CPU/Mobo AMD Bulldozer Discussion Thread

Status
Not open for further replies.
Well pricing is the key here

If bulldozer's performance is at par with intel or even 5-10 percent less. but if it consumes lesser power and is priced at-least 30 percent lesser compared to intel then i guess amd might well be in the the games.

cuz over all price to performance ration would be much better in amd, cuz amd mobos are cheaper compared to intel counter parts.

personally I would still bet on intel SB on current scenario
 
you cant gauge CPU performance at 1440P resolution

As resolution increases the role of CPU in final FPS numbers diminishes. The actual processing load does not change between resolution (CPU) whereas graphics subsystem starts getting loaded at resolutions above that.

When testing for CPU performance, the resolution should be kept at 1440x or 1680x to minimise impact of GPU.

Agree that 8-core vs 4-core is not a fair fight. Eventually though if you move to a fully programmable scalar structure like GPUs have done, the number of cores will be at best an academic discussion.

Most apps today are multi-threaded anyway, though some have limitations on the number of cores. I expect given these results, a lot of benchmarks will reveal Bulldozer to be slower than the Core i gen 2.

But then we've seen even their own Sandy Bridge-E chips to be slightly slower than these chips with a good combination of high IPC, clock rates and power footprints.

The key decision to be taken will be whether this is quick enough. An Athlon II quad or Core i3 (last gen) is quick enough for most of us and I wager no one can tell the difference between that and an i7 gen 2 in everyday tasks. In any case upgrading your SSD will get you a far bigger speed boost than upgrading your processor unless you're stuck with some dinosaur from over three years ago.

If the Bulldozer retains this level of performance, has a good power envelope, maintains compatibility with earlier coolers, keeps TCO lower than Intel and is available at a competitive price (say 10-15% cheaper), it's a decent enough offering for the value-conscious buyer. Frankly, I will be waiting for this as my 550BE is now growing long in the tooth.
 
  • Like
Reactions: 1 person
As far as I am concerned, its another failed attempt by AMD. They put back the releae and had to reduce prices so that the performance is comparable to Sandy bridge. And with SB E series lurking around the corner, AMD will remain budget king and thats it. Performance and AMD will not appear in one sentence, but value for money will.
 
^^^ value for money is what matter to 90% of the ppl.

of-course there are 10% who will negate price and chose the beast [does those high end CPU's/GPU's use there all potential in our daily works]
 
read this quote

If you can get 8 cores for the price of 4 cores, you're going to get lower performance on any program that doesn't utilize all 8 cores (as in, most).

So, unless you're encoding video EXCLUSIVELY, the same rules apply to BD as 1090T - buy Intel.

get that and understand it!
 
People who say 'cores doesnt matter' and 'overall performance' is what counts are misinformed. read the sentence quoted in above post.
 
^^What applications do you use, and for what percentage of the time?

There is theoretical performance ceiling and there are everyday applications. Most apps used by home users don't push even an Athlon II quad-core to the brink, let alone more powerful products.

The moment you actually use the computer for demanding tasks, you can talk about performance and look for a processor that meets your performance expectations. The truth is that today computing power is far ahead of application demand.

If you spend 80% of your time working on Photoshop or 3DMax or [insert a professional application here] you will need a lot of grunt from your CPU. If all you do is the additional decompression using WinRAR, you'll save twenty minutes for every 1TB of data you decompress. Not worth losing sleep over. If you have a high resolution screen and a capable enough GPU, the CPU role is minimised.

Horses for courses. The word 'Failure' depends on who is defining goals and what the definition of succes is. All we can say is that single-thread performance is still lower than Intel. That is probably true. I don't know that that is a 'failure' unless we see cost-performance and platform interoperability, etc.

I don't think there is 'victory' or failure' here. That is a game played by fanboys and flame-baiters. We don't want that here, and it would be good if a technical discussion happened instead of the +1 posts.
 
AMD FX-Series CPUs to Launch on October 13

Source : news.softpedia.com/news/AMD-FX-Series-CPUs-to-Launch-on-October-13-222238.shtml

Nordichardware also managed to get hold of one 8150 and they benchmarked it using Super Pi benchmark... 8150 managed to complete the test in 21 seconds...
 
I'm not sure if this has already been discussed before but apparently the bulldozer part FX-4170 will have a base clock of 4.2Ghz!!which is the highest for any processor till date.Can we expect this to at least match the Core i5 2500k in performance considering that the clock is so much higher?

 
nuke'em said:
I'm not sure if this has already been discussed before but apparently the bulldozer part FX-4170 will have a base clock of 4.2Ghz!!which is the highest for any processor till date.Can we expect this to at least match the Core i5 2500k in performance considering that the clock is so much higher?

A good question. May I also say, finally.

At about 65% IPC of the Intel the BD 4.2GHz part would roughly equal a SB processor with similar core structure of about 2.8GHz. If they got the IPC to 75% (unlikely) then you're looking at it matching a 4-core 3.2GHz SB. So I would expect performance to fall between a 2300 and 2400 (?, not sure about SB speeds). Also expect the BD part to be unlocked, and not have much headroom above the 4.5GHz mark. You can expect higher power consumption under load simply because of the higher clock. Idle depends on how the processor logic is wired. If it can turn off cores, then maybe overall consumption wil be lower.

The octals seem to be far more sensible than the quads. Someone posted a silly benchmark of 8GHz with some exotic chemicals, but even then it was a dual core processor (8-Cc with 6 inactive cores) so it may seem that it is possible to turn off cores externally.

I guess it will come down to consumption efficiency, as I have posted earlier absolute performance is going to be a subjective experience. If the top dog costs $245, it's not competing with the top of Intel's range anyway. Unless they surprise us.
 
nuke'em said:
I'm not sure if this has already been discussed before but apparently the bulldozer part FX-4170 will have a base clock of 4.2Ghz!!which is the highest for any processor till date.Can we expect this to at least match the Core i5 2500k in performance considering that the clock is so much higher?


Nothing can be said before the real world tests..On papers ,yes it will compete i5 2500k but in real world,nothing can be said..
 
cranky said:
A good question. May I also say, finally.

At about 65% IPC of the Intel the BD 4.2GHz part would roughly equal a SB processor with similar core structure of about 2.8GHz. If they got the IPC to 75% (unlikely) then you're looking at it matching a 4-core 3.2GHz SB. So I would expect performance to fall between a 2300 and 2400 (?, not sure about SB speeds). Also expect the BD part to be unlocked, and not have much headroom above the 4.5GHz mark. You can expect higher power consumption under load simply because of the higher clock. Idle depends on how the processor logic is wired. If it can turn off cores, then maybe overall consumption wil be lower.

The octals seem to be far more sensible than the quads. Someone posted a silly benchmark of 8GHz with some exotic chemicals, but even then it was a dual core processor (8-Cc with 6 inactive cores) so it may seem that it is possible to turn off cores externally.

I guess it will come down to consumption efficiency, as I have posted earlier absolute performance is going to be a subjective experience. If the top dog costs $245, it's not competing with the top of Intel's range anyway. Unless they surprise us.

I had no idea AMD was so far behind in terms of IPC.Also, the FX-4170 has its turbo mode at only 100Mhz higher probably because of the high power consumption you talked about.And with high clocks and power consumption comes a lot of heat too.Besides,considering that the OctCore are selling in the 200-250$ range it looks like AMD might not even be trying to compete core for core against Sandy Bridge.
 
I had no idea AMD was so far behind in terms of IPC.Also, the FX-4170 has its turbo mode at only 100Mhz higher probably because of the high power consumption you talked about.And with high clocks and power consumption comes a lot of heat too.Besides,considering that the OctCore are selling in the 200-250$ range it looks like AMD might not even be trying to compete core for core against Sandy Bridge.

Why is core for core an important parameter now? Most of the developers have realized that multi threading environment is the logical way ahead, so they should be developing their codes. Besides small bumps in speeds, leads to high thermal output. In an 8 core processor, I believe would be better bet in the longer run, and should have better thermal envelope for everyday loads (I believe not all cores would be used for everyday applications, which would shutdown some cores, and since thermal envelope per core would be lower, should lead to better utilization).
 
Core for core is an important matter, but more for the chipmakers than us users.
If an Intel dual core is about half the size of an AMD quadcore, and they perform and cost about the same, then Intels going to make larger profits, have more money for research, and will extend it's technological lead.
They'll have a big pricing advantage, being able to cut prices at will, while AMD will have to match Intels prices, irrespective of whether they're making a profit or loss.
Also, a dual core will consume less power, and that should be enough to sway enterprise customers.
So while we may feel happy about being able to afford an 8-core (in reality 4 cores + hyper threading on steroids), AMD won't be able to stay in business long unless it makes more efficient processors.
 
Core for core is an important matter, but more for the chipmakers than us users.

If an Intel dual core is about half the size of an AMD quadcore, and they perform and cost about the same, then Intels going to make larger profits, have more money for research, and will extend it's technological lead.

Rather then cores its die size and efficiency which will determine margins. Intel with better fab tech is able to cram in more transistors per area. In addition if the foundry is having efficiency issues (a lot chips failing in the fabrication tray), this will impact company's profitability big time. So if AMD or Intel can have more cores in the same space, manufacturing costs should not be different from a chip which is dual core but has same die size.

Also, a dual core will consume less power, and that should be enough to sway enterprise customers.

I feel given a normal workload, a multi-core processor would consume lower power.

So while we may feel happy about being able to afford an 8-core (in reality 4 cores + hyper threading on steroids), AMD won't be able to stay in business long unless it makes more efficient processors.

Arent AMD's FX 8xxx series actual 8 core processors?
 
Rather then cores its die size and efficiency which will determine margins.

I agree, and that's what I meant when I said If an Intel dual core is about half the size of an AMD quadcore. When you've got less cores, and you're manufacturing on a better process - 32nm/22nm v/s 45nm/32nm, you're obviously paying less per processor than your competitor.

Arent AMD's FX 8xxx series actual 8 core processors?
Yes, but not in the true sense. They've got 4 dual-core modules, and the cores share resources within each module. So it's more like 4*1.75 cores.
Inside the AMD Bulldozer Architecture | Hardware Secrets

I feel given a normal workload, a multi-core processor would consume lower power.
Only if the idle cores can be completely disabled, which isn't the case right now.
 
Aces170 said:
Why is core for core an important parameter now? Most of the developers have realized that multi threading environment is the logical way ahead, so they should be developing their codes. Besides small bumps in speeds, leads to high thermal output. In an 8 core processor, I believe would be better bet in the longer run, and should have better thermal envelope for everyday loads (I believe not all cores would be used for everyday applications, which would shutdown some cores, and since thermal envelope per core would be lower, should lead to better utilization).
I had gaming performance as the only metric when I said that that,that's why core for core performance was important.I was planning to buy something at core i5 2500K's price point next summer.So I was really hoping for a quad core which can match the i5.I think you will agree that more than 4-cores won't make a difference for gaming performance at least at the moment,but the scene next summer could be entirely different after Ivy Bridge.
 
nuke'em said:
I had gaming performance as the only metric when I said that that,that's why core for core performance was important.I was planning to buy something at core i5 2500K's price point next summer.So I was really hoping for a quad core which can match the i5.I think you will agree that more than 4-cores won't make a difference for gaming performance at least at the moment,but the scene next summer could be entirely different after Ivy Bridge.

I think it will take developers still quite a while before >4 cores are properly utilized by the game. Heck, they have such a difficult time balancing GPU vs. CPU and then cores thrown in the mix causes all sorts of issues. Yea, GPU drives are not in control of the gaming studios -- but coming back to the point 4-cored dies are just fine for now.
 
Status
Not open for further replies.