Monday, January 24th 2011

Bulldozer Shines in 3D Gaming and Rendering: AMD

Close to two weeks ago, reports surfaced about AMD claiming that its upcoming "Zambezi" 8-core desktop processor based on the company's new Bulldozer architecture is expected to perform 50% faster than Intel's Core i7 and its own Phenom II X6 processors. The slide forming the basis for the older report surfaced, and it's a little more than a cumulative performance estimate.

Slide #14 from AMD's Desktop Client Solutions presentation to its industry partners reveals that the company went ahead and provided a breakdown on which kinds of applications exactly does its new 8-core chip perform better compared to present-generation processors. The breakdown provides an interesting insight on the architecture itself. To begin with, AMD's 8-core Bulldozer "Zambezi" processor is 1.5X (50%) faster overall compared to Intel Core i7 "Bloomfield" 950, and AMD Phenom II X6 1100T. Breaking down that graph, the processor performs similar to the other chips in media applications, but features huge gains in gaming and 3D rendering, which is where most of its gains are coming from.
To put this into perspective, games and 3D graphics applications, which still favour processors with higher clock speeds with lesser number of cores/threads to processors with lesser clock speeds and higher number of cores/threads, performing well on Bulldozer indicates that AMD is concentrating on higher performance per core, in other words, higher instructions per clock (IPC). The modular design of Bulldozer, perhaps, is contributing to high inter-core bandwidth, which helps 3D games that can do with lesser number of cores.

AMD described the Zambezi-powered "Scorpius" enthusiast desktop platform to have "the best graphics features and performance". A comparative table also reminds us that apart from the radical design, Bulldozer might benefit from a vastly upgraded SIMD instruction set compared to the previous generation. Bulldozer packs SSE 4.1, SSE 4.2, and AVX (Advanced Vector Extensions). With socket AM3+ motherboards already seeing the light of the day in pre-release photo shoots, AMD's new processor doesn't seem too far.
Source: DonanimHaber
Add your own comment

122 Comments on Bulldozer Shines in 3D Gaming and Rendering: AMD

#101
Unregistered
guys, check out this JF-AMD post on anandtech
Why couldn't it be 95W? There are 95W client parts today.

There will be a 35W TDP 8-core server part and there will be a <95W TDP 16-core server part as well.
its look like it will be a power efficient bulldozer

source
#102
Thatguy
JF-AMDNot confirming the info because I can only confirm the things I have control over. If that was a server slide I could tell you with 99% accuracy because I make 99% of them.




That is very true, but look at the bigger picture. Let's say you are going to sell a million processors in Q1. But then there is a price war. You lower the price $10. Doesn't sound like a lot.

$10M

And that is only Q1. Most products live ~6 quarters. So that little price move cost you $60M in pure profit.

And worse yet, if you look at the typical processor market, new products come in at the price of the old products, so your $10 price cut carries on to the next generation.

Companies tend to not cut prices, but instead push new technology in at those established price points. Taking a price cut also disrupts the stack, so while you think you just need to change one price, the new price is too close to the one below. So you have to drop that.

It becomes a snowball that eats up profits. And the guy with the bigger share has more to lose in a price war, not the other way around.
I like your points and they are very valid with regards to profit per unit and total scaled profits. Its usually better to raise prices and move a few less units then lower prices and sell more units. People forget the economic scalling falls apart at certain volumes. It just depends on where those volumes fall on the distrobution side that makes it worth while for the producer.

Would be nice to see some client side info however

hint hint.
Posted on Reply
#103
newtekie1
Semi-Retired Folder
TheLaughingManWait. You are comparing a i3 540 to a AM2+ 920? That processor wasn't even the flagship of its generation and no one is thinking about it now.
I used the 920 because they didn't have a comparison with the 925, which is the same processor with the same performance, but DDR3 support enabled which is pretty accepted as not really helping performance any.
TheLaughingManTry comparing your i3 540 to say the Phenom II X2 565 or Athlon II X4 645. Both of those are $10 cheaper and you will see your comparison gets real grey real fast.

www.anandtech.com/bench/Product/204?vs=143

www.anandtech.com/bench/Product/188?vs=143
Actually, as per the post you quoted, I did compare the 565 to the i3-540, and the 540 wins.(I just linked to the wrong comparision.) Comparing it to the 645 still sees a lot of wins in the 540 column, but I think it is closer, but the 540 still crushes the 645 in games, and even some multi-threaded apps...

And again, I'm not saying the 540 is an amazing super great processor that everyone should buy over AMDs offerings. I'm saying that the idea that AMD is better price for the buck in the areas the compete in isn't a given fact. Intel stays pretty competitive going down into the lower end.
TheLaughingManPrice/performance is AMD stomping ground because they price their processors according to their audience and Intel inflation.
Yeah, if you believe this, I pitty you. This is coming from someone that paid $900 for an AMD processor, AMD prices their products where they need to be to sell. Don't kid yourself and think that AMD would still be selling their highend processors at $300 if they were on top and Intel wasn't. They sell their processors at low prices because they have to, because that is where they perform. You aren't getting incredible bang for the buck by going with AMD, you are just getting a cheaper processor that doesn't perform as well, but performs good enough for you to not notice.
TheLaughingManWho gave regular folks quad cores for less than $100?
AMD-Because the $99 quad-core performed worse than Intels $99 dual-cores. But minor details like that don't mater, I'VE GOT FOR CORES AND YOU DON'T! HAHA
TheLaughingManWho gave gamers the first "sweet spot" of our current gen or offerings?
Sweet spot if you consider only Intels current generation, but since Intel kept that last generation going to compete with AMD's "sweet spot", Intel still had that title...they just held it with the last generation products that were still competiting with AMDs current gen.:laugh:
TheLaughingManFor me it has always been 3 groups:
* People who want to computer on a budge, AMD is your best friend.
* People who want to swing their e-penis in public and brag about how fast their processor is in applications they don't own or use, Intel to the rescue.
* People who try to mix budget with performance, I feel sorry for your because this middle ground's competition is ugly and confusing. This is my buying area and I am always torn for weeks before I make a final decision. And to be honest, I don't really think me picking one or the other ever really matters.
Exactly, that is my point. Intel competes very well in the middle-ground. The only area that AMD really shines is low-end, and even there Intel's Celerons compete pretty well, they just don't have something to compete with the single cores from AMD.(but really, would you recommend a single core to anyone today?) But, honestly, once you get down this low, most people buying here don't care about a few percentage points difference, it really won't help them checking their email any faster, and won't let them watch movies any better. Where AMD shines in the low end market is their integrated graphics, they really do walk all over Intel, who are still relying on the G31/G41 chipset in this market.
TheLaughingManI think AMD should go after that middle ground more aggressively which is what they seem to be doing. I don't think the initial flagship will truly compete with Sandy's top end, but I expect it to go blow for blow with Sandy's mid-range processors in the same price range. I am just hoping this time AMD will take the mid-ranged crown so they can say, "We beat Intel overall in every price segment, unless you are spending $800+." And they can say, "And we offer better overall server processors in every price segment, unless your budget is unlimited. Then I think we can help you with our GPU based servers."
I expect the same, but I fear that people are getting their hopes up and over hyping what Bulldozer will really do.
Posted on Reply
#104
TheLaughingMan
newtekie11. Actually, as per the post you quoted, I did compare the 565 to the i3-540, and the 540 wins.(I just linked to the wrong comparision.) Comparing it to the 645 still sees a lot of wins in the 540 column, but I think it is closer, but the 540 still crushes the 645 in games, and even some multi-threaded apps...

2. Yeah, if you believe this, I pitty you. This is coming from someone that paid $900 for an AMD processor, AMD prices their products where they need to be to sell. Don't kid yourself and think that AMD would still be selling their high end processors at $300 if they were on top and Intel wasn't. They sell their processors at low prices because they have to, because that is where they perform. You aren't getting incredible bang for the buck by going with AMD, you are just getting a cheaper processor that doesn't perform as well, but performs good enough for you to not notice.
1. i3 540 does not win against the AMD PII 925 or AII 645 in much of anything. Here you go: www.techpowerup.com/reviews/Intel/Core_i3_540_530/1.html

But I digress as synthetic benchmarks are mostly crap IMO.

2. Them adjusting price based on compared performance is exactly what I was talking about with the Intel Inflation comment. And yeah, I kinda do believe that if AMD had the top of the heap processor it would as over priced as Intel. I don't think it would be $300 range, but I seriously doubt it would be in the $1000+ range either.
Posted on Reply
#105
Wile E
Power User
JF-AMDNot confirming the info because I can only confirm the things I have control over. If that was a server slide I could tell you with 99% accuracy because I make 99% of them.




That is very true, but look at the bigger picture. Let's say you are going to sell a million processors in Q1. But then there is a price war. You lower the price $10. Doesn't sound like a lot.

$10M

And that is only Q1. Most products live ~6 quarters. So that little price move cost you $60M in pure profit.

And worse yet, if you look at the typical processor market, new products come in at the price of the old products, so your $10 price cut carries on to the next generation.

Companies tend to not cut prices, but instead push new technology in at those established price points. Taking a price cut also disrupts the stack, so while you think you just need to change one price, the new price is too close to the one below. So you have to drop that.

It becomes a snowball that eats up profits. And the guy with the bigger share has more to lose in a price war, not the other way around.
They also move more units to make up the difference. Intel already matches AMD's price cuts every time they happen. Whether with a new part or not is irrelevant to a purchaser. AMD cuts prices, and Intel has a cpu right there at the same perf level in the same price range. I don't see that changing any time soon.
TheLaughingMan1. i3 540 does not win against the AMD PII 925 or AII 645 in much of anything. Here you go: www.techpowerup.com/reviews/Intel/Core_i3_540_530/1.html

But I digress as synthetic benchmarks are mostly crap IMO.

2. Them adjusting price based on compared performance is exactly what I was talking about with the Intel Inflation comment. And yeah, I kinda do believe that if AMD had the top of the heap processor it would as over priced as Intel. I don't think it would be $300 range, but I seriously doubt it would be in the $1000+ range either.
You are deluded then. They charged $1000 for the FX cpus when they were ahead of Intel last time. They will do it again if they ever get up there again. The only way top end might drop, is if both AMD and Intel are on equal performance ground, and a price war commences.
Posted on Reply
#106
HillBeast
So an upcoming product from a new generation is beating a CPU which Intel has replaced 3 times over (i7 Bloomfield -> i7 Lynnfield -> i7 Gulftown -> Sandybridge) by a mere 50%. I was expecting WAY more than that. Why didn't they compare it to the 980X or a Sandy Bridge chip? Because they were too expensive for them to buy at the time? They lost the chips on the way home? They forgot which chip was the best from Intel?

Come on AMD, that's a crock and you know it.
Posted on Reply
#107
erocker
*
HillBeastSo an upcoming product from a new generation is beating a CPU which Intel has replaced 3 times over (i7 Bloomfield -> i7 Lynnfield -> i7 Gulftown -> Sandybridge) by a mere 50%. I was expecting WAY more than that. Why didn't they compare it to the 980X or a Sandy Bridge chip? Because they were too expensive for them to buy at the time? They lost the chips on the way home? They forgot which chip was the best from Intel?

Come on AMD, that's a crock and you know it.
Why so mad? The guy said it himself that AMD didn't release these performance comparisons. At least let AMD and actual reviewers come out with some concrete performance numbers first. :rolleyes:
Posted on Reply
#108
newtekie1
Semi-Retired Folder
TheLaughingMan1. i3 540 does not win against the AMD PII 925 or AII 645 in much of anything. Here you go: www.techpowerup.com/reviews/Intel/Core_i3_540_530/1.html

But I digress as synthetic benchmarks are mostly crap IMO.
It all depends on where you look, they go back and forth, which is my point. They are close in performance and priced accordingly.
TheLaughingMan2. Them adjusting price based on compared performance is exactly what I was talking about with the Intel Inflation comment. And yeah, I kinda do believe that if AMD had the top of the heap processor it would as over priced as Intel. I don't think it would be $300 range, but I seriously doubt it would be in the $1000+ range either.
Again, coming from someone that paid $900 for a AMD processor, I can assure you they would if they could.
Posted on Reply
#109
jsfitz54
My last build was an Asus A8N32SLI-deluxe with an AMD dual core FX-60 that cost $400.00.
My new i7 950 quad core from Intel cost $280.00.
I only built AMD systems for 10 years and I made the decision to go with Intel for my current build. I made that choice because Intel had a great product. This is also the first time I have had an SLI setup. I can say I am not loyal to either camp at this time but will make future decisions based on need and cost.

What is of interest to me now is Tri channel memory systems, will it last or go away like Rambus memory? Will dual channel win out? AMD has stayed with Dual channel but right now Tri channel feels like it was an experiment for a few years like Rambus.

While the CPU's are getting faster, memory does not seem to be keeping pace...almost disproportionately.

While we all would like the next best thing most of us don't need it, we just like to tinker a bit. To that end, all we can do is wait and see.

I am interested to see 5GHz on Air. That would make us all...Buy..Buy..Buy.
Posted on Reply
#110
Thatguy
jsfitz54My last build was an Asus A8N32SLI-deluxe with an AMD dual core FX-60 that cost $400.00.
My new i7 950 quad core from Intel cost $280.00.
I only built AMD systems for 10 years and I made the decision to go with Intel for my current build. I made that choice because Intel had a great product. This is also the first time I have had an SLI setup. I can say I am not loyal to either camp at this time but will make future decisions based on need and cost.

What is of interest to me now is Tri channel memory systems, will it last or go away like Rambus memory? Will dual channel win out? AMD has stayed with Dual channel but right now Tri channel feels like it was an experiment for a few years like Rambus.

While the CPU's are getting faster, memory does not seem to be keeping pace...almost disproportionately.

While we all would like the next best thing most of us don't need it, we just like to tinker a bit. To that end, all we can do is wait and see.

I am interested to see 5GHz on Air. That would make us all...Buy..Buy..Buy.
the next jump in memory will be on die ram becuase the bus is quickly reaching the specs of what it can reliably transmit.
Posted on Reply
#111
HillBeast
erockerWhy so mad? The guy said it himself that AMD didn't release these performance comparisons. At least let AMD and actual reviewers come out with some concrete performance numbers first. :rolleyes:
Was I mad? No. Was I disappointed in AMD? Yes. They have spent 6 or 7 years raving on about Fusion and how it's going to be the best CPU ever and how it'll revolutionise computing, and all I have seen is steam and every benchmark that has been leaked hasn't been anything I would deem as being 'ground breaking'.

For one, AMD came up with the idea of an APU for desktop computing, and in the time it took for AMD to rave on about it and produce nothing, Intel built a so-called APU, TWICE. Yes the GPU is rubbish but can you honestly tell me the video encoder/decoder that uses the GPU in Sandy Bridge isn't amazing? Unless ALL the figures that have been mentioned so far are a major underestimation, I can see AMD Fusion become AMD Flop.

Besides, this slide is released by AMD, leaked yes, but it's still from AMD. That still means the information in this should be taken as an exaggeration and that this will in fact be HIGHER than what we should expect.

And anyways, was I ever taking a jab at JF-AMD? No. I posted my comment straight after reading the article. It wasn't until after I read the thread that I saw his posts.

Besides he said himself:
JF-AMDI can't say whether this data is real or not. It might be, but it is not my department, I am in server.
So how does he know that it's not official information?
Posted on Reply
#112
JF-AMD
AMD Rep (Server)
HillBeastSo how does he know that it's not official information?
I know that I was not leaked by me or my team or the desktop team. We share silicon. If they were going to purposely leak this information they would need my approval first (just like everything I do goes past them for review as well.)

That is the stock AMD template, but I just can't vouch for any slides that I did not create.

We did change our template in January with the launch of the APU, so if it is an AMD slide it would have been from last year. The dates auto populate.
Posted on Reply
#113
btarunr
Editor & Senior Moderator
JF-AMDI know that I was not leaked by me or my team or the desktop team. We share silicon. If they were going to purposely leak this information they would need my approval first (just like everything I do goes past them for review as well.)
Again, nobody claimed it to be a purposeful leak by your client team. The client guys may have shared that slide with the motherboard industry (they need to, it's part of their partner relations), and someone in the motherboard industry may have leaked it.
Posted on Reply
#114
TheMailMan78
Big Member
btarunrAgain, nobody claimed it to be a purposeful leak by your client team. The client guys may have shared that slide with the motherboard industry (they need to, it's part of their partner relations), and someone in the motherboard industry may have leaked it.
Hes just making it clear to the people who do not know how to read in this thread man. How many jackasses have you seen on here screaming how AMD sucks from these slides? You know slides AMD may or may not have released?
Posted on Reply
#115
Dent1
Wile EYou are deluded then. They charged $1000 for the FX cpus when they were ahead of Intel last time.
To be fair Intel charged even more for the P4 Extreme Editions when they were losing!
Posted on Reply
#116
TheoneandonlyMrK
regardless of some peoples dislike of the new bulldozer and APu chips it does do a lot of good NV quad core +GFX on way intel using shaders etc it pushes up the for want of a better word shitter level , the place where peeps play net games and sims and wow and if more pcs can play games at framerates on low-med settings at last (lookin at u intel) as seems amd's plan then gloat and be happy the fight goes on i luck t the discount bin for hope of a gleamin summat
Posted on Reply
#117
JF-AMD
AMD Rep (Server)
Nobody has ever been able to show me the parallel universe where more competition has led to a wose environment for customers.

Everyone should be happy that there is lots of competition and new products are coming out. They should want every product to be better than what they are expecting, not worse.

Bad products lead to complacency, not innovation.
Posted on Reply
#118
jpierce55
JF-AMDNobody has ever been able to show me the parallel universe where more competition has led to a wose environment for customers.

Everyone should be happy that there is lots of competition and new products are coming out. They should want every product to be better than what they are expecting, not worse.

Bad products lead to complacency, not innovation.
agreed
Posted on Reply
#119
LAN_deRf_HA
largonAs long as it's not Intel-made-CPU I'm a buyin' it.
That sounds reasonable.
Posted on Reply
#120
Wile E
Power User
JF-AMDNobody has ever been able to show me the parallel universe where more competition has led to a wose environment for customers.

Everyone should be happy that there is lots of competition and new products are coming out. They should want every product to be better than what they are expecting, not worse.

Bad products lead to complacency, not innovation.
I can get behind that statement.
Posted on Reply
#121
MarcusTaz
Wile EI can get behind that statement.
Agreed!
Posted on Reply
#122
largon
LAN_deRf_HA
largonAs long as it's not Intel-made-CPU I'm a buyin' it.
That sounds reasonable.
I consider myself having a strong sense of justice.
If a company shows such indifference to laws and good sales practices, as Intel has been, that company does not get my $$$ no matter how good their product is.
Posted on Reply
Add your own comment
May 5th, 2024 06:28 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts