Fudzilla: Bulldozer performance figures are in

[grkM3]

If they wont show a screenshot of AIDA64 results then they are worthless. By now everyone should know the ES scores zero on AIDA64 L2 and L3 write. As long as that is true I consider these parts extremely crippled. And if someone does not post results then they are deliberately contributing FUD.

I found this online somewhere,I hope its true for BD

 

[Abwx]

Source of that quote?

edit - this?

Seems to be the good one..
thanks, i didn t have it myself...

 

[sm625]

Re #851... how do we know that isnt a screenshot of a 5.0GHz 2600k. That is exactly what it looks like to me.

 

[grkM3]

Re #851... how do we know that isnt a screenshot of a 5.0GHz 2600k. That is exactly what it looks like to me.

I was just trying to give you a adia64 run,wow you nailed it right on the head,thats a 2600k at 5ghz lol

 

[Vesku]

Q4 starts Oct 1.

Yes and that Sept 1st demo doesn't seem to be much more involved than the gaming demo at HardOCP, judging by their own description of the upcoming event. Seems that at best they will get it out the door 1-2 weeks before Oct. 1st.

 

[AnandThenMan]

Super PI

. I can't believe that ancient thing is still being used.

 

[RussianSensation]

Super PI

. I can't believe that ancient thing is still being used.

While no one benchmark is end-all-be-all, read my post #787.

SuperPi has been extremely useful in predicting single threaded performance per core and coincidentally gaming performance which relies on integer computation (i.e., superPi!)

 

[Tuna-Fish]

SuperPi has been extremely useful in predicting single threaded performance per core and coincidentally gaming performance which relies on integer computation (i.e., superPi!)

SuperPi isn't integer. The computation is all x87 floats. However, as modern computers have vastly more FP horsepower than the ones that SuperPi was optimized for, it's performance mostly depends on how good caches you have.

As caches are among the most expensive parts of the chip, measuring just the caches is a passable heuristic for total speed -- as there would be no point in having a chip with better caches than it's execution parts need.

 

[Abwx]

SuperPi isn't integer. The computation is all x87 floats. .

Unlikely.
There s scores of FPs for sure , but there is likely way higher
number of integer computation since it use some kind of limited
develloppement built upon elliptic integrals identities.

 

[RussianSensation]

SuperPi isn't integer. The computation is all x87 floats. However, as modern computers have vastly more FP horsepower than the ones that SuperPi was optimized for, it's performance mostly depends on how good caches you have.

As caches are among the most expensive parts of the chip, measuring just the caches is a passable heuristic for total speed -- as there would be no point in having a chip with better caches than it's execution parts need.

Setting integer vs. floating aside, SuperPi is a great benchmark to measure single core / cache performance as you said. In fact, it was very widely used even 5 years ago. Back then A64/X2 were faster than Intel offerings and ironically not many people complained about using it then.

 

[AnandThenMan]

Unlikely.

AFAIK SuperPi uses x87 instructions, think 80486 optimized. ^_^

 

[Schmide]

AFAIK SuperPi uses x87 instructions, think 80486 optimized. ^_^

Yes it uses FP instructions but it's dealing with numbers way way bigger than any register can hold. The actual FP calculations are not really that taxing, it's more the accumulation of all the stray digits that causes the bottleneck. As Abwx was eluding to, it probably takes a bit of regular instructions to do so, i.e. integer.

To put a fine point on it, it's optimized for very large memory access.

 

[Michael Meio]

I have the feeling that AMD struggles to surface a new line. Many people think that some Bulldozer will climb up some hill and strike a foot on INTEL's neck. I find this kinda funny.

I'm a true believer on the potential of AMD's architecture but for some reason, I don't have faith in it's ability to retain any title simply because one can't ignore each company's lineup.

We all expect the degree of performance to be in direct proportion with the price segment. That's why people talk so much about the USD$320 neighborhood. I think that if they really want to put the gloves on, they should expose a chip that outperforms a $320 at $250. THEN we could talk about real competition in this particular economic era. The rest are simple numbering and name calling games.

I truly wish to consider the Bulldozer line as AMD's ace in the sleeve as many naive enthusiasts do but the truth is that INTEL has aces, jokers, swords, cavalry, alices and jabawookies... I bet they even have some Tony Montana chip somewhere that we have no clue about.

INTEL's strategy is grotesque, to put it in PG13 words. They deplete a lineup as far as AMD can float their zodiac. Then, they just release whatever can cope with that plus some useless gimmick which enthusiasts can't live without and the wheel goes around.

 

[86waterpumper]

I have always liked amd, it's hard not to want to root for the underdog. As plenty have pointed out we need them to at least stay in the game. I fondly remember the k6 days, I loved especially my k6III+ systems @ 112 x 5 Overclocking used to be alot more fun when you actually had to manipulate jumpers and work at it some. Newbies would freak out nowdays if their rigs failed to even boot after a failed o/c haha.

I will have to say though I am very impressed with my new 2500k system even at stock speeds. I didn't like the rough launch of the sandy bridge setups, especially the early chipsets, but with z68 systems it's hard to complain. I know this is the cpu forum, but to me the motherboards and chipsets is where amd really falls behind. Bulldozer may be a fast cpu, but unless there is a MUCH improved ram and sata controller onboard, things are not going to look too bright. We are quickly reaching the point where power consumption matters alot too. If I decide to get rid of my sandy bridge system, I can easily move it into a htpc role to replace my x2 amd rig I am running now. Low power and idle consumption is a area where intel seems to have the bull by the horns, but hopefully LLano is a step in the right direction on that front.

I think that having tons of cores is a cool concept but it's not really something most people care alot about. Sure there are people that run servers or render farms or whatever that can use gobs of cores, but I care more about single core performance and the ability to build smaller, quiet systems that still have alot of power.

 

[AnandThenMan]

I have the feeling that AMD struggles to surface a new line. Many people think that some Bulldozer will climb up some hill and strike a foot on INTEL's neck. I find this kinda funny.

I frequent several hardware forums, and the overwhelming sentiment is Bulldozer is going to disappoint, I rarely see people claiming it is going to outright take the performance crown and all that.

 

[videoclone]

It will be fine, it wont be great but it wont be bad... lol pretty much sums up what most people are saying!

 

[Arg Clin]

I frequent several hardware forums, and the overwhelming sentiment is Bulldozer is going to disappoint, I rarely see people claiming it is going to outright take the performance crown and all that.

Claims around hardware forums based on more or less qualified speculations. Hardly holds much value IMO. Let's wait for some hard numbers to come out - (and I don't mean obscure leaked "maybe-genuine" ES benchmarks) - then pass judgement on AMD and Bulldozer.

Here's my prediction, which is based on pure speculation on my part based on simple common sense in a marketdriven siatuation: BD will outperform at least Intels mid-end offerings (SB i5), and possibly go higher. Intel will respond ASAP with higher clocked SB's. (SB seems to have the potential)

Who takes the performance crown at the end of the day? - odds are in favour of Intel, because of their massive R&D ressources. What's more interesting though is the $200-$300 segment - I think BD will fint nicely in the many years tradition on AMD's part to offer competitative mid-end cpus.

 

[AnandThenMan]

Claims around hardware forums based on more or less qualified speculations. Hardly holds much value IMO.

I agree, I'd take it a step further and say the speculations hold almost no value. We know quite a lot about Bulldozer as an architecture, but there are so many unknowns that predicting actual performance is really difficult. The thing is, a pattern we oft see is when a product is late and much secrecy surrounds it, it usually ends up falling short. Not a hard and fast rule of course..

Let's wait for some hard numbers to come out - (and I don't mean obscure leaked "maybe-genuine" ES benchmarks) - then pass judgement on AMD and Bulldozer.

Sure, but the wait has become unbearable. Just out of sheer curiosity at this point I want to know what Bulldozer is all about, it's not often we see an all new CPU architecture.

 

[Arg Clin]

I agree, I'd take it a step further and say the speculations hold almost no value. We know quite a lot about Bulldozer as an architecture, but there are so many unknowns that predicting actual performance is really difficult. The thing is, a pattern we oft see is when a product is late and much secrecy surrounds it, it usually ends up falling short. Not a hard and fast rule of course..

I agree - the delays is a bad sign. Certainly! On the other hand it's a "once it's done it's done"-kind of thing. If AMD were to release an underperforming BD right now, the imageloss would be unrecoverable for the most part, even if they were able to rework it to much greater performance later. Same or maybe even worse - a well performing BD that AMD couldn't bring in volume to the market. Once BD is out the door, the clock is ticking really fast on Intels anticipated countermove. Also marketing ressources should be considered, taking into acount that Zacate and Llano is fresh out the door.

Speculation is fun, but there could be a number of reasons BD is held back. All we can do is guess and hope.

Sure, but the wait has become unbearable. Just out of sheer curiosity at this point I want to know what Bulldozer is all about, it's not often we see an all new CPU architecture.

Without a doubt! Unfortunately, in the bigger picture, we enthusiasts do not really pull that heavy a weight. In a corporate sense Llano might be bigger news than BD.

 

[Topweasel]

Consistantcy

Consistantcy? I would at least expect some consistency in the spelling of consistency.

 

[jhansman]

While we're all predicting, here's mine (worth, BTW, exactly what you pay for it): BD comes out, the other side minimizes it's abilities, contribution, and value. Its rooters overstate the same. It garners a respectable amount of business, but the other side never really feels any pain. And, we all wait and speculate on the next round of magical transistors.

 

[grimpr]

Sure, but the wait has become unbearable. Just out of sheer curiosity at this point I want to know what Bulldozer is all about, it's not often we see an all new CPU architecture.

Hot Chips 23 is just around the corner.

http://www.hotchips.org/program/program-23

 

[Abwx]

Hot Chips 23 is just around the corner.

http://www.hotchips.org/program/program-23

- Practical power gating and dynamic voltage/frequency
scaling issues (AMD)

LOL.....

Perf Power-Efficient Bulldozer-Core x86-64 Server, WS & Desktop Procs

 

[Abwx]

Some numbers ???..

http://browse.geekbench.ca/geekbench2/compare/425710/412187

 

[notty22]

AMD Considers Equipping FX Chips with Liquid-Cooling Solution.

If they are putting out feelers with this idea. I think its great.
It would give it a nice advantage over most usual smaller high rpm stock hsf.