Bulldozer "delayed" until September 2011 (Rumor)

[OCGuy]

Now that the delay is confirmed (?) what do you guys suggest for an almost complete build that was holding out for June 20th release?

1) Z68 + Sandy bridge (ex Asus P8Z68-V Pro + 2500k)
2) 990 + budget AMD chip (with the plan to go BD in the fall)

Other tidbits:

• will go with a 580 to drive games at 2560x1600 and other... tasks
• coming from an E6850 @3.6 with an HD4860, yea, it's ancient.

Buying a cheaper setup now just to upgrade in a few months to a platform with unknown performance? No thanks.

 

[RyanGreener]

Now that the delay is confirmed (?) what do you guys suggest for an almost complete build that was holding out for June 20th release?

1) Z68 + Sandy bridge (ex Asus P8Z68-V Pro + 2500k)
2) 990 + budget AMD chip (with the plan to go BD in the fall)

Other tidbits:

• will go with a 580 to drive games at 2560x1600 and other... tasks
• coming from an E6850 @3.6 with an HD4860, yea, it's ancient.

It's very tough to recommend something because we know nothing about Bulldozer still....so if we say to go 990 + budget AMD chip and Bulldozer is a flop, then that won't be too good for you. At the same time, there is a chance that Bulldozer can be huge, and since you're coming from an E6850 which is a C2D, a safe recommendation would to try and get a cheap Phenom II X4/X6 with a 990 motherboard. Phenom II's provide somewhat comparable performance to the C2Q series so you'll notice an upgrade while at the same time, you'll have upgradeability for the future. But this is just my opinion...

 

[drizek]

I'd get a dual core and try and unlock it. Nothing to lose.

 

[nonameo]

Now that the delay is confirmed (?) what do you guys suggest for an almost complete build that was holding out for June 20th release?

1) Z68 + Sandy bridge (ex Asus P8Z68-V Pro + 2500k)
2) 990 + budget AMD chip (with the plan to go BD in the fall)

Other tidbits:

• will go with a 580 to drive games at 2560x1600 and other... tasks
• coming from an E6850 @3.6 with an HD4860, yea, it's ancient.

*sigh* I think everyone here wants bulldozer to perform well, but I just don't see it being leaps and bounds faster than SB in anything except possibly the highly multithreaded stuff. The Z68 chipset has lots of nice features... you might spend extra money on an intel setup, but it will surely run what you want it to for at least a few years. Not only that, intel stuff tends to hold its value better. Just look at the prices that 775 quads are going for. For people who have the bug, just get go ahead and build sandy bridge. Although, I can't really say there's any harm in waiting. Supposing AMD hits the mark with bulldozer, it will almost certainly be a better feature/value proposition.

 

[gramboh]

Now that the delay is confirmed (?) what do you guys suggest for an almost complete build that was holding out for June 20th release?

1) Z68 + Sandy bridge (ex Asus P8Z68-V Pro + 2500k)
2) 990 + budget AMD chip (with the plan to go BD in the fall)

Other tidbits:

• will go with a 580 to drive games at 2560x1600 and other... tasks
• coming from an E6850 @3.6 with an HD4860, yea, it's ancient.

If your main purpose is gaming, get the SB and overclock it. SB at 4+GHz driving a single GPU at 2560x1600 will be more or less GPU limited anyway. Even if BD somehow blows away SB (unlikely), it won't give you higher FPS in your specific situation. Plus, you will waste months of gaming time on an underpowered setup, why cripple a top of the line card (580) while it's still top of the line?

 

[LucJoe]

If your main purpose is gaming, get the SB and overclock it. SB at 4+GHz driving a single GPU at 2560x1600 will be more or less GPU limited anyway. Even if BD somehow blows away SB (unlikely), it won't give you higher FPS in your specific situation. Plus, you will waste months of gaming time on an underpowered setup, why cripple a top of the line card (580) while it's still top of the line?

lol i'm not even using the 580 yet because it doesn't fit in my case and i don't want to swap all the old stuff over to the new case.... what a waste

Anyway, I agree with all the responses. Seems like a 2500k asap is the way to go; too bad, really was trying to hold out for that supposed June 20th launch

 

[Skott]

Now that the delay is confirmed (?) what do you guys suggest for an almost complete build that was holding out for June 20th release?

1) Z68 + Sandy bridge (ex Asus P8Z68-V Pro + 2500k)
2) 990 + budget AMD chip (with the plan to go BD in the fall)

Other tidbits:

• will go with a 580 to drive games at 2560x1600 and other... tasks
• coming from an E6850 @3.6 with an HD4860, yea, it's ancient.

You basically have three choices:

1. Do nothing and wait for BD to release later in August or whenever it is. Or even wait for IB next year if you want Intel and can wait.

2. If you want to build AMD but cannot wait then get an AM3+ mobo and a Phenom II chip now and then when BD releases upgrade to a BD chip once they release. Unknown if you need a bios update. Depends on mobo.

3. If you want Intel now buy a Intel SB chip and a Z68/P67 mobo and when IB releases Q1 2012 update bios and buy IB chip.

So there are your choices choose which you want to go with.

I'm of the opinion BD will not be able to match or beat Intel on the high end spectrum so they'll have to go for cheaper chips and play the low to middle spectrum like they currently do. In short Intel will stay king of high performance chips just as they are now.

Just my own opinion. I'd love to see BD smack SB around and trounce it but I seriously doubt it can. That doesn't mean a BD rig won't be good enough for gaming though. It will.

 

[RussianSensation]

Anyway, I agree with all the responses. Seems like a 2500k asap is the way to go; too bad, really was trying to hold out for that supposed June 20th launch

But if you buy a Phenom system today and then decide to upgrade to BD later, that means you will have wasted at least $100 on the Phenom chip. Even if you consider reselling it, why create that unnecessary hassle? 2500k @ 4.4ghz+ should be fast enough to max out your GTX580 in most scenarios. Where BD is likely to be more competitive are multi-threaded scenarios such as rendering, encoding, etc. But gaming wise, at 2560x1600 you are almost entirely GPU limited (as has been mentioned).

lol, I seriously doubt it. Have you seen the insane claims these companies make?

While the 335% claim is probably ridiculous, the fact is Asrock xFast does have the fastest USB 3.0 copy speeds out of all the current implementations. Read and write speeds are also class leading.

 

[dx11101]

hmmm, amd finally figured out reverse hyperthreading? 16 physical cores registering as 8 threads to windows?

 

[Tuna-Fish]

hmmm, amd finally figured out reverse hyperthreading? 16 physical cores registering as 8 threads to windows?

Um wtf what?

The very idea of reverse hyperthreading is insane. Anyone who mentions it just doesn't know much at all about programming or processors.

 

[Arachnotronic]

Anyway, I agree with all the responses. Seems like a 2500k asap is the way to go; too bad, really was trying to hold out for that supposed June 20th launch

BD is not due out for another 60-90 days.

 

[RyanGreener]

BD is not due out for another 60-90 days.

Which is why he said it was too bad that it didn't come out on June 20th

 

[formulav8]

Asrock xFast does have the fastest USB 3.0 copy speeds [/URL]out of all the current implementations. Read and write speeds are also class leading.

I've used that xfast utility and it really makes a difference in both read and writes on the usb 2.0 drives I tested. Do you know what the utiity actually does? Different Caching scheme or something? I remember searching awhile back but didn't find anything concrete.

 

[Arachnotronic]

Um wtf what?

The very idea of reverse hyperthreading is insane. Anyone who mentions it just doesn't know much at all about programming or processors.

*Although*, if you think about it, two FMACs working together to produce one 256 bit AVX operation...

 

[dx11101]

i knew if i typed "reverse hyperthreading" i would piss off at least 1 .

Yeah i know its not that but i did read that they are somehow sharing L2 cache between cores and that the 8 core bulldozer actually only has 4 modules, but it registers as 8. Pretty weird....

 

[Voo]

i knew if i typed "reverse hyperthreading" i would piss off at least 1 .

Yeah i know its not that but i did read that they are somehow sharing L2 cache between cores and that the 8 core bulldozer actually only has 4 modules, but it registers as 8. Pretty weird....

Wait, you're not a novelity troll account? Surprising. Usually uninformed posts from users with <10 posts are..

Anyways - the first research papers about SMT are decades old and Tullsen's that pretty much inspired intel's approach are from the 90s

 

[Topweasel]

i knew if i typed "reverse hyperthreading" i would piss off at least 1 .

Yeah i know its not that but i did read that they are somehow sharing L2 cache between cores and that the 8 core bulldozer actually only has 4 modules, but it registers as 8. Pretty weird....

Basically they include 2x of all of the normal portions of a CPU and integrate them into one module and then remove the secondary pieces that don't have a use (things that 2 of them don't have an affect on performance). Then instead of doubling the pathway for the FPU to support rare 256-bit code, they continued to use 128bit FPU's that within a module can work together to handle this code.

 

[Idontcare]

Anaphase?

 

[RobertPters77]

My main concern with BD will be performance per clock.

I don't expect it to beat sandy, I expect it to come close.

 

[Vesku]

My main concern with BD will be performance per clock.

I don't expect it to beat sandy, I expect it to come close.

That's a lot to ask when they've deepened the pipeline. I'm still expecting closer to i5-750 IPC, is that what you meant by close?

 

[Nemesis 1]

Anaphase?

http://www.computer.org/portal/web/csdl/doi/10.1109/TPDS.2007.70797


The proposed approach is based on predicting/computing thread input values via software, through a piece of code that is added at the beginning of each thread (the pre-computation slice). A pre-computation slice is expected to compute the correct thread input values most of the time, but not necessarily always.

This code would be like onto the prefix of VEX

 

[Nemesis 1]

Basically they include 2x of all of the normal portions of a CPU and integrate them into one module and then remove the secondary pieces that don't have a use (things that 2 of them don't have an affect on performance). Then instead of doubling the pathway for the FPU to support rare 256-bit code, they continued to use 128bit FPU's that within a module can work together to handle this code.

It wasn't me that you were expecting I been waiting on someone to mention it . Than IDC gave me the perfect in.

 

[Khato]

So the seemingly conflicting reports from Anandtech and Xbitlabs regarding Bulldozer B0/B1 stepping frequency suddenly make far more sense after the new information revealed today. The Anandtech overview of AMD's presentation of Bulldozer at Hot Chips 2010 as well as all other information I'd seen stated that frequency would be controlled at the module level... But the OverDrive screen shots from E3 today show core specific frequencies, which would require that the integer cores have clock crossings on their interfaces to all shared logic. So it could well be that the integer cores are easily hitting 3.8 GHz as was mentioned in the one Anandtech article, while other logic in the module couldn't clock up to competitive frequencies.

It certainly would be an interesting design decision... Though there's of course a chance that it's just another bug in the software (like temperature) and there aren't different clock domains for each portion of logic in a module.

 

[RussianSensation]

I've used that xfast utility and it really makes a difference in both read and writes on the usb 2.0 drives I tested. Do you know what the utiity actually does? Different Caching scheme or something? I remember searching awhile back but didn't find anything concrete.

I am not sure how it works exactly. But this video clearly shows that it's faster in the real world

http://www.youtube.com/watch?v=Pq6NBNHs7zY&feature=player_embedded

 

[RobertPters77]

That's a lot to ask when they've deepened the pipeline. I'm still expecting closer to i5-750 IPC, is that what you meant by close?

I mean within 10% of each other. Like Core 2 and Phenom 2.