I'm not saying anything about the validity of these results but generally speaking, i don't think i've ever seen a semiconductor product that was really late and offered great performance...
D:
A lot of websites have flash (which Apples hates due to it using a lot of resources). Even my 6950 2GB doesn't help despite it supposedly having Adobe Flash acceleration. I have a secondary rig that I often do tests on. It runs Core 2 Duo E6600 2.4ghz with OCZ 60GB and I stick my videocard in it. It cannot handle 50-60 tab browsing across 3 web-browsers. I can easily hit CPU usage of 100%, and that's without watching 1080P content on youtube. Also, if you are going to bring up video encoding as the "only CPU intensive program that 'most' people use", then you can't dismiss the idea that a lot of people who do video encoding are converting video to their tablet or smartphone. Under such usage, QuickSync is actually superior since it doesn't sacrifice the image quality and performs much faster.
Also, when most people refer to "single core performance" they are not discussing running programs from 1998 that run 1 thread. Most programs today use 2-4 threads and outside of Arma 3 I can't think of a single game that uses > 4 cores. So in other words, if you are running 1-4 threads (i.e., most real world programs), then single core performance is critical.
Consider this, when Fermi GTX470/480 arrived some 6 months late, they were still 10-15% faster than HD5850/HD5870, respectively, and delivered far superior tessellation performance. On top of that, NV shipped those cards with 2-3 free games in a bundle (Mafia 2, Metro 2033, Just Cause 2). However, they were still deemed to be a "failure" by a lot of people. Here we have BD that is 9 months late vs. SB - a CPU that so far looks like it is barely faster than the X6 1100T in multi-threaded apps and barely beats a $220 2500k (which is only $179.99 at MC btw). But once you compared 2500k @ 4.5+ ghz, then not even an overclocked FX-8150 will be able to beat it (at least based on these benches). That's nothing to say about its power consumption either. Honestly, this is shaping up to be a far greater failure than Fermi ever was. If you are going to be late by 9 months, you need to be either faster or cheaper or both.
There are quite a few examples. Opteron was late, Fermi was late, and although flawed, had excellent performance. But it's hard to say how late Bulldozer really is, the original design was supposedly **** canned, so how late is Bulldozer exactly? Compared to the competition, about 4 years I guess.I'm not saying anything about the validity of these results but generally speaking, i don't think i've ever seen a semiconductor product that was really late and offered great performance...
Because they clearly were. It was late, too hot, too noisy, and in no way significantly superior to what AMD had at the time. Them shipping them with games is simply a bonus, and the higher tessellation has so far been mostly theoretical and made very little difference in actual games. The GTX 480 was 15% faster than the HD 5870, but had absolutely horrible efficiency coupled with high noise and its price tag was too high. The GTX 470 was only 5% faster than the HD 5850 and costed significantly more as well, not to mention it also had the other drawbacks I mentioned earlier.
For the most part Bulldozer also seems to be a failure, but for different reasons.
Not forgetting these CPU tests.......Hmm...
http://lab501.ro/procesoare-chipseturi/amd-fx-8150-bulldozer-preview/7
For what it is worth,
(from that preview)
3dMark11 is pretty new. Is AVX in use here? :hmm:
the BIOSv 9905 doesn't have turbo core pstates so it is only 3.6GHzAnd wPrime.......
Does seem to correlate with that FX-8120 @4GHz (since FX-8150 turbos at 3.9GHz to 4.2GHz). :hmm:
Gross, I would never want $1000 CPUs
Agreed, his preview stinks to high heaven and has the nerve to conclude that everybody buys 1000$ cpus, what a mor**.
Go troll somewhere else, like AMDZone or something.
But with our short passing test suite, we had the most brutal awakening since test components and write reviews, pure and simple, it is difficult to find a scenario where the bulldozer is a competitive solution. Handbrake only approach the performance of i7 2600K and with lower price may be considered a smart choice, assuming you only x264 encoding, but what about heat dissipation and energy consumption, of the approximately 250 RON save will have to buy a cooler better and we will remain current for many bills. Or what of the lack of vigor in multithreaded applications less, or downright tragic performance in floating point calculations? How justified that in some benchmarks is considerably slower than its predecessor, the Phenom II? How explain that encourages overclocking AMD producer, but Bulldozer double their energy consumption and so much just by raising the frequency and voltage in modest limits?
wPrime it would probably just be easier to use an application that uses AVX
speaking of AVX look at these scores (AIDA 1.80 FPU Sin/Julia):
http://lab501.ro/procesoare-chipseturi/amd-fx-8150-bulldozer-preview/9
4 cores 2600K is 78% faster than "8 cores" FX8150, it was expected that Bulddozer will be a dog for AVX-256 code since its specs (*1) were disclosed more than one year ago but I was hoping for less real world difference
*1: http://forums.anandtech.com/showpost.php?p=32387549&postcount=74
It's always the OS, because the apps don't "know" the hardware and only interact with the OS.Who distributes the load, the OS or the application?
yeah, AMD had some sort of live streaming event on twitch.tv yesterday where they were showing off Bulldozer's capabilities, although it wasn't particularly encouraging because they were showing it off @ 7+GHz keeping it stable and cool by pouring liters and liters of LN2 on it throughout the presentation
so apparently BD is a viable solution as long as you can keep it under permanent LN2
FPU Sin/Julia
This benchmark measures the extended precision (also known as 80-bit) floating-point performance through the computation of a single frame of a modified "Julia" fractal. The code behind this benchmark method is written in Assembly, and it is extremely optimized for every popular AMD and Intel processor core variants by utilizing trigonometric and exponential x87 instructions.
Boy, that's a very rough conclusion from the Romanian site's Bulldozer preview.
excuse me but your links are broken, can you fix them? IIRC AIDA was the first to release an AVX-256 path right at Sandy Bridge launch
btw I don't expect Bulldozer to be so much slower for legacy 80-bit x87 code, it makes no sense
"The latest AIDA64 update further optimizes the previously introduced AVX-acccelerated benchmark suite"
"Further optimized 64-bit AVX-accelerated fractal and security benchmarks"
"Julia" is the fractal test, so it looks like they have replaced x87 with SSE2, then AVX a long time ago
A product that is utter [trash] getting concluded as such? Boy, that's a surprise.
The FPU Sin/Julia is x87
It's always the OS, because the apps don't "know" the hardware and only interact with the OS.