
FireSnake - Thursday, February 26, 2009 - link
Taking off. by TA152H: "Europeans broadly seem to favor AMD"This is one of the stupidest things i have heard lately.
Go back to the cave you came from and live in!
I am from Europe and what you are saying is a bunch of BS!
mikeepu - Wednesday, February 25, 2009 - link
You mean Dual 4-Core for the Nehalem and Quad 6-Core Istanbul and Quad 4-Core for Shanghai?Mikey - Wednesday, February 25, 2009 - link
I agree that the Nehalem is the better processor out there. There's no comparison between a quad core vs a dual-core setup. Its like comparing 4 apples to 2 oranges. Regardless of its gap-closing performance, I would have to agree that its much less practical to go forth with AMD's quad core computer setup. Nehalem is hands down the better x86 processor. This doesn't make AMD's processors bad...it's still one of the best...just NOT the best.Mike
JarredWalton - Wednesday, February 25, 2009 - link
Okay, let me get this straight: an article is titled "How AMD's Istanbul might close the gap with Nehalem EP" and Johan is now totally AMD biased. Amazing! We could call it, "AMD isn't totally dead in the water" and also be AMD biased, I suppose?http://it.anandtech.com/weblog/showpost.aspx?i=555">This wasn't very pro-AMD as one example. Staying in the slipstream of Intel means you can be close, but you'll still be behind. There are http://it.anandtech.com/weblog/showpost.aspx?i=480">more articles from Johan discussing how Nehalem is shaping up to be the IT wunderkind CPU. All this particular article shows is that there are some improvements to Istanbul that ought to help it remain in the ballpark of Intel for some situations.
TA152H - Wednesday, February 25, 2009 - link
Jarrod, I know you're smarter than that post.You took one line, and even then we can talk about that, and from that extrapolate the entire article? OK, so why write the article?
The latter response would not have been AMD biased, but even the title gives AMD more credit than it deserves, but, again if you read my post, you probably saw the writing after the title is what I objected to.
Mind you, I am often defending AMD products to people that only consider Intel. I do consider them fine for mainstream users, at least the Phenom II (I see absolutely no use for any previous processors at this time, but, that's probably a bit extreme), and often times tell people that Intel only is too simplistic, since in some cases the AMD solutions, particularly the 790GX and and Phenom II make more sense than an Intel with G45. So, I'm not against AMD at all.
But, Istanbul isn't even close to the Nehalem. When one says closes the gap, there's a purely literal interpretation, which could mean even a 1% improvement, and the interpretation one perceives when reading it. If you are writing, you have to know the difference, and the choice of words does reflect a gross exaggeration of the capabilities of the Istanbul when you consider how even the title is interpreted.
Let's put it a different way. If someone with 15 career home runs hits a homer, do you think a correct headline for it should be that the player has closed the gap to Hank Aaron (let's forget Bonds for the sake of discussion)? It's technically correct, but the implication is inaccurate, and that's how people who are biased properly approach their trade. They can't be so overtly so that it's recognized easily by the hoi polloi, so they try to be delicate about it with their choice of words, and mood they create while discussing the products. That's what Johan does.
Normally, I like it, because I think for servers Intel was getting too much credit. Well, until the Nehalem came out, and then, well, the game was over. But, this article he really overstepped his limit and was overtly pro-AMD.
Now, let's say you were writing this article. You wouldn't have brought up the power considerations? You wouldn't have emphasized the difference between two and four processors? Really, you wouldn't? And for Johan, who to his credit always emphasizes how important power considerations are (and he's so correct on that), to somehow kind of pass over this, shows a clear bias.
Don't get me wrong, I do respect his intelligence, and his diligence in tests, but I do perceive a clear bias. For the rest of the people who write, I do not detect a bias one way or another. For what it's worth, I think Tom's Hardware has a clear AMD bias, since they gave a clearly inferior processor, the Phenom II, a Gold Award. Why? It's not even better than the Penryn, and it doesn't even approximate the Nehalem. A Gold Award? It's still a decent processor for the price, and isn't always a bad choice, but since when does a processor get this award for being inferior to the competitions last generation, and way behind their current? It's almost insulting to AMD that they expect so little from them, that a decent processor generates so many accolades.