AMD "Bulldozer" FX Model 8150 CPU Overclocking Preview @ [H]

ebduncan said:
Honestly i would care less if the IPC was 4 times better, than one part, the other part could have 4 times the clock speed and it would be just as fast.
Click to expand...

This would be a great argument...

...if it weren't for the fact that Sandy Bridge is hitting 5Ghz on air.

Seeing that the dual 120 radiator water cooling system they demoed in that video only could get bulldozer up to 4.8ghz, means that AMD can't play the Ghz game to make up for IPC.
 
Rauelius said:
I would take a preview like this:

http://www.tomshardware.com/reviews/core-i7-3960x-x79-performance,3026.html

Over an overclocking video any day of the week.

Lets see...Sandy Bridge-E is due for late November...and we already have benchmarks. Bulldozer has nothing in respect to reputable benchmarks. I'm saying it! Bulldozer is going to be delayed until Q2 2012 maybe Q3. Prove me wrong AMD, I dare you!

On the upside SB-e doesn't seem as ULTRA fast as I was guessing. So if BD has an IPC close to or similar to Nehalem, and OC's better then SB, we might actually have a contender! That is if it does come out this year. Like I said, I foresee Q2 2012 at the earliest. I mean, come on, even the rumored delay says October, but no benchmarks, SB-e is due for November and we have benchmarks....Solidifies my belief that buying AMD at this time is pointless as there isn't going to be an upgrade path till half-way through next year...Like I said, prove me wrong AMD....they won't because they either can't or have a garbage chip.
Click to expand...

Thats more like a review, rather than preview.
but they cannot release any information on how fast it is so they can only release this kind of information answering one question alot of poeple are curious about, will it oc well on air.
2nd. (not far to many but still some)
coldbug
3rd. ( very few)
LN2 OC etc

And I wudnt complain about any information we can get regarding a product under NDA.
 
AMD's choice to be so secretive about it shows a lack of faith in their product. The fact that Intel allowed a preview if SB-e with actual benchmarks (and frankly I'm not ultra impressed) while AMD puts out this overvlocking video which is much less relevant to most if us.
 
Mize said:
Now we just have to hope that [email protected] GHz is reasonably close to [email protected] GHz. Would love to see some competition for intel...most every i7-2600k will hit 5GHz on stock air so AMD have their work cut out.
Click to expand...

Best joke ever :rolleyes:

I hardly keep it under "decent" temps with NH D14,HAF932 and i got stock HAF932 rear fan replaced with Zalman ZM F4 and the temps are hardly getting normal, they are quite high around 83c..during prime 5ghz 1.42vcore, its summer here.

Now, show me how you run prime 95 (with max memory setting - custom) @5ghz, under 90c and ill give you my D14 for free.
 
Rauelius said:
I would take a preview like this:

http://www.tomshardware.com/reviews/core-i7-3960x-x79-performance,3026.html

Over an overclocking video any day of the week.

Lets see...Sandy Bridge-E is due for late November...and we already have benchmarks. Bulldozer has nothing in respect to reputable benchmarks. I'm saying it! Bulldozer is going to be delayed until Q2 2012 maybe Q3. Prove me wrong AMD, I dare you!

On the upside SB-e doesn't seem as ULTRA fast as I was guessing. So if BD has an IPC close to or similar to Nehalem, and OC's better then SB, we might actually have a contender! That is if it does come out this year. Like I said, I foresee Q2 2012 at the earliest. I mean, come on, even the rumored delay says October, but no benchmarks, SB-e is due for November and we have benchmarks....Solidifies my belief that buying AMD at this time is pointless as there isn't going to be an upgrade path till half-way through next year...Like I said, prove me wrong AMD....they won't because they either can't or have a garbage chip.
Click to expand...

Yes, let's make it Q3 2012 to be super sure. :rolleyes:

Prove you wrong? Seriously, who the fk are you? Some random Joe on a internet forum making challenges to a major corporation. :rolleyes:
WTF is this, kindergarten? I DOUBLE DOGGY DARE YOU.
 
psyside said:
Best joke ever :rolleyes:

I hardly keep it under "decent" temps with NH D14,HAF932 and i got stock HAF932 rear fan replaced with Zalman ZM F4 and the temps are hardly getting normal, they are quite high around 83c..during prime 5ghz 1.42vcore, its summer here.

Now, show me how you run prime 95 (with max memory setting - custom) @5ghz, under 90c and ill give you my D14 for free.
Click to expand...

Pretty much this. Posted at 5ghz, on stock cooling, my 2500k was loading around 98c, lol.
 
Rauelius said:
Been there done that AMD. What's the point of 8Ghz if the IPC is crap? Might as well get this:
20110128203738!Trollface.png
Click to expand...
They wanted to show how it overclocks.
Your point sir?
 
I've used Intel chips for as long as I can remember, yet I've chosen a BD build because I feel that AMD has a great product here; not because I think it'll outperform SB.

As far as performance goes, I'm good long as it performs in the manner that I expect it to. Having read up on the product and viewed the various demonstrations, I'm confident it will exceed my needs as an avid gamer and graphic designer.

I feel no need to weigh in on Intel or the size of my penis for having made the decision to buy AMD this go round. I did, however, predict with accuracy that AMD haters and staunch Intel users would ring through tech forums all across the land with "That don't mean nothing," after yesterday's news. Indeed, the appearance of "2500K" and "2600K" in AMD related forms everywhere will increase ten fold as the saviors of the world attempt to besmirch their fellow gamers with information that everyone already knows.

Utter silliness.
 
Zarathustra[H];1037750762 said:
This would be a great argument...

...if it weren't for the fact that Sandy Bridge is hitting 5Ghz on air.

Seeing that the dual 120 radiator water cooling system they demoed in that video only could get bulldozer up to 4.8ghz, means that AMD can't play the Ghz game to make up for IPC.
Click to expand...

Some hit 5ghz on air. It's not a given. Beat that drum and every prospective 2500k or 2600k buyer will think they can hit that and/or be sorely dissappointed that they might not be able to with their particular sample.

Btw this is only the first production silicon of the 'high performance' 32nm Global Foundries process. I would strongly believe Llano is made to a slightly different recipe. Intel's 32nm is VERY mature, whereas the GF one is very new. If the progress Deneb made over it's lifetime, or even say from early Deneb (x4 940) to Late Thuban (1100t), then we might be in for quite the mhz scaling ride with Bulldozer. That's a 'might' of course.
 
I do see the big deal of hitting 8.xx GHz, the machines didn't boot at that speed they were just changing the multiplier in the OS. If they ran at 8.xx GHz and were able to post and boot at that speed that would be one thing, but changing the multiplier until it blue screens means nothing.
 
Action_Parsnip said:
Some hit 5ghz on air. It's not a given. Beat that drum and every prospective 2500k or 2600k buyer will think they can hit that and/or be sorely dissappointed that they might not be able to with their particular sample.

Btw this is only the first production silicon of the 'high performance' 32nm Global Foundries process. I would strongly believe Llano is made to a slightly different recipe. Intel's 32nm is VERY mature, whereas the GF one is very new. If the progress Deneb made over it's lifetime, or even say from early Deneb (x4 940) to Late Thuban (1100t), then we might be in for quite the mhz scaling ride with Bulldozer. That's a 'might' of course.
Click to expand...

Good point. Process maturity should add to the scalability of Bulldozer with time, but if it came to a clock*IPC vs. clock*IPC battle today, with each manufacturer releasing new clock speeds every other month to best the other, I don't think it would be too long until AMD would exhaust their clock speed head room, especially since they are not going to release higher clocked parts until they can do so within the desired TDP envelope. (125W) (We enthusiasts overclock without regard for power usage. My Thuban, though rated 125W is probably putting out close to 200W...)

This war is already kind of starting, with Intel announcing the 2700K shortly before the anticipated Zambezi launch.
 
Kingofl337 said:
I do see the big deal of hitting 8.xx GHz, the machines didn't boot at that speed they were just changing the multiplier in the OS. If they ran at 8.xx GHz and were able to post and boot at that speed that would be one thing, but changing the multiplier until it blue screens means nothing.
Click to expand...

I don't think they are trying to convince you that every BD can get to 8 GHz with some TLC. But this stuff is done on every CPU out there and as Guinness has stated this was the first to get up that high. It implies that it should have decent potential in overclocking in normal circumstances.
 
I suppose that Asus board is also under NDA ?? I want to see that also.

Still going to be same socket layout am3+ or the FX will like the new A8-3850 cpu socket ??
 
Eisenblut said:
I guess I will be waiting for this instead of getting a 2600K, hah.

8 cores > 4 cores + HT
Click to expand...

BD cores aren't true cores but hybrid ones...so you never really know.

Kyle can you at least say if they are comparable BD vs SandyBridge?
 
Sweet thnx looked like one but wasn't quite sure, I though i notice cross-hair briefly in the demo, but with new cpu never cant tell if they will make last board version changes as well.
 
Topweasel said:
I don't think they are trying to convince you that every BD can get to 8 GHz with some TLC. But this stuff is done on every CPU out there and as Guinness has stated this was the first to get up that high. It implies that it should have decent potential in overclocking in normal circumstances.
Click to expand...

the 4.8Ghz on that mid range closed water cooling setup they had is a pretty good clue on its overclocking potential with high end air cooling/water cooling.
 
gigatexal said:
BD cores aren't true cores but hybrid ones...so you never really know.

Kyle can you at least say if they are comparable BD vs SandyBridge?
Click to expand...

At this point.... I don't think he will be able too. I am sure he wants to say something but he can't even confirm or deny having an engineering sample or when the NDA expires so I am pretty sure he can't comment on performance. :(

Well... he can comment on the Overclocking I suspect though lol.
 
sirmonkey1985 said:
the 4.8Ghz on that mid range closed water cooling setup they had is a pretty good clue on its overclocking potential with high end air cooling/water cooling.
Click to expand...

Agreed.

It also depends on how cherry picked these samples were.

They did hint at some pseudoscience being used in picking the right chips for the overclocks, so I would consider that 4.8Ghz number a best case, or at least close to it, with current production maturity.
 
Mosher said:
At this point.... I don't think he will be able too. I am sure he wants to say something but he can't even confirm or deny having an engineering sample or when the NDA expires so I am pretty sure he can't comment on performance. :(

Well... he can comment on the Overclocking I suspect though lol.
Click to expand...

hes only really allowed to comment about the overclocking at the event since it was a media event held by AMD.
 
sirmonkey1985 said:
hes only really allowed to comment about the overclocking at the event since it was a media event held by AMD.
Click to expand...

Yep. At least that's something, he doesn't have to completely not comment since they had the event then that's really fair game now.

If the thing performs decently to SB with some extra Mhz, then there is hope, and who doesn't want to cross over that 5Ghz barrier with ease :D :p
 
Zarathustra[H];1037752130 said:
Agreed.

It also depends on how cherry picked these samples were.

They did hint at some pseudoscience being used in picking the right chips for the overclocks, so I would consider that 4.8Ghz number a best case, or at least close to it, with current production maturity.
Click to expand...

true but basing it on some of the numbers the overclocks look pretty consistent, even when they showed on the chips with the writing on them the majority hit 7Ghz or higher. which is typically a good sign that the chips are consistent with their overclocks so its a good sign for the people that do traditional overclocking, may not be so much for the extreme overclockers though.


Zarathustra[H];1037752155 said:
Cool. I had heard Dan suggest that maybe there would be an N-Way comparo of 990FX boards once Zambezi hit as well.

Is this still planned?
Click to expand...

pretty sure dan said they were going to retest all the am3+ boards with BD later but i doubt they have time to do it in the launch review..
 
Zarathustra[H];1037752155 said:
Cool. I had heard Dan suggest that maybe there would be an N-Way comparo of 990FX boards once Zambezi hit as well.

Is this still planned?
Click to expand...

using Nietzsche as your forum name?
he'd roll over in his grave
 
ebduncan said:
no offense you are rather ignorant. I say this because my hexcore thuban idles barely over 1ghz would be lower if my fsb wasn't overclocked. only has a multiplier of 4. So a Stock setup would yield you 800mhz idle. Just because of the clock rate of the cpu is low doesn't mean it is power efficient. It just means its more power efficent than it is when running a full frequency. Sandy Bridge is efficient due to its die shrink and power saving features. Bulldozer will be on the same process as sandy bridge 32nm it also features advanced power saving features.

You are correct about the little guy sticking it to you as well. When the orginal FX processors were released some of those were in the 1000$ range. A company will charge you based on performance, at that time nothing matched the FX series processors in performance so there was a huge premium for them. Same as intel is doing with its hexcore cpu's.
Click to expand...

Thanks for proving my point. Due to your overclocking you sacrificed your idle speed. 1Ghz instead of 800Mhz. In your case that isn't much of a power difference but a 5.8Ghz idle system will use a lot more power than if the same system were idling at a low speed. Many different factors contribute to the power consumption of a processor. Processor speed, design (architecture, manufacturing process and stepping), voltage and load are the major factors. Processor design is locked in once you buy the system. You do have control over the processor speed and voltage as a user. Many of the existing systems require users to sacrifice idle power for max OC performance. SB overclocking allows for high overclocks when under load yet retains stock idle speeds and power consumption. In my case it goes from 1.6Ghz low power to 4.9Ghz under high power. That is a large dynamic range of power usage and performance. BD didn't show this behavior in the video.

BD will be built on a 32nm process like SB but not the same process. A die shrink alone does not mean a processor design will use less power. AMD was able to show how BD clocks very high but that is it. None of the systems were under heavy load and no measurable work was done in the demo. This would have been a great stunt back in the P4 days but everyone knows Ghz alone won't cut it.

I wasn't referring to AMD's ability to release $1000 processors because of their position. Currently they have processors listing over $1200. What they did was raise the price on already released products. What would you say if Apple raised the prices on their products 3 months after they were released?

Ignorance is bliss. Have a good day.
 
Zarathustra[H];1037752116 said:
800px-80386with387.JPG


:
Click to expand...

I miss that style of logos. The i386 logo looks like a developer for an Atari game. I hope that 80s graphics/processor marketing/overall look and feel will one day be promoted into a genre similar to steampunk.
 
Zarathustra[H];1037752155 said:
Cool. I had heard Dan suggest that maybe there would be an N-Way comparo of 990FX boards once Zambezi hit as well.

Is this still planned?
Click to expand...

We are going to go back and revisit some boards in an OC roundup.
 
Zarathustra[H];1037752116 said:
Hah, good luck getting any kind of performance answer out of anyone with legit access to parts until the NDA lifts.

Kyle has probably already stretched the limits a little with what he is allowed to say with this comment:
Click to expand...


That is the problem with all of this speculation and rumor mill crap that has been going on for many months now. NOONE known to be a reputable member of the community is ever out there leaking stuff. Charlie every now and then hits the nail on the head, but he is no longer considered reputable either. Sure Kyle might get blacklisted and lose a ton of advertising revenue, but all of this tin foil crowd discussions wouldn't be happening if he or Anand or someone like that leaked some stuff. Outrageous speculating is why I havn't stepped into the AMD processor section in a while as it just gets old and boring at times. I think we just want some solid numbers after many years of waiting for this processor.

As for me, I just want it to be competitive since I want an option this holiday season when the ole PII 965 gets upgraded regardless of whether the Zambezi 8 cores are available in volume or not.
 
Zarathustra[H];1037752116 said:
Well, they are true cores if you think of them as cores in the traditional sense, namely Integer cores.

It wasn't until fairly late in the CPU world that FPU units were built into the main CPU chip.
800px-80386with387.JPG


Again, I've gone into more detail than I really needed to (I tend to do that) but essentially if you think of it from the traditional CPU terminology, a CPU was an integer core and some L1 cache only. No memory controller, no L2 or L3 caches, and no FPU.

In this sense, it is a true 8 core (or CPU) design. We've just come to expect more out of our CPU's over the years, like FPU's, caches and memory controllers.
Click to expand...

It's a telling illustration of just how long ago it was that we thought that way, that those chips are running naked, with no coolers at all. So, while true, its kind of like arguing that "video processors" only require 2D and that all the 3D performance we look for doesn't really matter, because they used to not do 3D at all.

Not saying I disagree with your assessment (which is that for practical purposes it is a true core with "morphing" capabilities) just pointing out that in these cases, perception is reality, and the perception is that the BD core is "less" than a SB core.
 
gigatexal said:
Kyle can you at least say if they are comparable BD vs SandyBridge?
Click to expand...

The closer we're getting the launch, the more I'm getting the sense that AMD never intended for a BD core to have more than 50% overall perf of a Sandy core. That instead AMD is hoping each pair of cores (core units) will come close enough to a single Sandy core, and price accordingly. I would hope to be proven wrong, but when chew* over at XtremeSystems (the guy who did this 8.4Ghz BD overclock for AMD) has hinted as much over the past few weeks, I'm not expecting any surprises.

The only thing you need to keep in mind when review hit, is that Intel Hyper-Threading on a 4-core processor has the performance of ~3.8 to 5 threads on the CPU without HT, not 8 threads... AMD with it's 8-cores/4CU should be better perf/thread with high thread counts, but worse perf/core. All I know is Bulldozer with it's 8 'cores' but only 4 'core units' is no doubt going to be a tricky CPU to review and draw solid conclusions on. How does 4 & 8-threads on a 2500K + 2600K compare to 4 & 8-threads on a FX8150? Is Bulldozer only competitive when maxed out with 8-threads? I think we can count on Kyle and [H] coming through with a solid review and not succumbing to marketing BS, but if not the initial review, I hope we get a follow-up that compares thread scaling (1, 2, 4, 6, 8) on an 8-core BD vs 4-core+HT Sandy to get a true sense of what were dealing with performance wise.
 
Is Bulldozer only competitive when maxed out with 8-threads?
Click to expand...

I expect the 8150 to be competitive with the i7 2600 in less than 6 threads with its aggressive turbo. In 6+ threads I expect the BD to win most benchmarks. Overclocking is more complicated. AMD should still win the 6+ threads provided it can reach the around same frequency as the i7 2600K. In less than 6 threads I expect BD to need more frequency than the i7 to achieve the same performance.
 
ebduncan said:
IPC doesn't mean anything. It stands for instructions per clock, it doesn't mean the cpu will be faster you also have to factor things like l2, l3 and frequency .

Honestly i would care less if the IPC was 4 times better, than one part, the other part could have 4 times the clock speed and it would be just as fast.

Point is Bulldozer is going to have a 400mhz clock speed advantage over Intel. So it can suffer a bit by having a 10% slower ipc. It makes up the difference in clock speed. Of coarse you will just blab about how you can overclock. Truth is 95% of users don't even know how to overclock.
Click to expand...

IPC is very important. IPC was why a 2.2Ghz AMD Barton could process just as much as 3.2GHz Northwood P4. I have no idea how come people nowadays all of a sudden say IPC doesn't matter when estimating a processors performance.
 
IPC is very important. IPC was why a 2.2Ghz AMD Barton could process just as much as 3.2GHz Northwood P4. I have no idea how come people nowadays all of a sudden say IPC doesn't matter when estimating a processors performance.
Click to expand...

In the second line he was basically saying that (IPC x frequency) is what is important so that if AMD could achieve less IPC with more frequency that would be the same. I have no argument with that.
 
drescherjm said:
I expect the 8150 to be competitive with the i7 2600 in less than 6 threads with its aggressive turbo. In 6+ threads I expect the BD to win most benchmarks. Overclocking is more complicated. AMD should still win the 6+ threads provided it can reach the around same frequency as the i7 2600K. In less than 6 threads I expect BD to need more frequency than the i7 to achieve the same performance.
Click to expand...

Those were originally my exact thoughts a couple months ago as well, but now I'm beginning to have doubts, and that 8-core BD with 8-threads will be slightly worse than 4-core HT Sandy with 8-threads... That AMD's goal all along was to be competitive against Sandy Bridge Hyper-Threading, not Sandy Bridge cores... Price/perf could potentially make up for it, but that's exactly why weeding through BD reviews is going to be so tricky...
 
Last edited:
drescherjm said:
In the second line he was basically saying that (IPC x frequency) is what is important so that if AMD could achieve less IPC with more frequency that would be the same. I have no argument with that.
Click to expand...

While this is true and would make them initially competitive, the point of concern (at least for AMD) is that you can only raise the clock speed so high as the processes are refined. If Intel holds a sizable lead in IPC over bulldozer, they will hold a solid advantage over AMD as the competing chip series mature.
 
You must log in or register to reply here.
Back
Top