Bulldozer Review! Legit?

[frostedflakes]

I've mentioned this before, but I think performance of BD is going to vary wildly due to the shared resources approach between cores. Cinebench is mostly floating point, for example, and for programs not compiled with FMA4, XOP, and other Bulldozer optimizations, they will only have four FPUs to work with in a four module BD. With optimizations, though, hopefully things will perform better as then I think they will be able to send two 128-bit floating point instructions to each 256-bit FPU. At least this is how I understand it.

Integer workloads should perform great without any optimizations, though, since each core in a module has its own dedicated integer resources.

Am really anxious for a full review from AnandTech that can explain the architecture in detail and put it all into perspective.

 

[classy]

I think the issue is, we're not seeing any real improvements here. The PhenomII X6 1100T is already between a 2500k and 2600k in multi-threaded (scoring a 5.84, see below):

This while clocked @ only 3.3Ghz.

Then you have "8 core" BD, clocked @ 3.6Ghz, and it's barely any faster, despite higher clocks and 2 extra cores.

In addition to this, you have worse single-threaded performance than an 1100T, again, due to the clock speed differences listed above.

That, to me, is a fail. Is that what everyone has been waiting and waiting all this time for? A slightly less IPC PhenomII "X8" that performs slightly worse than a Thuban X6 (considering the higher clock\turbo speeds)?

You put up one benchmark. A benchmark that I posted earlier has an approved list of cpus. With a brand new architecture not every piece of software will show the true potential. We looked at Battlefield 3 beta benchmarks and the 2600K loses out to a 1100T, put that benchmark up.

 

[hooflung]

Well. This means that I don't have to upgrade my 1075 anytime soon (I just got the cpu not too long ago afterall). But when I do I'll see a huge boost in my Virtual Machine environments. The fact I can probably get 16G of DDR3 and an 8 Cores of goodness cheap... Hell I should be able to play Star Craft 2 while pimping 4 Ubuntu Servers at 2G a pop without a hitch running J2EE and a smaller VM to act as a load balancer... while I kick some humang arse.

It would be nice for BD to outdo the latest Intel but lets be real a second. Only complete fanbois would expect a David vs Goliath win like the original amd64 vs P4 was. I just want lower temps, less power use, performance that is at least as good as what I have with more lateral movement.

 

[bunnyfubbles]

I like the 5GHZ @ < 1.5v overclock though.

I'm not, as that's really no better than what current SB can do, and considering that BD needs a hefty clock rate advantage to even compete at stock clocks we need BD to be in the 5.5 OC range to remain competitive.

 

[drizek]

Ya, that's the most attractive part to me. I already have my 16GB ram. I want to be able to run WS2008R2 constantly in the background, even while gaming.

 

[Venom20]

There is something called "Sandy Bridge". What are you, living in 2008?

$990 USD
$1280 CAD

if you meant the i7-965, then I repeat the question.
I assume you were referring to the i7-2600? Of course I would trust that you meant what you said.

 

[frostedflakes]

I'm not, as that's really no better than what current SB can do, and considering that BD needs a hefty clock rate advantage to even compete at stock clocks we need BD to be in the 5.5 OC range to remain competitive.

To be fair, 5GHz is pretty high for SB, I didn't think most people could hit that with a reasonable voltage whereas 5GHz may be a pretty typical OC for the FX-8150. Also can you run SB at 5GHz with HyperThreading or do you need to disable it to hit those clocks?

 

[drizek]

Not all 1.5v are created equal.

 

[Arkaign]

You put up one benchmark. A benchmark that I posted earlier has an approved list of cpus. With a brand new architecture not every piece of software will show the true potential. We looked at Battlefield 3 beta benchmarks and the 2600K loses out to a 1100T, put that benchmark up.

The BF3 beta results are so borked (i7-920 beats 2500k stock speeds? what?) that it's essentially meaningless right now. Also the framerate caps across the board are so low as to make no sense.

Final code, new drivers, and tests at both high and low resolution will tell the tale.

 

[classy]

The BF3 beta results are so borked (i7-920 beats 2500k stock speeds? what?) that it's essentially meaningless right now. Also the framerate caps across the board are so low as to make no sense.

Final code, new drivers, and tests at both high and low resolution will tell the tale.

That really is not the point. The point is that some of the posts are silly because the comparisons we have seen are against the BEST Intel has to offer. So we'll see when the final numbers are released.

 

[Idontcare]

Make you wonder how Bulldozer will do in Battlefield 3. Knowing it loves more cores.

Yep. Really any new apps that stand a chance of having been compiled with a "BD aware" compiler.

CB11.5 came along because CB10 just wasn't taking advantage of the then modern CPU's.

CB12 will be coming, and it will be BD aware.

Benches on legacy apps are relevant if you are running legacy apps and don't plan to update them.

 

[Idontcare]

I'm not, as that's really no better than what current SB can do, and considering that BD needs a hefty clock rate advantage to even compete at stock clocks we need BD to be in the 5.5 OC range to remain competitive.

price/performance though, right? It needs 5.5Ghz to be same value to us OC'ing enthusiasts IF it were priced the same as a 2600K that could OC to 5GHz.

 

[RussianSensation]

i think the 870 has dynamic turbo on..

which means it only gets 3.6 on 2 cores.. and 3.2ghz on all 4.

I dont remember if they could do 3.6 on all 4 cores with turbo on without overclocking.

You guys been around sandy too long...


So its still not making sense..

In R10, you run single-threaded bench. Under such a scenario, max turbo is enabled on the i7-870.

 

[StrangerGuy]

$990 USD
$1280 CAD

if you meant the i7-965, then I repeat the question.
I assume you were referring to the i7-2600? Of course I would trust that you meant what you said.

If you are too dense to figure out a $300 2600K is faster than a $1000 outdated chip you shouldn't be opening your load here.

 

[RussianSensation]

That really is not the point. The point is that some of the posts are silly because the comparisons we have seen are against the BEST Intel has to offer. So we'll see when the final numbers are released.

But 2500k is not the best Intel has to offer. Also comparisons to i7-965 are there probably because that's the chip they had on hand for testing the Nehalem generation. They could have put a $200-250 i7 930 @ 4.2ghz in there, or an i7-860 @ 4.0ghz, etc. The point is it is good to see how the older Intel generation compares to 2500k/2600k and BD in those benches. Most 920/930 chips hit 4.0-4.2ghz with ease; so the i7-965 is just a place holder.

You put up one benchmark. A benchmark that I posted earlier has an approved list of cpus. With a brand new architecture not every piece of software will show the true potential. We looked at Battlefield 3 beta benchmarks and the 2600K loses out to a 1100T, put that benchmark up.

I am pretty sure almost no one cares about a situation where 1 processor gets 49 fps and another 53 fps., esp when no single modern GPU can achieve those frames with AA at 1920x1080 in BF3. You know a game is not really CPU limited when a Core i7-920 and 2600k are achieving identical frame rates. Also, the higher clocked Phenom II X4 is keeping up with the X6, which means BF3 does not use more than 4 cores. Among modern CPUs, BF3 is almost entirely GPU limited. How about we put a real CPU limited game up and see how AMD does? If BD beats SB in Starcraft II, that would be impressive.

 

[classy]

Huh? 2500k is not the best Intel has to offer.

Dude I listed 4 of the 5-6 fastest cpus Intel has. You single out the 2500K, okay. It is one the fastest gaming cpus you can buy. It sounds completely retarded to read posts about how poor the performance of the Bulldozer leaked marks and they are equal or superior to the fastest Intel cpus. If there are some secret cpus that I don't know about, then please point them out. No one knows if the benches we have seen are true or not. But the last 2-3 weeks all of the info leaked would appear to support the fact that indeed, AMD has a very competitive chip.

And to prove just how stupid some of the arguments are. Here are the bench comparisons of a 2600K vs a 2500K.

http://www.anandtech.com/bench/Product/287?vs=288

There is very little difference. So to see a post criticizing an AMD benchmark compared to any of the cpus I listed is totally silly. They are the fastest that are available. If they were comparing these cpus to 655K or an I3 type, then clearly such an argument is warranted. But that is not the case.

 

[frozentundra123456]

That really is not the point. The point is that some of the posts are silly because the comparisons we have seen are against the BEST Intel has to offer. So we'll see when the final numbers are released.

Why would you NOT compare to the best that Intel has?? Bulldozer is the "best" that AMD has, or at least it had better be. If you take price into consideration, then you can compare to the best that Intel has in the same price range, which would be the 2500 or 2600. What do you want to compare it to a Celeron?

 

[classy]

And let me add this. I think in being fair, if they have made these kinds of gains with their first production silicon of this chip they have done a very good job. People forget Intel has been building on yonah for quite some time now. To see AMD not overtake, but leap up into the upper cpu performance arena if that is the case, then they should commended, not criticized.

 

[classy]

Why would you NOT compare to the best that Intel has?? Bulldozer is the "best" that AMD has, or at least it had better be. If you take price into consideration, then you can compare to the best that Intel has in the same price range, which would be the 2500 or 2600. What do you want to compare it to a Celeron?

Huh? Many of the leaked benches are against the 2500K and 2600K. But to pretend that the the I7 965 and 980X are somehow slow dogs ain't nothing more than fanboyism.

 

[Black96ws6]

Here is a link to another of their reviews on the A8:

http://www.pcmweb.nl/tags/amd/a8-3850

The format is the same. Check it out.

99.99&#37; chance these leaked benches are accurate.

Which is sad.

Because it means that BD gets about the same scores, in both single and multi-threaded, as an 1100T.

But the problem is, BD is an 8 core chip, vs a 6 core, and it's also clocked higher.

If you downclocked it to match the 1100T's 3.3Ghz stock speed, BD would LOSE to it. And BD has 2 more "cores"!

Before pulling the trigger on a 2500k\2600k, I waited to see if BD was worth the wait to upgrade, and now, it looks like I wasted my time...

 

[Olikan]

more benchmarks

well OBR leaked some of those slides, but not all of them...
interesting the slide about windons 8

 

[bunnyfubbles]

To be fair, 5GHz is pretty high for SB, I didn't think most people could hit that with a reasonable voltage whereas 5GHz may be a pretty typical OC for the FX-8150. Also can you run SB at 5GHz with HyperThreading or do you need to disable it to hit those clocks?

hitting 5GHz on SB isn't that hard when you force 1.5v down its throat, which is what this review got with BD...and 1.5v isn't exactly reasonable for 24/7 use. It'd be a much different story if they were able to easily get BD to 5GHz @ 1.35v without ridiculous cooling and prove it to be rock solid stable

even if BD turns out to be a golden overclocker and can regularly hit above 5GHz with reasonable volts, again, that doesn't mean much if its consistently outperformed by 2500K/2600K rigs that have been taken to their own reasonable limits

price/performance though, right? It needs 5.5Ghz to be same value to us OC'ing enthusiasts IF it were priced the same as a 2600K that could OC to 5GHz.

well we still need to see some more/better (or even real?) reviews to see where everything shakes out, with numbers like this if BD can only "reliably" hit ~5GHz @ 1.5v that doesn't exactly put AMD in any sort of favorable position when Sandy's K series can regularly hit 4.7-4.8. We still don't know enough to make any definitive conclusions from from this one "review" (we don't know what kind of cooler or temps they were getting, let alone to what degree of stress testing they employed and resulting stability or lack thereof)

BD might certainly make an interesting price/performance option for highly threaded situations, but with early numbers like this showing it to be poorer than AMD's own Phenom 2 chips in IPC, it simply doesn't bode well for its overall value in today's software landscape.

 

[Accord99]

And let me add this. I think in being fair, if they have made these kinds of gains with their first production silicon of this chip they have done a very good job.

Assuming all the leaks and slides are correct, I disagree, AMD hasn't made any gains with BD. It is the most disastrous CPU launch for AMD since the K5. They have never been farther behind.

People forget Intel has been building on yonah for quite some time now. To see AMD not overtake, but leap up into the upper cpu performance arena if that is the case, then they should commended, not criticized.

They haven't leaped anywhere, they would be worst off now than when it was Thuban versus Lynnfield, where at least Thuban had a small throughput advantage against competing Hyperthread-ed Lynnfields. Now, it appears they don't even have that. For all that talk of CMT versus SMT, it turns out that Intel has lapped AMD and built a core that matches or exceeds the throughput of a full BD module with two threads. With one thread, the advantage looks like it'll be 50% or more.

I doubt anybody 12 months ago would have thought that to be possible.

 

[RussianSensation]

Huh? Many of the leaked benches are against the 2500K and 2600K. But to pretend that the the I7 965 and 980X are somehow slow dogs ain't nothing more than fanboyism.

I am not understanding what you are trying to get across. Are you saying that because BD puts up mixed #s with i7-965 and 2500k/2600k that's it's suddenly a great chip? From those leaked benches, if true, it can't beat either in single threaded performance. This means that an overclocked Nehalem, or 2500k/2600k will still smoke it even if BD is overclocked to 4.8-5.0ghz. And considering those Cinebench results, it's barely better than a $190 X6 1100T that has been available for a long time.

This is what a lot of people are thinking:

1) If I want a top-notch multi-threaded CPU, I'll get a $159.99 X6 1090T and overclock it to 4.0ghz. And based on those 'leaks' the $189.99 FX-6100 wouldn't stand a chance against the X6 1090T.

2) If I want a top-notch all-around performance CPU, I'll get a $220 2500k.

3) If I want a top-notch fastest performance CPU without breaking my wallet, I'll get a 2600k.

Where do FX-8120 for $220 or the FX-8150 for $260 fit in this picture? If those benchmarks are true, those prices are way too high, esp. since most programs can't use more than 4 threads. It was somewhat acceptable to pay $160-180 for the X6 since you were getting unbeatable multi-threaded performance for the price. However, at $250+, the FX-8150 would need to beat the 2500k in all situations: (1) lower power consumption in overclocked states; (2) faster performance in single and multi-threaded apps. But it looks like it barely performs faster than the X6?

I am obviously going to wait for official benchmarks, but not a single leak thus far has shown that the FX is really worthy of the "FX" brand name. The Athlon FX dominated Pentium 4 in power consumption, performance per core, IPC and overclocking headroom!

Wouldn't you have expected an 8 core FX-8150 to at least beat 2600k in the most heavily multi-threaded apps such as Cinebench? That should have been a given.

 

[classy]

Assuming all the leaks and slides are correct, I disagree, AMD hasn't made any gains with BD. It is the most disastrous CPU launch for AMD since the K5. They have never been farther behind.


They haven't leaped anywhere, they would be worst off now than when it was Thuban versus Lynnfield, where at least Thuban had a small throughput advantage against competing Hyperthread-ed Lynnfields. Now, it appears they don't even have that. For all that talk of CMT versus SMT, it turns out that Intel has lapped AMD and built a core that matches or exceeds the throughput of a full BD module with two threads. With one thread, the advantage looks like it'll be 50% or more.

I doubt anybody 12 months ago would have thought that to be possible.

Okay, I want you to explain to me how this would be a disaster to release cpus equally as fast as your competitor? As I mentioned before all the technical talk doesn't change if product A is equally as fast as product B. You guys just make no sense. Its like saying a porsche isn't as good as a vette because it has a higher reving engine.