Fudzilla: Bulldozer performance figures are in

[BlueBlazer]

Though "considerably" is not a very accurate word..

Here anand s review at the time :
http://www.anandtech.com/show/2045/11

For IPC, you need to compare them clock-to-clock (same clock speed) and immediately the gap is noticeable. In other applications and CPU bound gaming, these gaps can be much wider depending on application/game.

Pre-order prices of AMD FX-Series CPUs
If price reflects the relative performance then I have a bad feeling that BD is going to be a flop. :hmm:

Somehow it does correlate a bit to the recent benchmark leaks. Seems those FX-81xx CPUs are within Sandy Bridge Core i5 territory. :hmm:

$190 for the FX-6100? That's the same as the current X6 1100T...

Those prices look too low.

Just initial/introductory pricing. It may go lower than that. Quote from above link....

AMD FX-6100 has only 6 CPU cores, or 3 Bulldozer modules. As such, it has only 6 MB of L2 cache. The processor is clocked at 3.3 GHz, and up to 3.9 GHz when Turbo Core feature is active. Like eight-core CPUs, this model has 8 MB L3 cache. The FX-6100 is listed at $188, and its expected official price is $155 - $170, or close to current official price of Phenom II X6 1055T.

 

[Accord99]

Yes , in sysmark tests it was quite better......
Though "considerably" is not a very accurate word..

Here anand s review at the time :
http://www.anandtech.com/show/2045/11

The 2.4GHz E6600 easily matches a 2.8GHz FX-62 in applications and has a measurable advantage in games. Therefore AMD needs to clock their K8 CPUs 15-25% higher to match Conroe, which is a pretty significant disadvantage at that time relative to past matchups. And that ignores the other factor that made Conroe a hit with enthusiasts and enthusiast sites, its higher overclocking limits.

And really, it only looks not so large today because of the 40+% single-thread and 70+% throughput advantage Sandy Bridge cores have today over a K10 core. But even after Conroe's release, would any bother in 2006 seriously believed that in 2011, a modern Intel core would nearly have the same throughput as two AMD cores of the same clock speed?

 

[Riek]

$190 for the FX-6100? That's the same as the current X6 1100T...

Those prices look too low.

No i don't think so for the 6100. It has the same frequency range as the 1100T (3.3-3.9 vs 3.3-3.7).
It has a drawback in multithreaded due to shared front end which gives lower scaling. (Thats like comparing 6*~0.95 vs 3* ~1.8)

So if the FX6 is equallying the 1100T in overall performance that would be great... That would indicate that the FX6 would be more competitive in low threaded tasks while being competitive in high multithreaded threaded. Just what the current lineup was missing compared to the i5. Better in MT but to slow in LT. All this with a much lower TDP. (llano cannot reproduce that feat with the current process. Given BD uses the same process i would give them some slack for that.)

I would believe the FX4170 (if that one would exist) would be priced the same. Given it will have a higher LT performance and slightly lower MT compared to the FX6100.

 

[AtenRa]

No i don't think so for the 6100. It has the same frequency range as the 1100T (3.3-3.9 vs 3.3-3.7).
It has a drawback in multithreaded due to shared front end which gives lower scaling. (Thats like comparing 6*~0.95 vs 3* ~1.8)


So if the FX6 is equallying the 1100T in overall performance that would be great... That would indicate that the FX6 would be more competitive in low threaded tasks while being competitive in high threaded. Just what the current lineup was missing compared to the i5. Better in MT but to slow in LT.

I would believe the FX4170 (if that one would exist) would be priced the same. Given it will have a higher LT performance and slightly lower MT compared to the FX6100.

So, a dual core Phenom II will have 1.9 vs 1.8 of BD module.

That means that BD only needs a 10% more IPC to much Phenom II in Multithread and be 10% faster in single thread, correct ??

I have the feeling BD will have more than 10% higher IPC than Phenom II

 

[Riek]

For IPC, you need to compare them clock-to-clock (same clock speed) and immediately the gap is noticeable.

Wrong, you would measure the performance/clock. not the instruction per clock which can differ alot between code path the cpu follows. Bu yeah there is a gap, always was with Core Duo.

So, a dual core Phenom II will have 1.9 vs 1.8 of BD module.

That means that BD only needs a 10% more IPC to much Phenom II in Multithread and be 10% faster in single thread, correct ??

I have the feeling BD will have more than 10% higher IPC than Phenom II

Actually i believe they won't differ that much. I expect at the same frequency a dual core llano would be slightly better then a module.

But also at the same frequency BD uses less energy. (8core BD 3.1GHz is possible under 95W TDP, whereass llano is flirting with 70-80W TDP with 4cores at 2.9GHz).
BD can have a much higher frequency where llano is at its limit around 3.5GHz(or so it seems).

I use llano as a reference because they use the same procedee and they are having issues with that procedee. As itcurrently stands llano in 6core form on this procedee won't be competitive with x6 1100T either. I don't even think they could fit 6llano cores at 3GHz under 95W TDP wihtout extreem binning.

I also believe that AMD was expecting FX8150 frequencies under 95W, which should be doable if the process would be good around 1.2-1.25V instead of 1.35Vas seen with llano and which would completely change the situation.

 

[AtenRa]

Wrong, you would measure the performance/clock. not the instruction per clock which can differ alot between code path the cpu follows. Bu yeah there is a gap, always was with Core Duo.


Actually i believe they won't differ that much. I expect at the same frequency a dual core llano would be slightly better then a module.

But also at the same frequency BD uses less energy. (8core BD 3.1GHz is possible under 95W TDP, whereass llano is flirting with 70-80W TDP with 4cores at 2.9GHz).
BD can have a much higher frequency where llano is at its limit around 3.5GHz(or so it seems).

I use llano as a reference because they use the same procedee and they are having issues with that procedee. As itcurrently stands llano in 6core form on this procedee won't be competitive with x6 1100T either. I don't even think they could fit 6llano cores at 3GHz under 95W TDP wihtout extreem binning.

I also believe that AMD was expecting FX8150 frequencies under 95W, which should be doable if the process would be good around 1.2-1.25V instead of 1.35Vas seen with llano and which would completely change the situation.

Llano has almost the same or better performance per clock as Phenom II, but i suspect that Llanos CPU cores use smaller(Gate Length) transistors for higher transistor density that translated in to smaller die.

Smaller transistors in the same process will need more Vt and will have lower performance making Llanos CPU cores use more power and not be able to raise higher clocks.

If AMD managed to pull out another 5-6% more IPC from Athlon II core with only minor upgrades in the CPU core, just imagine what they would be able to make with BD core and all those extended upgrades in the Front end, execution units and more.

So, comparing Llano with BD will not make you come to a correct conclusion of BDs performance.

 

[Riek]

If AMD managed to pull out another 5-6% more IPC from Athlon II core with only minor upgrades in the CPU core, just imagine what they would be able to make with BD core and all those extended upgrades in the Front end, execution units and more.

So, comparing Llano with BD will not make you come to a correct conclusion of BDs performance.

But bulldozer isn't based on phenom. So while they can up the ipc of a known design, designing something completely knew will have a completely new baseline. So yeah comparing ipc is difficult between them because they have nothing in common.
There are absolutely no restrictions to where Bulldozer would land based on where phenom2 was.

But performance leaks seem to indicate a similar ipc (although it will be reached differently) for both designs. (at least for integer)

 

[LOL_Wut_Axel]

No i don't think so for the 6100. It has the same frequency range as the 1100T (3.3-3.9 vs 3.3-3.7).
It has a drawback in multithreaded due to shared front end which gives lower scaling. (Thats like comparing 6*~0.95 vs 3* ~1.8)

So if the FX6 is equallying the 1100T in overall performance that would be great... That would indicate that the FX6 would be more competitive in low threaded tasks while being competitive in high multithreaded threaded. Just what the current lineup was missing compared to the i5. Better in MT but to slow in LT. All this with a much lower TDP. (llano cannot reproduce that feat with the current process. Given BD uses the same process i would give them some slack for that.)

I would believe the FX4170 (if that one would exist) would be priced the same. Given it will have a higher LT performance and slightly lower MT compared to the FX6100.

The way I'm looking at it is: if the FX-6100 and FX-4170 (if that one exists) are priced at around $190-200, does it mean that AMD is trying to be very competitive with Intel and get them to lower prices for the Core i5-2500 and 2500K, or is it a message that AMD's performance is low in comparison, therefore the price is lower as well? If it's priced so low it probably means AMD has made very little to improve their IPC, especially with the 4170 running at 4.2GHz.

 

[Mem]

The way I'm looking at it is: if the FX-6100 and FX-4170 (if that one exists) are priced at around $190-200, does it mean that AMD is trying to be very competitive with Intel and get them to lower prices for the Core i5-2500 and 2500K, or is it a message that AMD's performance is low in comparison, therefore the price is lower as well? If it's priced so low it probably means AMD has made very little to improve their IPC, especially with the 4170 running at 4.2GHz.

My best speculative guess is probably performance is not as high as we and also AMD had all hoped so pricing is adjusted and according to performance.

It'll be interesting to see what the real official BD benchmarks look like.

 

[Riek]

The way I'm looking at it is: if the FX-6100 and FX-4170 (if that one exists) are priced at around $190-200, does it mean that AMD is trying to be very competitive with Intel and get them to lower prices for the Core i5-2500 and 2500K, or is it a message that AMD's performance is low in comparison, therefore the price is lower as well? If it's priced so low it probably means AMD has made very little to improve their IPC, especially with the 4170 running at 4.2GHz.

Well currently AMD has:
Phenom II X4 980 ‘Black Edition’(3.7GHz, 125W, , 2MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1090T 'Black Edition' (3.2GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1100T 'Black Edition' (3.3GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$205


So they are doing the same thing as they did previously. Have their highest clocked 4 core part around the price of the lower clocked 6core part.

AMD should always have an advantage compared to intel if they want to sell. They are the underdog.

So they either have to deliver same performance for lower price
Or higher performance for same price
Or same performance, same price but additional possibilities, selling points.

Not sure which route they will (or are forced to take).

 

[LOL_Wut_Axel]

Well currently AMD has:
Phenom II X4 980 ‘Black Edition’(3.7GHz, 125W, , 2MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1090T 'Black Edition' (3.2GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1100T 'Black Edition' (3.3GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$205


So they are doing the same thing as they did previously. Have their highest clocked 4 core part around the price of the lower clocked 6core part.

AMD should always have an advantage compared to intel if they want to sell. They are the underdog.

So they either have to deliver same performance for lower price
Or higher performance for same price
Or same performance, same price but additional possibilities, selling points.

Not sure which route they will (or are forced to take).

Hmm, true. But then, if IPC were improved by a decent amount, it would mean that both the Six and Quad-Cores are trying to be cheaper AND faster, which I doubt is true. If performance was competitive they wouldn't need to get out so many different steppings and revisions, and to also wait for the process to mature more. They also wouldn't need to launch the FX-4170 at such a skyrocketing frequency, one which is probably a bit difficult to meet at AMD's target 95W TDP and given the process node of the new chips isn't anywhere near as mature as Intel's.

 

[Riek]

Hmm, true. But then, if IPC were improved by a decent amount, it would mean that both the Six and Quad-Cores are trying to be cheaper AND faster, which I doubt is true. If performance was competitive they wouldn't need to get out so many different steppings and revisions, and to also wait for the process to mature more. They also wouldn't need to launch the FX-4170 at such a skyrocketing frequency, one which is probably a bit difficult to meet at AMD's target 95W TDP and given the process node of the new chips isn't anywhere near as mature as Intel's.

I thought the target for the rumoured FX4170 was rumoured to be 125W.

Well you can consider that delaying and steppings point to an issue and most likely a performace issue. But all that shouldn't matter to the actual performance. the reason why they delay is to reach the performnce goal at which they want to release. So the slipping shouldn't affect the performance but just the release date.

I think the FX4170 and FX6100 are a good deal. The user can choose between exceptional low thread performance or more multitask oriented performance when they are on a tight budget. When they have more budget they can choose both with the 8120 and 8150.


It could be the FX6 it a little to slow between 1-4 threads and relies on the 5-6 threads to become really competitive over the whole line.
The FX4 might be faster in all cases (given it has 4 cores running at wazoo speed) but if it has a 125W TDP is also has a measurable on the box drawback.

 

[RussianSensation]

Well currently AMD has:
Phenom II X4 980 ‘Black Edition’(3.7GHz, 125W, , 2MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1090T 'Black Edition' (3.2GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$185

Phenom II X6 1100T 'Black Edition' (3.3GHz, 125W, 3MB total dedicated L2 cache, 6MB L3 cache, 4000MHz HyperTransport™ bus, socket AM3)
$205

Real world is lower:

X4 980 = $169.99, but who would buy that over the $20 more expensive i5-2400, or even over the 1090T? This CPU is pointless.

X6 1090T = $159.99-169.99 <This chip really makes the X4 980 irrelevant when considering an AMD CPU>

X6 1100T = $189.99. Again, a very difficult buy vs. an i5-2400 and is still only $30 cheaper than the i5-2500k. Given the performance parity between an 1100T and the 2500k, the 1100T would likely need to be priced at $129-139 to make sense. That means AMD knows that people who buy 1100T want 6 cores and don't care about IPC or single threaded apps. Which means the same buyers don't need higher IPC for the FX-8100 series to justify their purchase since they prioritize cores over per core performance.

Looking at AMD's pricing, you would think their CPUs are very close in performance to Intel's, and yet the performance gap is actually far more than their pricing strategy indicates.

I think the FX4170 and FX6100 are a good deal. The user can choose between exceptional low thread performance or more multitask oriented performance when they are on a tight budget.

We don't know that. If FX-8120 is ~$225, then it will be positioned right against 2500k. If it has exceptional low threaded performance (i.e., Nehalem or SB style), then it would be 70-90% faster in 8 threaded apps vs. the 2500k. Why would AMD sell a processor with performance similar to a 2500k in 4-threaded apps and yet almost 2x faster in multi-threaded apps for the same price? Doesn't make any sense considering they are selling way slower CPUs now just $20-30 away from processors that mop the floor with them. If anything, it makes sense that AMD decided to focus on more cores since that likely is the only way for them to compete. They may have decided to focus on the lucrative server market and catering to users who want more cores (likely the same people who paid $190 for the 1100T).

 

[sequoia464]

I thought the target for the rumoured FX4170 was rumoured to be 125W.

I have seen charts showing the 4 and 6 core units at 95W - don't know if they are accurate or not. One of the lower frequency 8 cores is also - eventually - supposed to be released as a 95W part.

I'm sure someone can correct this if it is wrong - did a search for the chart but can't locate it now.

 

[LOL_Wut_Axel]

I thought the target for the rumoured FX4170 was rumoured to be 125W.

Well you can consider that delaying and steppings point to an issue and most likely a performace issue. But all that shouldn't matter to the actual performance. the reason why they delay is to reach the performnce goal at which they want to release. So the slipping shouldn't affect the performance but just the release date.

I think the FX4170 and FX6100 are a good deal. The user can choose between exceptional low thread performance or more multitask oriented performance when they are on a tight budget. When they have more budget they can choose both with the 8120 and 8150.


It could be the FX6 it a little to slow between 1-4 threads and relies on the 5-6 threads to become really competitive over the whole line.
The FX4 might be faster in all cases (given it has 4 cores running at wazoo speed) but if it has a 125W TDP is also has a measurable on the box drawback.

Even worse, then. And the pricing makes no sense, especially considering these are all unlocked CPUs. Why buy the FX-4170 over the 6100 if you're gonna overclock, anyway? If there's one thing that history has taught us it's that the CPUs of a family with the highest cores typically overclock the same as the variants with less cores, but produce more heat due to consuming more power. Given that, the only ones that really make sense are the FX-8100 and the FX-8120. The FX-4170 will probably be abysmal in everything overclocked, as it'll OC the same as the Six and Eight-Cores.

EDIT:

And why would someone get the FX-4170 over the A8-3870?

 

[Riek]

Real world is lower:

X4 980 = $169.99, but who would buy that over the $20 more expensive i5-2400, or even over the 1090T? This CPU is pointless.

X6 1090T = $159.99-169.99 <This chip really makes the X4 980 irrelevant when considering an AMD CPU>

X6 1100T = $189.99. Again, a very difficult buy vs. an i5-2400 and is still only $30 cheaper than the i5-2500k. Given the performance parity between an 1100T and the 2500k, the 1100T would likely need to be priced at $129-139 to make sense. That means AMD knows that people who buy 1100T want 6 cores and don't care about IPC or single threaded apps. Which means the same buyers don't need higher IPC for the FX-8100 series to justify their purchase since they prioritize cores over per core performance.

Looking at AMD's pricing, you would think their CPUs are very close in performance to Intel's, and yet the performance gap is actually far more than their pricing strategy indicates.



We don't know that. If FX-8120 is ~$225, then it will be positioned right against 2500k. If it has exceptional low threaded performance (i.e., Nehalem or SB style), then it would be 70-90% faster in 8 threaded apps vs. the 2500k. Why would AMD sell a processor with performance similar to a 2500k in 4-threaded apps and yet almost 2x faster in multi-threaded apps for the same price? Doesn't make any sense considering they are selling way slower CPUs now just $20-30 away from processors that mop the floor with them. If anything, it makes sense that AMD decided to focus on more cores since that likely is the only way for them to compete. They may have decided to focus on the lucrative server market and catering to users who want more cores (likely the same people who paid $190 for the 1100T).

The x6 cores are still faster in threaded applications than the i5 4core series. The Deneb is also pretty competitive against the low end i5s and i3s. Yes they don't perform like intel in gaming. But overal the distance isn't that big as you believe. (they really do pretty good in some more intensive applications). If the FX6 can get better in the low thread performance while giving a little in the high threaded applications it would become a great competitor in its price range.

We don't know that. If FX-8120 is ~$225, then it will be positioned right against 2500k. If it has exceptional low threaded performance (i.e., Nehalem or SB style), then it would be 70-90% faster in 8 threaded apps vs. the 2500k. Why would AMD sell a processor with performance similar to a 2500k in 4-threaded apps and yet almost 2x faster in multi-threaded apps for the same price?

Again with the errors.
FX8120 would run 1-4Threads at 4GHz but it would 8thread at 3.1GHz.

That would make that you get the theoretical double performance of 25% less.

So if you first scored 100, you would score with 8cores 150. Thats a 50% increase and not a 70-90%.

If you give HT to the 2500 without anything else the FX8120 would just be faster by 20% in certain applications that scale well with cores.


However the 8120 would also be 5% (insignificant difference imo) slower in 1-4 threads then the FX4170 or FX8150, so its 1-4thread performance can still be lower then 2500 by a few %.


Also if you don't find 2500 low threaded(1-4) performance exceptional than i don't know what you find exceptional? given the 2600, 2700, SB-E will all perform within 5-6% from that.

 

[Abwx]

Given the performance parity between an 1100T and the 2500k, the 1100T would likely need to be priced at $129-139 to make sense. .

That is , at the same price as an i3 21XX...

That s total non sense , mind you...

 

[LOL_Wut_Axel]

*snip*

Gonna have to disagree with you on that one. The Phenom II X6 has comparable multi-threaded performance to the Core i5 at the same clock speed, so the price is definitely right in line.

 

[996GT2]

Gonna have to disagree with you on that one. The Phenom II X6 has comparable multi-threaded performance to the Core i5 at the same clock speed, so the price is definitely right in line.

1100T and 2500K are similar in heavily multithreaded apps when both are at stock speeds. 1100T is significantly slower in single threaded apps.

However, the 2500K has a lot more overclocking headroom than the 1100T. You'd be lucky to break 4 GHz with an 1100T (and that's using a lot of voltage in most cases), whereas almost every 2500K can hit 4.7 GHz at moderate voltages. That furthers that gap between the two.

Not to mention that a 2500K system at 4.7 GHz still draws less power than an 1100T system at stock.

If you look at prices, 2500K is $220 while 1100T is $190. Pretty much a no-brainer in favor of the 2500K.

 

[AtenRa]

AMD at IFA 2011

http://www.youtube.com/watch?feature=player_embedded&v=sQWVk6ZDLHg

 

[LOL_Wut_Axel]

1100T and 2500K are similar in heavily multithreaded apps when both are at stock speeds. 1100T is significantly slower in single threaded apps.

However, the 2500K has a lot more overclocking headroom than the 1100T. You'd be lucky to break 4 GHz with an 1100T (and that's using a lot of voltage in most cases), whereas almost every 2500K can hit 4.7 GHz at moderate voltages. That furthers that gap between the two.

Not to mention that a 2500K system at 4.7 GHz still draws less power than an 1100T system at stock.

If you look at prices, 2500K is $220 while 1100T is $190. Pretty much a no-brainer in favor of the 2500K.

Yeah, but then you look at the Phenom II X6 1055T and the price/performance argument shifts back to AMD. Overclocked it reaches 4GHz on average. Also, the average for Sandy Bridge is 4.5GHz, not 4.7. You're overestimating. Just like some SB CPUs can reach 4.7GHz, some Phenom II X6s can reach 4.2GHz. But since the 2500K would OC higher, then it'd be 10-15&#37; faster. It also costs 45% more, though. For a cheap machine for encoding, I'd say the X6 is the best bang-for-buck.

 

[996GT2]

Yeah, but then you look at the Phenom II X6 1055T and the price/performance argument shifts back to AMD. Overclocked it reaches 4GHz on average. Also, the average for Sandy Bridge is 4.5GHz, not 4.7. You're overestimating. Just like some SB CPUs can reach 4.7GHz, some Phenom II X6s can reach 4.2GHz. But since the 2500K would OC higher, then it'd be 10-15&#37; faster. It also costs 45% more, though. For a cheap machine for encoding, I'd say the X6 is the best bang-for-buck.

I do agree that for a low-budget gaming or encoding build, a Phenom II X4 or X6 makes sense. After all, you can spend the extra $50-100 on a better GPU.

But once you get into that midrange and up segment, 2500K/2600K is really a no-brainer. And judging by Bulldozer's pricing, it seems that AMD might once again be relegated to the low-end even with Bulldozer.

 

[LOL_Wut_Axel]

I do agree that for a low-budget gaming or encoding build, a Phenom II X4 or X6 makes sense. After all, you can spend the extra $50-100 on a better GPU.

But once you get into that midrange and up segment, 2500K/2600K is really a no-brainer. And judging by Bulldozer's pricing, it seems that AMD might once again be relegated to the low-end even with Bulldozer.

Well, I wouldn't say low-end. Remember that Llano is Mainstream. Bulldozer was supposed to be part of the Performance market, but as of now it looks like not of the higher-end tiers (where the Core i7-2600K is).

2500K is Performance (high-end, but not enthusiast), BTW.

Looks like Bulldozer will only be good for HPC, encoding, content creation, and rendering.

 

[RussianSensation]

The x6 cores are still faster in threaded applications than the i5 4core series.

Which apps are those? 7-zip?

i5 4 core still beats it in most image editing, audio processing and archiving apps. And that's without taking overclocking into account.

The Deneb is also pretty competitive against the low end i5s and i3s. Yes they don't perform like intel in gaming. But overal the distance isn't that big as you believe.

When AMD can't beat a 2 core HT i3 2120, that's pretty bad to me. The more important question is, why would I pay more for a slower CPU with worse power consumption and worse overclocking?

Also if you don't find 2500 low threaded(1-4) performance exceptional than i don't know what you find exceptional? given the 2600, 2700, SB-E will all perform within 5-6&#37; from that.

I have no idea where you got this idea from my post.

That is , at the same price as an i3 21XX...

That s total non sense , mind you...

Is it?

$130 X6 + $130 mobo + $50 DDR3 8GB ram = $310 for the setup.
$220 i5 2500k + $130 mobo + $50 DDR3 8GB = $400

30% more expensive but you get much lower power consumption, QuickSync for video encoding to your tablet / smartphone and more than 30% faster in overclocked states with nearly half the power consumption. See 996GT2 posts. Even at $130, I still can't justify the X6. AMD is trying to throw more cores like a mhz myth of Pentium 4 era.

Right now at $170-190, the X6 is not even a consideration, unless you just want to support AMD and completely disregard performance, power consumption and overclocking. The X6 needs a $30-40 price cut to make sense vs. a quad core SB. Looks like we'll get that with the FX-6100 series.

In the context of the overall system cost, it gets even worse. You save what $30-40 over $600-700 it costs to build a good rig to buy a far inferior CPU? Please explain how that makes sense esp. since such a system will last 2-3 years at least. It's no wonder that AMD's strategy to offer 6 cores against 4 core Intel CPUs didn't do much for them overall. Perhaps, 8 cores this time may sway more people.

It's interesting how AMD users purchased AM3+ boards with slow X4/X6 CPUs, and are eager to upgrade those for BD. Maybe finally their upgrade path will allow them to catch up to a 2008 Core i7 920 @ 4.0ghz...albeit 3 years later.

 

[RussianSensation]

Originally Posted by Jarred Walton

Unfortunately, all signs point to Zambezi being less than stellar; I suspect that clock for clock, a single BD core will be slower than current K10.5 stuff, but you'll get more cores. It will also be interesting to see how Turbo Core plays out; if it's as cautious as some of the other chips, the eight core chips will only run at the base (or base + 1) frequency for anything more than dual-core workloads.

That should pretty end the argument of Nehalem-like IPC.