What Were They Thinking

L

larkin

Gawd
Joined
Oct 24, 2005
Messages
907
This is the question that has to be on the minds of everyone right now.

100% less power efficient than SB, bested by their previous generation CPU in most cases.

The engineers have had ample time (see 7 years) to test this thing. What were they thinking continuing down this road? I don't know if I would use this CPU if it were free.
 
I am asking myself the same question. I was ready to buy it tonight. Less power efficiency under load than the last gen CPU on 45nm?

Real reviews are showing BD to be worse than what the fake reviews showed.

Comparatively speaking this is the worst CPU AMD has ever released. There is not a single positive area about this CPU, well except for meaningless LN2/Helium overclocking with all the cores disabled.
 
they went for a high branch and missed

AMD is a fraction of the size of Intel
 
Check out this thread here at HardOCP (Possible explanation for poor single threaded performance)

High transistor count is because of all the damn cache, I reckon.

Summary: Better OS scheduling to take advantage of the quirky architecture, thrashing within L2/L3 cache?
 
It only seems to be good for full-time x264 video encoders. Still, given that the prices of Thubans will now fall, i don't think it's worth the price/performance difference over the Thuban.

Also, yes, power consumption is a disappointment. I for one, usually go for 45W CPUs (Thuban is the exception and i undervolt it). This thing altough at 32nm consumes more than the lower clocked Thuban...

It's unbelievable that they delivered a much anticipated CPU that doesn't seem to have a single strong point over the competition and considering price/performance, not even over the Phenom II x6!
 
Yossarian22 said:
Check out this thread here at HardOCP (Possible explanation for poor single threaded performance)

High transistor count is because of all the damn cache, I reckon.

Summary: Better OS scheduling to take advantage of the quirky architecture, thrashing within L2/L3 cache?
Click to expand...

Tom's actually did a comparison of benchmarks in win7 vs win8 with it's better scheduler and there is improvement but only about 10% at best in the game example. Somehow, 2500K is still trouncing 8150 at SC2, WOW, CIV V and others. I'm really interested to know how it can be beat so astonishingly (20fps behind 2500K @ SC2).
http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-23.html

The cache issue has been discussed but looks like only, again, a 10% increase at best. There seems no rescuing this chip from it's abysmal power consumption and performance shortcomings although over the next days and weeks I would like to see some technical explanations for the more pronounced shortfalls, maybe even from AMD directly.
 
Yeah good on Tom's for actually doing an SC2 benchmark. That was the last nail in the coffin for me. It's slower clock per clock than Thuban/Deneb and Intel beats it by a huge margin.
 
When the Phenom 1s came out, people were speculating that they had "reverse hyper-threading" to make up for their lack of... goodness. Let's see what stories about BIOS fixes and stuff people come up with for these.

I was an AMD fanboy back when my s754 Sempron got a 50% overclock and played FEAR on high settings. It's pretty difficult now though.
 
The worst in all this, is if the flagship can't cut it, what's the point in buying one of the lower clocked BD or the 6core BD, when you can still grab a Phenom II at less...
 
larkin said:
Tom's actually did a comparison of benchmarks in win7 vs win8 with it's better scheduler and there is improvement but only about 10% at best in the game example. Somehow, 2500K is still trouncing 8150 at SC2, WOW, CIV V and others. I'm really interested to know how it can be beat so astonishingly (20fps behind 2500K @ SC2).
http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-23.html

The cache issue has been discussed but looks like only, again, a 10% increase at best. There seems no rescuing this chip from it's abysmal power consumption and performance shortcomings although over the next days and weeks I would like to see some technical explanations for the more pronounced shortfalls, maybe even from AMD directly.
Click to expand...

Thanks. I typically don't visit Toms.
Check out this post over at xtremesystems and the guy's site translated from Korean here.

I agree the power draw is insane. With phase change you can get around 5.9GHz @ ~1.63vdc atleast on the engineering samples when they gave a sneak peek [H]ardOCP reported on. (at about 3min05sec)

If FX-4100 is binned/chip harvested 8xxx could be interesting to mess around with, since bulk of the single threaded performance seems to come from just 4 cores, as that Korean guy's results show.
 
i wonder if we will see any sites or even [H] revisiting with retail chips, since i saw they mentioned they were samples given to them by AMD. doubt performance would be any different but i wonder if the retail chips are a little more power efficient. either way 200ish for the 8120 is probably what i'll end up getting and throw it in a asus sabertooth motherboard. i still can't bring myself to ever support intel with my own money even though i still recommend them to others.
 
Most crazy thing is that they started working on it when Pentium IV was around.

What killed PIV was technical process blocking them from reaching projected frequencies they needed for long pipeline to perform well. At that time AMD enjoyed biggest success in their history due to sticking with high IPC architecture of short pipeline.

So it makes me scratching my head why they went into excactly the same direction Intel went and got burned.

My hipothesis currently is that they assumed much much higher adoption and better scalling of multi threaded apps (remember they started in PIV times where cpu were 1 core with HT or 2 cores) so ST performance wouldn't matter.

Rest is probably result of doing both tick and tock at the same time.

2011 should have been 32nm Phenom III - llanos for low end, 6 cores 32nm high clocked parts for AM3+
 
Michaelius said:
Most crazy thing is that they started working on it when Pentium IV was around.

What killed PIV was technical process blocking them from reaching projected frequencies they needed for long pipeline to perform well. At that time AMD enjoyed biggest success in their history due to sticking with high IPC architecture of short pipeline.

So it makes me scratching my head why they went into excactly the same direction Intel went and got burned.

My hipothesis currently is that they assumed much much higher adoption and better scalling of multi threaded apps (remember they started in PIV times where cpu were 1 core with HT or 2 cores) so ST performance wouldn't matter.

Rest is probably result of doing both tick and tock at the same time.

2011 should have been 32nm Phenom III - llanos for low end, 6 cores 32nm high clocked parts for AM3+
Click to expand...


Go back and look , bulldozer changed in the last 3 years or so. The original was scrapped .
 
larkin said:
This is the question that has to be on the minds of everyone right now.

100% less power efficient than SB, bested by their previous generation CPU in most cases.

The engineers have had ample time (see 7 years) to test this thing. What were they thinking continuing down this road? I don't know if I would use this CPU if it were free.
Click to expand...

It doesn't work like that, if they had this 7 years ago they would be much further along. Sadly AMD has performance issues they are still trying to catchup.

Some benchmarks do show that the performance is there.

Marketing wise this was not the best move. The grand standing on the overclocking world record seems silly now.
 
larkin said:
Grand standing on the OC seems silly indeed especially when a mild 4.8Ghz OC uses 586 watts (!) vs. 2500K at 5Ghz using only 311. This is not an OC friendly chip by any account.
http://www.bit-tech.net/hardware/cpus/2011/10/12/amd-fx-8150-review/10
Click to expand...

gotta say something is fishy there.

at idle its 110/126 and at load its suddenly 244 and 586 ? something isn't right at all.

He said they are only using prime so its the cpu that is stressed in this. So the cpu is already using 134w more non overclocked , If we compare the over clock its usings 460 watts more than idle ?

Can the motherboard even supply 460w of power to the cpu ?


looking at kyle they have the fx 8150@ 4.6ghz using 452 watts vs 120 idle. That means the cpu overclocked is only pulling 332w more. An extra 200mhz is supposed to cause this thing to pull another 130w of power ? something has got to be wrong there
 
GTX590@bit-tech vs GTX570@[H]. Nothing is wrong, you just didn't care to look for difference in system specs.
 
It doesn't beat a 1st generation Core i7 handily, why would you invest in dead end architecture before its even been released?

I SO wanted this to work for AMD, cause it would have worked for the consumer, but this, is a total and complete waste of resources that AMD did not need, and will not recoup.

If their stock doesn't shit the bed today I'll be shocked.
 
larkin said:
Grand standing on the OC seems silly indeed especially when a mild 4.8Ghz OC uses 586 watts (!) vs. 2500K at 5Ghz using only 311. This is not an OC friendly chip by any account.
http://www.bit-tech.net/hardware/cpus/2011/10/12/amd-fx-8150-review/10
Click to expand...

What is funny that i have seen 5ghz (on air even)from a "local" website while [H] and Anand were stuck at 4.6 .

Im not to shocked about the extra power it uses (while overclocked) supposedly that is one of it's quirks and was reported several months ago. Btw the [H] review uses less power but 450 (minus 120 Watt from idle) Watt is still somewhat impressive.
 
stiltner said:
If their stock doesn't shit the bed today I'll be shocked.
Click to expand...

Outside of never having owned an Intel CPU since AMD came to the market, I also own AMD stocks. I'm down almost 40% on my investment. I was truly hoping that this release would at least bring their stock back to a level that I could exit without a lose. Looks like I'll be holding my stock with them for another product lifecycle.
 
I think I might be shorting some AMD stock today.

I really wanted BD to do well, too.
 
stiltner said:
It doesn't beat a 1st generation Core i7 handily, why would you invest in dead end architecture before its even been released?

I SO wanted this to work for AMD, cause it would have worked for the consumer, but this, is a total and complete waste of resources that AMD did not need, and will not recoup.

If their stock doesn't shit the bed today I'll be shocked.
Click to expand...

You do realize that AMD sales of the Bulldozer were never a big part of their income, they make money on low end stuff more then anything else.
 
vdozer.jpg
 
Pieter3dnow said:
You do realize that AMD sales of the Bulldozer were never a big part of their income, they make money on low end stuff more then anything else.
Click to expand...

That is pretty much keeps them afloat and the video card business as well..
 
I'll tell you what they were thinking. JF-AMD and his buds in server won the battle. They can move more BD CPUs to a single massive client than they can to all the fanbois put together. Someone had to lose out and it was us. Strictly a dollars and cents choice aggravated by the fact that GloFo is as good at making silicon as Paula Deen is in making diet meals.

R.I.P. AMD (at least on the high end consumer side)
 
I'm not too well versed in what's happening server side, but does that mean that Opteron chips based on BD are any different? Or did you just mean that with its seemingly near-exclusive multithreaded performance that its an attractive, cheap server chip?

Overall, I'm really bloody disappointed if this turns out to be as bad as it looks. I don't see why they're release anything branded for the desktop at all if it had shoddy single-core performance. Anyone banking on quad-plus threaded apps had to have been a moron - for most desktop software being dual-threaded has been a slow adoption rate for risk of alienating those in the desktop mainstream.

Why sell a "next generation" at all if you can't 1) Beat your OWN previous high-end across the board by a considerable measure 2) Match/Beat Intel's current generation mainstream at lower/equal prices, more or less across the board in real-world usage. If you can't beat them, come close but undercut on pricing etc... I'm not asking BD to wipe the floor with SB-E Socket 2011's Extreme Edition but... blah. This seems really disappointing.

So much so in fact that I'm waiting to see if there are any further developments with real-world performance both at stock and OC'ed on air and water....
 
I've lost all hope in AMD when it comes to "high end" CPU's, face it, they have disappointed us for many years and to catch up to Intel is 1 to 100.
 
Xaeos, wait 7 more years. Maybe by then TrinityPiledriverWhateverTheHellItIs will beat Intel. Maybe by then we'll also all be commuting to work on flying pigs.

John Vuong, 1 to 100000000000000000000000000000000000000000000000000000 maybe...
 
I think someone owes me an apology for insulting me a couple of weeks ago after I posted about rumors that BD would have lower IPC than Thuban...

Today's results are about what I expected.

Clock for clock, worse than Thuban, but able to compensate for it by hitting higher clocks. Reducing the lead slightly on Intel, but nowhere near competitive with 2600K.

I;'m still getting one. I refuse to buy from the Evil Empire (Intel).
 
Here I am after having a 2600K for months not wanting to wait for Bulldozer and feeling just fine. Not sure what everybody is complaining about, though, you guys all seem to be gamers and BD does just fine for gaming. I mean, I'm actually impressed at the gaming performance it displays (but I do a lot more than gaming). I wish there were an easy way of benchmarking virtualization performance so I'd know if it were more worth my time to get Thuban for my home server or Bulldozer.
 
Weeth said:
I'll tell you what they were thinking. JF-AMD and his buds in server won the battle. They can move more BD CPUs to a single massive client than they can to all the fanbois put together. Someone had to lose out and it was us. Strictly a dollars and cents choice aggravated by the fact that GloFo is as good at making silicon as Paula Deen is in making diet meals.

R.I.P. AMD (at least on the high end consumer side)
Click to expand...

Server parts are identical to desktop ones feature wise, they are even older steppings ( since they needs more validation you need to stick with an earlier one ). So if desktop is B2G, as example, server might be B2E. Of course, the server die will go by a different name but it's more or less the same chip.

BD uarch was a conscious design choice; they decided to stop trying to fight for best single threaded performance. Instead try to see how much you can simplify the core without losing too much performance and try to cram as many as possible on a given area.
 
I must admit I wont be rushing out to buy one.

I felt this would be the case. I was watching a few folks on this forum setting themselves up for a very big fall over the past few months and their lack of response so far can only be the fact they are currently hanging by a belt in their closet or fallen on their swords.

You know who you are. With tech, an air of skepticism never hurts.

Bulldozer may be a fine CPU for what it does but its not extraordinary in any shape or form. Its performance doesn't warrant the sheer number of tech presentation slides AMD have made for it.
 
savantu said:
Server parts are identical to desktop ones feature wise, they are even older steppings ( since they needs more validation you need to stick with an earlier one ). So if desktop is B2G, as example, server might be B2E. Of course, the server die will go by a different name but it's more or less the same chip.

BD uarch was a conscious design choice; they decided to stop trying to fight for best single threaded performance. Instead try to see how much you can simplify the core without losing too much performance and try to cram as many as possible on a given area.
Click to expand...

Sure, but the BD Opterons are better suited to the specific demands of servers than the BD FX are to the needs of high end enthusiast-gamers. On the latter, BD is an epic phail. They could have added two cores to an X6 and had the damn thing out years ago. FX is just another boring mid-range chip. Big deal. Who cares.

daglesj said:
I must admit I wont be rushing out to buy one.

I felt this would be the case. I was watching a few folks on this forum setting themselves up for a very big fall over the past few months and their lack of response so far can only be the fact they are currently hanging by a belt in their closet or fallen on their swords.

You know who you are. With tech, an air of skepticism never hurts.

Bulldozer may be a fine CPU for what it does but its not extraordinary in any shape or form. Its performance doesn't warrant the sheer number of tech presentation slides AMD have made for it.
Click to expand...

Hmm... what's that I hear coming from AMD Zone... just the gentle rustling sounds of the fanbois swinging from their belts! :)
 
I want that AMD marketing guy in here to see what he has to say about this epic train wreck. Many of us are AMD supporters, and we were hoping Bulldozer would be our Luke Skywalker, but it looks like he never escaped the trash compactor.
 
Monkey God said:
I want that AMD marketing guy in here to see what he has to say about this epic train wreck. Many of us are AMD supporters, and we were hoping Bulldozer would be our Luke Skywalker, but it looks like he never escaped the trash compactor.
Click to expand...

JF-AMD has nothing to worry about. His job is secure since BD Opterons will likely do mildly well in the server space and that's his dept. AMD sold the enthusiast-gamer market down the tubes but server guys care more about the color of their rackmount rails than they do about all the enthusiast-gamers on the planet put together. However, that's where FX belongs... in the trash compactor.
 
You must log in or register to reply here.
Back
Top