Fudzilla: Bulldozer performance figures are in

[Abwx]

Proof?

Pro tip: you won't find it. A Phenom II X4 at 3GHz is a bottleneck for a Radeon HD 6950 or GTX 560 Ti and higher.

The only bottleneck Llano suffers from is the need of more memory bandwidth for the IGP. When outfitted with 1600MHz RAM it stretches its legs.

Yes, but with 1866mhz memory, it doesnt do better, wich means
that the CPU is then the bottleneck at these memory frequencies.

 

[LOL_Wut_Axel]

Yes, but with 1866mhz memory, it doesnt do better, wich means
that the CPU is then the bottleneck at these memory frequencies.

Sorry, but this makes absolutely no sense. An A8-3850 with a Radeon HD 6850 is MUCH faster than with the Radeon HD 6550D IGP. That means the GPU is the bottleneck in the system with the A8-3850.

The fact that you stop getting performance increases at 1866MHz means that the IGP no longer needs any more bandwidth and can run optimally.

 

[Idontcare]

http://www.xboxmb.com/forum/93-pc-related/62252-fx-8150p-benchmarks.html

FX-8150P Benchmarks - prob fake?

3DMark 2011: P6265 points
Cinebench R10 (64-bit): 24 434 points
Fritz Chess Benchmark: 14 197 kilo nodes per second
PCMark 7: 3045 points
Super Pi 1M: 19.5 seconds
x264-Benchmark: 136.29fps 1 pass, 45.4fps 2 pass

Are all these guys just copying each others #s over the net?

Actually he claims those are the 8130 benches.

 

[LOL_Wut_Axel]

Actually he claims those are the 8130 benches.

Kinda funny how all of these are probably made up. The main guy responsible for this said he lied. Would be good if we could get some benchmarks from someone reputable that could show the chip on a good quality video and running a system.

 

[soccerballtux]

http://www.xboxmb.com/forum/93-pc-related/62252-fx-8150p-benchmarks.html

FX-8150P Benchmarks - prob fake?

3DMark 2011: P6265 points
Cinebench R10 (64-bit): 24 434 points
Fritz Chess Benchmark: 14 197 kilo nodes per second
PCMark 7: 3045 points
Super Pi 1M: 19.5 seconds
x264-Benchmark: 136.29fps 1 pass, 45.4fps 2 pass

Are all these guys just copying each others #s over the net?

Actually he claims those are the 8130 benches.

if that's true, there's no IPC improvement per clock in single threaded apps.
Actually there could be since I'm running my cpu-nb at 2.8ghz,
but I scored 17.4 or something at 1M on my 4ghz 965, so at 3.6ghz thats 19.3-ish
in which case, we're all doomed, because the dull-dozer won't be any better at playing games weighted towards single-threaded, poorly optimized engines (WoW, SC2, etc).
Maybe they've added some extra magic smoke that we don't know about that doesn't manifest itself in branch-light mathematical computations.
Its going to pretty ho-hum if all we can do is be excited about faster cinebench numbers.

 

[drizek]

lol

Image titled: umadbro.png

 

[PreferLinux]

Cheapest gulftown. it's very unlikely that a 6+ core SB-E is going to be less than the cheapest 6 core 1366 cpu, right?

Here's the post you were quoting:

I'd be very surprised if the lowest priced SB-E was much over $350 street. Then again, this is a seat of the pants opinion and worth exactly what you paid for it!

I would take that to be the quad-core.

 

[gramboh]

I check this (and other) threads every week and cannot believe there are still no verified benchmarks from AMD. What the hell are they doing over there.

 

[Rifter]

I check this (and other) threads every week and cannot believe there are still no verified benchmarks from AMD. What the hell are they doing over there.

still trying to respin to hit clocks that wont make BD look like a complete joke is what i think they are doing.

 

[Riek]

if that's true, there's no IPC improvement per clock in single threaded apps.
Actually there could be since I'm running my cpu-nb at 2.8ghz,
but I scored 17.4 or something at 1M on my 4ghz 965, so at 3.6ghz thats 19.3-ish
in which case, we're all doomed, because the dull-dozer won't be any better at playing games weighted towards single-threaded, poorly optimized engines (WoW, SC2, etc).
Maybe they've added some extra magic smoke that we don't know about that doesn't manifest itself in branch-light mathematical computations.
Its going to pretty ho-hum if all we can do is be excited about faster cinebench numbers.

Besides that those benchmarks are fake,
superPI has absolutely no relation to single thread performance or gaming performance whatsoever.

I check this (and other) threads every week and cannot believe there are still no verified benchmarks from AMD. What the hell are they doing over there.

AMD said no benchmarks until launch. The product isn't launched so no benchmarks. they stated this for a year or more, so don't expect performance leaks from AMD before launch. (neither happened with the 6850/6870/6950/6970

 

[jvroig]

I check this (and other) threads every week and cannot believe there are still no verified benchmarks from AMD. What the hell are they doing over there.

There are still no verified benchmarks simply because the product has not launched yet. AMD has been clear and consistent about this: no benchmarks until launch.

still trying to respin to hit clocks that wont make BD look like a complete joke is what i think they are doing.

That could certainly be the reason for the delay of the launch, so indirectly the reason why there are still no benches since benches will only come at launch.

if that's true, there's no IPC improvement per clock in single threaded apps.

IPC per clock is redundant, but besides this minor mistake in wording, the real problem with the supposed bench is that it isn't official, most likely fake, and contradicts the only official word from AMD about BD IPC, which is that IPC has increased. It's no bench, but that's the official word regarding IPC.

 

[RussianSensation]

Besides that those benchmarks are fake,
superPI has absolutely no relation to single thread performance or gaming performance whatsoever.

I wouldn't say that. In fact, it's extremely indicative of the 2 things you just mentioned.

Off the top of my head, Athlon 64 / X2 and Core 2 Duo/Quad, Core i3/i5/i7 and Sandy Bridge all outperformed their competitors in this benchmark. All of those processors turned out to be superior in both single threaded and in gaming performance scenarios relative to their competition.

You think it's just a coincidence?

As a side reference Super Pi 32M places:

Core i7 860 3.9ghz = 575 seconds
Core 2 Quad Q6600 3.4ghz = 949 seconds

Performance advantage = 39%

Core i7 was 20% faster than Yorkfield in IPC. 3.9ghz is 15% faster clock speed vs. 3.4ghz.

In total I should have expected a performance increase of 1.2 * 1.15 = 1.38x over my Q6600 3.4ghz. This benchmark showed exactly that!!

It's ironic that AMD supporters started to dismiss SuperPi benchmark as irrelevant only after Athlon 64 / X2 era ended because every single AMD processor after that time was noncompetitive in it (and as it turned out, overall as well).

 

[Riek]

I wouldn't say that. In fact, it's extremely indicative of the 2 things you just mentioned.

Off the top of my head, Athlon 64 / X2 and Core 2 Duo/Quad, Core i3/i5/i7 and Sandy Bridge all outperformed their competitors in this benchmark. All of those processors turned out to be superior in both single threaded and in gaming performance scenarios relative to their competition.

You think it's just a coincidence?

As a side reference Super Pi 32M places:

Core i7 860 3.9ghz = 575 seconds
Core 2 Quad Q6600 3.4ghz = 949 seconds

Performance advantage = 39%

Core i7 was 20% faster than Yorkfield in IPC. 3.9ghz is 15% faster clock speed vs. 3.4ghz.

In total I should have expected a performance increase of 1.2 * 1.15 = 1.38x over my Q6600 3.4ghz. This benchmark showed exactly that!!

It's ironic that AMD supporters started to dismiss SuperPi benchmark as irrelevant only after Athlon 64 / X2 era ended because every single AMD processor after that time was noncompetitive in it (and as it turned out, overall as well).

Why don't we still use win95 to measure wnidows performance? Because its outdated.

SuperPI unfortunatly uses outdated instructions that became obsolete since x64. The moment microsoft stepped to 64bit every driver didn't contain x87 instructions or legacy MMX... So testing an application that solely relies on these instruction sets is irrelevant for todays and tomorrows applications.

IPC isn't a constant thing, we already said and indicated and proofed that to you multiple times. So making up IPC statements and using it indicate your rightness is ridiculous. Even more so if you talk about Quad cpu's using 1 thread while both of them are running out of spec with different busspeeds..............
That said, ever wonder why previous designs seem to score similarly in superPI? Because they only upped the clockspeed and changed the frontend. The pipelines themselves didn't get such rework that would change the x87/MMX performance. Hence why there aren't known performance drops over previous generations.

However cpu's made now (and that includes Haswell) that have a reworked fpu unit will not focus on x87 throughput hence they will have different performance characteristics then previous designs. But considering the instruction set is obsolete, unused and basically unsupported for multiple years now its performance isn't relevant to overall cpu performance. Is has also no impact on the cpu ILP capabilities. It has also no relation to gaming performance and also no relation to any application that was created <11 years ago where SSE2 completely replaced x87.

 

[PorscheMaD911]

... Right now the average consumer has far more computing power than he has need of.

You are absolutely correct. I guess this is an enthusiast forum, so only the high-end stuff will be given the thumbs up, but this is a timely reminder that it's all relative.

I just upgraded my PC to the following:
- Antec 300 case
- Antec Neo Eco 450W power supply
- Asus M4A87TD/USB3 motherboard
- Phenom II X4 955 BE CPU
- 2 x 2 Gb Corsair 1600Mhz DDR3 RAM
- Gigabyte Radeon HD5750 video card
- Samsung 500Gb Spinpoint F3 hard drive
- Samsung SH-222AB optical drive
- Windows 7 Home Professional OEM

I know, I know, in the face of all the Intel I5 recommendations it's pedestrian madness, but a) the budget wouldn't stretch (saving up for my wedding), b) I like AMD, and c) to me it already feels crazy fast. Probably because I've been using my faithful little Asus N10J netbook as my sole computer for the last 18 months. I mean, such is the beauty of quad-core that on my new desktop you can rip / encode a DVD with Handbrake to H260 / AAC, run an AVG virus scan, watch a 1080p movie in VLC, have downloads going, windows update installing (all at the same time of course) and *still* have a super-responsive system. Pretty good for a $119 AUD CPU, and there's no way the average user won't be happy with that!

I'm finding that If you read forums widely, it's the same with enthusiasts in any field. Eg 'you can't seriously be thinking about buying a 997 Carrera, only the GT3 RS 4.0 / Carrera 2.7 RS / 993 GT2 / (insert personal favourite here) is a true driver's Porsche'.

Actually the main difference I notice in PC forums is the lack of a 'they don't do things like they used to' sentiment. There's a lot of that on the hi-fi / headphone / style forums! Maybe I browse too many forums, but I don't know, it's interesting to read about many different things

Anyway, just my 2 cents. I'm sure Bulldozer will be super-fast, and I'm sure Intel's response will be even faster. And one day when the prices fall far enough, my desktop feels slow and I've run out of overclocking room I might just upgrade to one of them!

 

[ed29a]

still trying to respin to hit clocks that wont make BD look like a complete joke is what i think they are doing.

[Citation needed]

 

[bryanW1995]

Here's the post you were quoting:

I would take that to be the quad-core.

Ok, I think that I've clarified it now. I was specifically referring to 6 cores cpus. I agree that it's likely that there will be ~ $350 quadcore SB-E cpus available, but as stated I don't see the value in them.

 

[Mopetar]

Ok, I think that I've clarified it now. I was specifically referring to 6 cores cpus. I agree that it's likely that there will be ~ $350 quadcore SB-E cpus available, but as stated I don't see the value in them.

It might be worthwhile if you want to get a 2011 rig together without dropping too much money at one time. Assuming that Intel doesn't retire it too soon, it should provide some room for future upgrades, perhaps to IB-E or something like that.

 

[podspi]

[Citation needed]

There is no proof that they are respinning for clocks (although they probably are, if there are big problems still in the silicon that isn't good) -- but they're definitely re-spinning. Otherwise wth are they waiting for? Whatever BD is, it isn't going to get MORE competitive by being delayed.

 

[ed29a]

There is no proof that they are respinning for clocks (although they probably are, if there are big problems still in the silicon that isn't good) -- but they're definitely re-spinning. Otherwise wth are they waiting for? Whatever BD is, it isn't going to get MORE competitive by being delayed.

Everyone is speculating left and right without any facts. They could have discovered a bug in the CPU and didn't want the repeat of the TLB episode? Plenty of reasons why it is delayed: subpar performance, last minute bugs found, head engineer's dog got hit by a car and he became depressive and quit, cosmic rays messed up the wafers, etc ... Some more plausible than others for sure, but until we know from reputable sources (not some random guy on a random enthusiast board), it's still baseless speculation.

 

[Schmide]

I wouldn't say that. In fact, it's extremely indicative of the 2 things you just mentioned.

Off the top of my head, Athlon 64 / X2 and Core 2 Duo/Quad, Core i3/i5/i7 and Sandy Bridge all outperformed their competitors in this benchmark. All of those processors turned out to be superior in both single threaded and in gaming performance scenarios relative to their competition.

You think it's just a coincidence?

As a side reference Super Pi 32M places:

Core i7 860 3.9ghz = 575 seconds
Core 2 Quad Q6600 3.4ghz = 949 seconds

Performance advantage = 39%

Core i7 was 20% faster than Yorkfield in IPC. 3.9ghz is 15% faster clock speed vs. 3.4ghz.

In total I should have expected a performance increase of 1.2 * 1.15 = 1.38x over my Q6600 3.4ghz. This benchmark showed exactly that!!

It's ironic that AMD supporters started to dismiss SuperPi benchmark as irrelevant only after Athlon 64 / X2 era ended because every single AMD processor after that time was noncompetitive in it (and as it turned out, overall as well).

My previous post on this.

Not so much a coincidence but more of a correlation, a better cache/memory system can lead to more IPC but doesn't necessarily guarantee it. In this case I would say Intel's superior cache/memory predictor leads to the increase.

 

[Davidh373]

My guess is it's delayed because they're pulling what companies always do. This is all a lot of marketing and hype. If everyone hears about it, most everyone will want one, and that's what matters. Do you guys think a lot of people are like us? I sure don't after knowing what my friends do. Most of them go out and buy a cruddy Dell or HP (probably an Alienware! o.o) and get whatever they have a budget for (be it 2&#37; better or 20% better). The other 'Enthusiast" types I know who custom build go on newegg, filter every part by "Price: High to Low" and buy the highest priced parts on there.

Just the other day I had one of my good friends wanting to get a computer for his GF to do 3D modeling and Photoshop. I recommended a machine at around $800. He said just the day before he was looking and said his build cost well above $2500. I can only assume he selected two 580s, a 590, two 6970s, or a 6990, maxed out his RAM, and had a 990X processor in there with an ASUS crosshair mobo. No doubt the best machine that kind of money can buy, but very stupid in the value department and only around 10% better for a total of $1500.

The average electronics consumer has no perception of value. They just say "higher numbers means better, I want the one with the higher numbers" that being core count, clock speed, voltage, frequency, wattage, capacity... the list goes on. There is no consideration of architecture, nor is there consideration of comparison clock for clock. They are just there to buy stuff, and unfortunately they make up a large segment of the market. All they really need to know is "Hey, that's that new processor, it's an 8 core... that means it's better somehow... I want it..."

As for the actual benchmarks, I'm kind of pulling for AMD, even if they're the Apple computers of the processor world. I'd love to see them do better than SB and maybe even SB-E. With a 130W TDP and higher clocks they better! I myself am waiting for around 2013 to upgrade my workstation again (around the time of Haswell I think), but I certainly have a lot of friends itching to upgrade right now. I don't do fanboyism, but this could get interesting if AMD actually bests Intel, since I have only recommended Intel for enthusiast builds thus far

 

[Idontcare]

<--- looks up at thread title...

...then reads the following:

baseless speculation.

you, you do realize this entire thread, the whole thing from top to bottom, is speculation, dare I say the baseless kind too...right?

 

[Abwx]

<--- looks up at thread title...

...then reads the following:


you, you do realize this entire thread, the whole thing from top to bottom, is speculation, dare I say the baseless kind too...right?

That s utterly right....

Beyond tablets, AMD this quarter will release server and desktop chips based on its new Bulldozer architecture, which has been in development for around five years. The Opteron server chip based on Bulldozer, codenamed Interlagos, has up to 16 cores and will be up to
40 percent faster than existing chips with up to 12 cores, while consuming the same amount of power, Bergman said.
The desktop chips include up to eight cores and will be used in high end desktops aimed at uses such as gaming.

http://www.computerworlduk.com/news...md-content-to-leave-smartphone-market-to-arm/

 

[Phynaz]

The Opteron server chip based on Bulldozer, codenamed Interlagos, has up to 16 cores and will be up to 40 percent faster than existing chips with up to 12 cores, while consuming the same amount of power, Bergman said.

An AMD exec claiming a 40&#37; performance improvement.

I have a vague sense of deja vu, can't quite put my finger on it though....

 

[996GT2]

The Opteron server chip based on Bulldozer, codenamed Interlagos, has up to 16 cores and will be up to
40 percent faster than existing chips with up to 12 cores

Wow, a 16 core chip being up to 40&#37; faster than a 12 core chip?

Who would have thought!