Fudzilla: Bulldozer performance figures are in

[Arzachel]

Super Pi accurately measures single core preformance.

It might measure something, but it's definitely not real world single core performance. Well, unless you're still living in the last decade I guess.

 

[sm625]

Sure it does. Take any two processors that have the same number of cores and cache. One has a superpi score double that of the other one. It is going to feel much faster in most applications. This is probably more true than for any other benchmark. Just name an instance where this isnt true.

 

[Arg Clin]

Looks good if this is true - but like so many others I don't take leaked benchmarks into consideration, since they hold too many unknowns. The extreme variation in leaked benchmarks on supposed BD performance proves this to the point.

I will however maintain my original prediction; BD will be competitative at least against Intels mid-end offering (SB i5 2500K). If BD wasn't AMD would not need to bother bringing it to the market, since they already got the low end covered with Llano.

If it in fact turns out to outperform SB i7 2600K -this would be really good news, as I strongly feel the CPU market needs some real competition again. Since the release of SB, AMD had not had a countermove, which means the K10.5 arch must have more or less run out of steam.

A well performing BD would probably mean a faster release of IB as a typical Intel strategy in a time when they are comfortably ahead is to keep the good stuff in the backhand, and try to push some older gen tech out first. (the less upgrade steps people can jump, the better for the cpu makers) So even the "strictly Intel folks" (for whatever reasons I never will understand) should hope for this

 

[yottabit]

Sure it does. Take any two processors that have the same number of cores and cache. One has a superpi score double that of the other one. It is going to feel much faster in most applications. This is probably more true than for any other benchmark. Just name an instance where this isnt true.

A while back I made a video comparing PPro, PIII, A64, Core 2, Core i5 in superPi
http://www.youtube.com/watch?v=dvibGuRe9E0

Honestly, truthfully, I expect Bulldozer to be only a so-so gaming processor. I think it will be better than Phenom II, but nowhere in Sandy Bridge territory. It was always designed as a many thread CPU which != gaming performance and I don't get why so many people think it will be a great gaming CPU. However I think it will have adequate gaming performance for those who want to game, and will rock multithreaded apps.

 

[podspi]

A while back I made a video comparing PPro, PIII, A64, Core 2, Core i5 in superPi
http://www.youtube.com/watch?v=dvibGuRe9E0

Honestly, truthfully, I expect Bulldozer to be only a so-so gaming processor. I think it will be better than Phenom II, but nowhere in Sandy Bridge territory. It was always designed as a many thread CPU which != gaming performance and I don't get why so many people think it will be a great gaming CPU. However I think it will have adequate gaming performance for those who want to game, and will rock multithreaded apps.

TurboCORE. Many games today use ~ 4 threads at the most. On a 2500K/2600K, turbo is limited because each core is engaged. On an FX-8xxx half the CPU can still be gated off and that TDP can be transferred to the remaining four cores.


Whether this pans out is still under debate, of course.

 

[formulav8]

That is the first thing that came to my mind.

I had to crack up at that

 

[soccerballtux]

Sure it does. Take any two processors that have the same number of cores and cache. One has a superpi score double that of the other one. It is going to feel much faster in most applications. This is probably more true than for any other benchmark. Just name an instance where this isnt true.

it doesn't provide an accurate reflection of performance in branch heavy code. Mathematical computations are ultra easy to accurately branch predict.

 

[OCGuy]

TurboCORE. Many games today use ~ 4 threads at the most. On a 2500K/2600K, turbo is limited because each core is engaged. On an FX-8xxx half the CPU can still be gated off and that TDP can be transferred to the remaining four cores.


Whether this pans out is still under debate, of course.

Who needs turbo when you can just OC the crap out of the Intel chip?

 

[Ares1214]

Who needs turbo when you can just OC the crap out of the Intel chip?

You can do the same out of both lol

 

[bridito]

Has anyone posted this yet? (if they have, I apologize for the doubling)

http://www.legitreviews.com/news/11064/

The FX-8130P is rumored to have a base core clock speed of 3.8Ghz, but the processor has a turbo mode that allows it to go up to 4.2Ghz for certain work loads. The FX-8130P is priced at the site for S$600. 600 Singapore dollars would equal $489.48 US dollars, so we hope that this individual is marking it up a bunch. The rumored tray price of the AMD FX8130P in quantities of 1000 is $320! The same seller has the FX-8110 listed for sale at S$500 or $407.90 USD.

Sounds like the pricing is NOT right on target if accurate.

 

[yottabit]

Has anyone posted this yet? (if they have, I apologize for the doubling)

http://www.legitreviews.com/news/11064/

The FX-8130P is rumored to have a base core clock speed of 3.8Ghz, but the processor has a turbo mode that allows it to go up to 4.2Ghz for certain work loads. The FX-8130P is priced at the site for S$600. 600 Singapore dollars would equal $489.48 US dollars, so we hope that this individual is marking it up a bunch. The rumored tray price of the AMD FX8130P in quantities of 1000 is $320! The same seller has the FX-8110 listed for sale at S$500 or $407.90 USD.

Sounds like the pricing is NOT right on target if accurate.

I think that's completely irrelevant IMO, on such an over-hyped CPU. Plus it's a preorder, so I'm sure if you want to be the very first to have in your hands a hyped up product you'll pay way over MSRP. Plus, computer component pricing varies substantially based on the market.

Remember when PS3s and Xbox 360s were on ebay for $1k+?

 

[Arkaign]

^^ International pricing is legendarily incomparable to US pricing, as we've seen when trying to help people from various countries around the world with their builds going by their own vendors. Something $250 here might be selling for $400+ say in India or Poland or whatever. And sometimes the other way around, although much less often.

Considering the amount of time left for the launch of this thing, I think it's just more unsubstantiated BS, or possible pricing for the failed original launch date.

 

[OCGuy]

You can do the same out of both lol

You have headroom numbers for how high BD can OC all cores stable on acceptable voltage?

I would love to see this source.

 

[Schmide]

Anybody else thinking it does so poorly in SuperPi because AMD said "Screw x87" seeing as how its dead? Might explain why it does so well in everything but SuperPi if these benchmarks are to be believed. Maybe its not poor single threaded performance.

I would think it's more the slow L2 screwing the calculation, not the x87 FPU. Despite the fact that most people see a calculation of PI as a purely mathematical operation, (yes it's math but...) it actually involves large numbers that overflow the general FPU registers and are encoded as BCDs (binary encoded decimals) or some hybrid of BCD. Although many of the new algorithms are self correcting conveniences, they still involve many variables that will overflow the L1, L2, L3 and most likely reach main memory depending on precision. Just run it (1M) and watch your CPU, it will barely jump a few % points during the pass.

I haven't studied the Jonathan Borwein and Peter Borwein's SuperPi algorithm, but I can't imagine they have some silver bullet to avoid the above situation.

 

[Arzachel]

Sure it does. Take any two processors that have the same number of cores and cache. One has a superpi score double that of the other one. It is going to feel much faster in most applications. This is probably more true than for any other benchmark. Just name an instance where this isnt true.

Well, if you remove "accurately" from your statement, then surely any program that isn't threaded works! If the number is irrevalent, why use superpi?

 

[bridito]

Arkaign & yottabit: Heck, I remember when 1MB RAM sticks were over a grand each. That's how long I've been around! IMHO, If they're really trying to move FX-8130Ps at $489.48, they had better beat SB-E Quads soundly across the board! So do you think that the long rumored $320 price is going to be close to the actual MSRP?

 

[podspi]

You have headroom numbers for how high BD can OC all cores stable on acceptable voltage?

I would love to see this source.

Without knowing those facts, neither of you can prove your point

If BD can't even eek out an additional 100mhz, then yea perhaps its Turbo doesn't matter much. If BD can easily hit 5ghz+ then that changes things completely.

We just don't know yet. I personally believe BD is going to have a clock advantage wrt SB. If it doesn't, 2500K will keep the sweet spot gaming crown for now.

 

[georgec84]

Intel supporters not impressed, AMD supporters excited. Big surprise there...

Reported benchmark - fake or not fake?

 

[yottabit]

Arkaign & yottabit: Heck, I remember when 1MB RAM sticks were over a grand each. That's how long I've been around! IMHO, If they're really trying to move FX-8130Ps at $489.48, they had better beat SB-E Quads soundly across the board! So do you think that the long rumored $320 price is going to be close to the actual MSRP?

Hah, that's a bit before my time Personally I think the rumored pricing is close to correct, but I of course have nothing to back that up. It makes sense that BD would be positioned against the 2600k to me given what we've seen so far.

 

[Arkaign]

Arkaign & yottabit: Heck, I remember when 1MB RAM sticks were over a grand each. That's how long I've been around! IMHO, If they're really trying to move FX-8130Ps at $489.48, they had better beat SB-E Quads soundly across the board! So do you think that the long rumored $320 price is going to be close to the actual MSRP?

At this point we have no idea, and unfortunately foreign rumored reserve pricing is of little use. I don't know if $320 is real or just more FUD, or was an intended price for a particular SKU but it's all gone out the window due to delays or what.

At this point we don't know jack honestly.

 

[Arkaign]

Intel supporters not impressed, AMD supporters excited. Big surprise there...

Reported benchmark - fake or not fake?

I'm not sure. I'm not a supporter of either camp, but from an independent perspective it doesn't seem legit to me.

Usually if things are going fairly well for a product, you get some leaks or previews from the mfg themselves that are pretty clear, as was the case with Conroe, Phenom II, etc. When things are going poorly, eg Prescott, Phenom I, Fermi, 2900, etc, you usually get very little legit info but sometimes some rumors turn out to be true as regards to problems.

It would make a lot of sense if AMD had a killer product on the way to put some legit benches out there from more credible sources to help slow sales on the Intel products and excite their shareholders and potential shareholders. They are a corporation, their sole purpose is to make $ for their shareholders, so it would be logical imho to promote as heavily as possible a winning product coming down the pipe, and to obfuscate info relating to poor/mediocre products.

 

[Ares1214]

You have headroom numbers for how high BD can OC all cores stable on acceptable voltage?

I would love to see this source.

Do you have a source saying it cant? If we are to believe these sources, then OB got his up to 5.1GHz at i think 1.57ish maybe 1.55ish volts. Thats if we are to believe, which it seems this is a rumour thread, so we might as well just pretend all these rumors are true in this context. Seeing as how its an AMD CPU, id say around 1.6v is the max safe voltage. This was also an ES, so that may or may not affect it, but obviously a production model might push a bit more.

 

[Gigantopithecus]

It would make a lot of sense if AMD had a killer product on the way to put some legit benches out there from more credible sources to help slow sales on the Intel products and excite their shareholders and potential shareholders. They are a corporation, their sole purpose is to make $ for their shareholders, so it would be logical imho to promote as heavily as possible a winning product coming down the pipe, and to obfuscate info relating to poor/mediocre products.

I don't disagree but there are reasons to conceal info about a good product. Namely, PC component sales - at least for consumers, I do not know about corporate customers - are typically terrible during the summer. There's no reason to slow sales on Intel products when those sales are already about as slow as they're going to get this year.

Further, AMD knows the economy is terrible right now. Neither people nor businesses are buying anything, let alone computers. Consumers will inevitably open their wallets for the back to school blitz and then again for the holidays, so I hope for AMD's sake that they have clear numbers available by the middle of August at least. It would be far more effective for a fall/winter sales blitz to get numbers out then, make a splash, and be fresh in consumer's minds. 99% of people who are going to buy computers later this year are not following the developing Bulldozer story every day.

 

[Arzachel]

Usually if things are going fairly well for a product, you get some leaks or previews from the mfg themselves that are pretty clear, as was the case with Conroe, Phenom II, etc. When things are going poorly, eg Prescott, Phenom I, Fermi, 2900, etc, you usually get very little legit info but sometimes some rumors turn out to be true as regards to problems.

Counterpoint: Cayman. That turned out fairly competitive for something they had to slap together at the last minute due to the canceled 32nm node.

Do you have a source saying it cant? If we are to believe these sources, then OB got his up to 5.1GHz at i think 1.57ish maybe 1.55ish volts. Thats if we are to believe, which it seems this is a rumour thread, so we might as well just pretend all these rumors are true in this context. Seeing as how its an AMD CPU, id say around 1.6v is the max safe voltage. This was also an ES, so that may or may not affect it, but obviously a production model might push a bit more.

Keep in mind that iirc, OBR "overclocked" an A8 to 4.6ghz through multiplier. The problem is, Llano limts the multiplier to stock, so that number only showed up on cpu-z.

 

[podspi]

Llano is more for BTS shoppers anyway, though. Cheap, energy efficient laptops that can game? They're going to fly off the shelves. I expect BD will mostly be for DIY enthusiasts and servers until Trinity.