Thuban, expectations?

[None]

The results vary on every website it seems, it's hard to guess which one to believe or whether the performance varies by large amounts on each CPU.

Overclock 3D Game performance
Tomshardware Crysis
Tomshardware Dirt 2

It's definitely a lot of mixed results, and it's hard to believe some of the results from Overclock3D, in my opinion it would be better priced at $250 and $160 rather than $300 and $200.

[tlitd]

For what it's worth, Tom's SC2B testing session likely configures Cataclysm's hardware requirements :

• framework
• CPU clock speed scaling
• CPU core scaling
• GPU scaling (or lack thereof ;D)

Unsurprisingly, the lightly-threaded World of Warcraft won't make good use of the new Thuban. A decently clocked i3 530 plus a 5770 equivalent (4870/GTX260 216SP) should do mighty fine in Cataclysm.

[Cilraaz]

I'm not certain on Overclock3D. They've never seemed very professional to me. When one of your admins is using internet slang in comments, I'm instantly turned off. If you can't spell properly, how can I be sure that you're handling something like benchmarking properly?

I also dislike Tom's Hardware (even though they show similar results to AnandTech on this one). In the past, they've shown signs of potentially skewing results, presumably due to advertising revenue or even outright bribery (pretty much my personal opinion there, but I've seen others say the same).

The biggest problem with benchmarking is that you can skew the results very silently. If you want one manufacturer to come out ahead, it's very easy to do so without people instantly being able to say "there's the problem".

For instance, Overclock3D shows the i7 930 @ 4.0GHz being equal to or ahead of the 1090T @ 4.1GHz in synthetic benchmarks, and then they show the 1090T demolishing all other CPUs in the gaming benchmarks. While synthetic and gaming benchmarks aren't usually a 1:1 type of thing, there is almost always a correlation. In fact, I would even think that the 1090T should win by a larger margin in the synthetic benchmarks, as they can push all 6 cores to their max, whereas a game won't. On top of this, Overclock3D doesn't show the resolution they used for their gaming benchmarks. That's one possible skew point.

Another thing about Overclock3D is their quote regarding their Crysis Warhead test:

Crysis Warhead responds as it always does to "the more the merrier". The major shock is that our average framerate shoes the overclocked i7 930 just edging ahead, yet the 1090T has triple the minimum framerate.

Despite what the graph shows the 1090T was a vastly smoother experience. That's why we log the data, but personal opinion says otherwise.

Firstly, if the 1090T had a higher minimum framerate and similar average framerate (as shown in their graph), then it would be a smoother experience. I'm not quite sure why it's "despite what the graph shows". They also have no data (to be fair, many sites don't) on how long/often framerate was below a certain point. The i7 930's framerate may have dropped to its minimum for a fraction of a second, setting a lower low, but maintained a better framerate otherwise, which would show why the average framerate was higher for the i7 930.

Secondly: "personal opinion says otherwise"? Are you not impartially testing these CPUs? Personal opinion leaves too much room for fanboy-ism to kick in and skew the results. That's why everything is graphed and laid out in numbers.

Finally, the fact that they're testing i7 930s vs the new 6-core AMDs is a bit of bias in and of itself. Where is the Intel 6-core CPU to show a true competition?

I know everything I'm saying makes me seem like an Intel fanboy, but I'm really not. I'm an enthusiast (on a minimal budget, lol). I enjoy the market. Whether Intel or AMD is on top is of no matter to me. I don't own stock in either, and so long as they're competing, prices come down for the consumer. Right now, I truthfully believe that Intel is ahead at pretty much all levels of CPU in performance (and most levels in price-per-performance). If AMD can make a comeback, more power to them. If not, hey, I still love their ATi video cards! Poor nVidia.

TLR - Overclock3D seems biased or perhaps just unprofessional/stupid. Tom's Hardware lost their trust in my eyes a long time ago.

[None]

I defiitely agree with the biased reviews from Toms and OC3D, I was actually banned from Tomshardware forums for bashing Cyberpower and Palit :

http://www.tweaktown.com/reviews/325...pu/index8.html

I do however trust Tweaktown far more than I trust either Toms or OC3D, and their review seems pretty normal and what would be expected of the Phenom II x6 CPUs. And based on their benchmarks I still think it would have been a better buy at $260 for the 1090T and around $160 - 180 for the 1055T

The i5 750 is a total knockout at 4.3ghz

Also another website I'd trust is Guru3D, though their results seem inconsistant with Tweaktown.

[Cilraaz]

I'm kind of curious how they hit 4.3GHz on the i5 750. It was either one of the best chips of any batch or they're overvolting beyond Intel's specs and it's going to last them another 6 months.

There's also something screwy with their results. Looking at their test, I don't doubt their methods (I've actually trusted Tweaktown for a while, myself), but for the overclocked X6 to beat everything in Modern Warfare 2 and Bad Company 2, and then turn around and lose to the stock i5 750 in Far Cry 2?!?! There's something screwy going on with that chip. It is new though, so perhaps it's certain programs conflicting with chipset drivers or something. It's tough to say, considering the lack of consistancy across the board.

[tlitd]

the biased reviews from Toms and OC3D (I)

Guru3D (II)

I. Oh, "anomalies" aren't specific only to the aforementioned reviewers. However, an experienced reader should easily spot a bias/dubious result by a. looking solely at the numbers they offer for results, and b. comparison.
II. Speaking of which, the conclusion of G3D on Thuban has a notable potential of inducing giggles : "But seriously, it's time for AMD to stop following Intel, and start to lead.".

[Cilraaz]

I fail to see where that statement is inaccurate though. For a time, AMD was being just as innovative as Intel (think K6 era). They innovated with the integrated memory controller, which took Intel forever to impliment. They innovated with MMX technology, where Intel only had SSE (or was it the other way around where AMD introduced SSE? In either case, it was an innovation). Now, Intel comes out with something and then AMD comes out with an "equal" product. AMD included "Turbo Core" with their X6 line, which is identical to Intel's Turbo Boost. As far as AMD's recent releases go, there's zero innovation.

[Enola]

Intel launched MMX, AMD launched 3Dnow shortly after, SSE its from Intel. The real evolution with amd was when they launched much after that the HT technology with the first 32/64bit processors, it was then for that short period of time that AMD hold the performance crown.
I had one of the first processors to come with mmx instructions, and my first amd was a K6-2 400(3dnow first model) it was also this generation that give the fame of being hot to amd, they weren't all that hot though, specially when compared with what Intel would do in future with the Prescott, It was a very good processor (k6-2) think its thanks to it that I used amd's from that moment one, till now, but no illusions what amd brings to the table nowadays is bang for the buck, the performance crown is securely owned by intel.

[Cilraaz]

3DNow. That's what I was thinking. Man, it's too many years ago, lol. I still think their biggest innovation was the IMC, though. When you can put the memory controller directly on the CPU and take it away from the northbridge essentially leaving it as just an information hub, you not only make the northbridge run cooler, but also lower your memory latency by a good bit. I'm really surprised it too Intel so long to follow suit on that one.

[None]

AMD's take on hyperthreading sounds great from what I've read. I have very high expectations for Bulldozer.

A bit early, but... Expectations for Bulldozer? :P

[Enola]

3DNow. That's what I was thinking. Man, it's too many years ago, lol. I still think their biggest innovation was the IMC, though. When you can put the memory controller directly on the CPU and take it away from the northbridge essentially leaving it as just an information hub, you not only make the northbridge run cooler, but also lower your memory latency by a good bit. I'm really surprised it too Intel so long to follow suit on that one.

Indeed, I agree it was the HT I mentioned Hyper Transport, think Intel problem was a bit due to patent issues bot not sure can be wrong.

[Cilraaz]

Indeed, I agree it was the HT I mentioned Hyper Transport, think Intel problem was a bit due to patent issues bot not sure can be wrong.

Ah, good call on the patent issues. And, for the record, my brain automatically translated HT into HyperThreading... and for some reason didn't have a second thought about a K6-2 using HyperThreading.

[Enola]

Lol, yeah the HT thing its usual , the K6-2 didn't have it, it was introduced much after, I think it was with socket 754 (1st's amd64).
Would love to see some performance jewels from amd, but not too confident, competing even at the lowest lev with Intel isnt easy, Intel is just massive the Intel business volume completely dwarfs amd, and that obviously impacts development a lot.

[chaud]

Disappointing, but hopefully 32nm will be more interesting.