Printable View
Uh yeah forgot about the 5570... :DQuote:
Originally Posted by ajaidev
Well it looses 30% on no aa and 40% on 4xAA, so yes thats quite some perforamnce that they give away... they easily could have gone with 320sps (like the 4670) and nearly would have got the same performance...
I still dont get it why they go with 480SPs...
In case anyone still wonders, this is not with the integrated graphics obviously.Quote:
Originally Posted by 570091D
I'd like to see a comparison of actual visuals, seeing as Intel's graphics didn't only used to give low framerates, but looked a lot uglier than nVidia and AMD's offerings too. And, will it support OpenCL? Or would Intel stifle this, just to try and bar AMD's GPU acceleration attempts?
Did you read the article? :confused:Quote:
Originally Posted by Calmatory
The SB sample did not have Turbo enabled.
The comparison parts (i7 880, for example) did.
The improvement was IN SPITE OF Sandy having no turbo, against a turboing Westmere. And it was in spite of this SB sku having 1.5MB of L3 per core instead of 2MB.
Bobcat IS the new low power CPU core from AMD-the new uarchitecture not related to K8,10h or Bulldozer(to which it resembles somewhat). This will be the x86 core inside the Ontario APU(Accelerated Proc. Unit). Ontario will have a low power GPU,the 80Stream Processors DX11 Evergreen-class part. Ontario is low power netbooks,mainstream notebook territory. Bobcat CPU core will be similar in speed with Propus/Regor or what you call Athlon II X2/X4 today.Quote:
Originally Posted by ~CS~
Llano is a mainstream desktop/notebok APU.It will have and improved "old" CPU core,based on Shanghai,but done on 32nm with some core level improvements(this means it should be somewhat faster at the same clock);there will be 2x more L2 per core but no L3. The GPu inside is speculated to have 480SP(and a lower count one),a Radeon 5650/5670 class.There will be a brand new integrated memory controller designed for this hybrid part,to ensure both CPU and GPU have a good balanced use of memory resources. Llano will be a 2 core /4 core part with probably 2 GPU models inside(one with higher and one with lower spec. GPU).
Well you get the IPC improvement per core + they said the gona improve HT, and by the looks they did. ;)Quote:
Originally Posted by Calmatory
Well the 5570 has 16 TAUs and 8 ROPs that can be also the reason's for its bad performance. There can be two reasons for keeping 480SP's one is for the added raw computing power needed for GPU processing like ATI stream and the second can be the presence of two GPU's one low power 2D/3D and one high power "80SP and 400SP" :shrug:Quote:
Originally Posted by Hornet331
No, because I am not interested in SB since I don't have enough money to get one. I just did a quick glance at the graphs.Quote:
Originally Posted by terrace215
Great. So one can't compare the CPUs by their base clockspeeds anymore since they twiddle dynamically? It gets cryptic.Quote:
The SB sample did not have Turbo enabled.
The comparison parts (i7 880, for example) did.
The improvement was IN SPITE OF Sandy having no turbo, against a turboing Westmere. And it was in spite of this SB sku having 1.5MB of L3 per core instead of 2MB.
Nice preview, but I'm tired of this Anand/Intel politic BS.
Quote:
Originally Posted by ajaidev
Both options are meh... Stream is nice to have, but still a few years off getting any real attention, maybe next cpu generation and even 320 would offer more then enough computing power.
The scond option is also not really plausible, why wast die space for 2 gpus when you can powergate the SPs, if the rework the gpu a bit, im sure it would be possible in a year or so.
I'd love to be tired of Anand/AMD politic BS, oh wait, AMD doesn't share anything with anyone. Their heads are stuck too far up their collective :banana::banana::banana::banana::banana::banana::b anana:s! :rofl:Quote:
Originally Posted by G.Foyle
wow, nice results, and thats without turbo
gamers will love that i5 2400 :)
that coupled with a HD6770 should make an awesome gaming rig for 2011...
What do you want them to share?Quote:
Originally Posted by freeloader
What exactly do you want AMD to share with Anand at this point? Other than Bobcat, I doubt anything is in anywhere near good enough shape to let it be "reviewed" this way at this time. (EDIT: lol, gazer)Quote:
Originally Posted by freeloader
It appears from the charts in the review that Intel made improvements to boost multi-threaded workloads apart from HT. The boost from the 760 indicates that well-threaded code should get a nice boost in performance, while enabling HT seems to have a lesser effect based on workload. Seems they are trying to address some of the shortcomings with the previous implementation of SMT, such that well-threaded code will still get a boost in performance and (likely) take less of a hit with HT enabled...
You can make a fairly good correction if you know the turbo bins, or Anand could have rerun some of the earlier results having disabled the turbo on the older parts. OTOH, IPC is a tad esoteric to most people, who just want to know how the part is going to perform, and, well, the parts will run with turbo.Quote:
Originally Posted by Calmatory
So, instead, you could adjust the SB results up. Say you're looking at a single threaded test. It will turbo 22.5% higher. (3.1 --> 3.8) So pick a scaling factor... say 85% for single-thread cinebench, and add (.85 * 22.5% = ) ~19% to the SB score.
Is it too hard for them to share one benchmark running anything? Simply stating that it's in a BETA stage and they wouldn't even have to reveal the clock speed. Simply state it's between 1ghz and 2ghz or 2ghz and 3ghz, etc...It's funny how Intel can share stuff within 6 months of a launch, but AMD can not. I'm not going to rant about it, it just seems stupid on AMD's part to crap on it's customers. That's my opinion, nothing more.Quote:
Originally Posted by terrace215
Well at least we know now francois wasn't exaggerating when he said we get a C2 like performancejump.Quote:
Originally Posted by terrace215
AMD is not within 6 months of a launch... maybe not even 12 months :p:Quote:
Originally Posted by freeloader
Sorry, that was too easy.
This is Intel Haifa we're talking about, they always Conroe the market with each of their new uarchs.Quote:
Originally Posted by Hornet331
Says who? They apparently have working silicon now. More than likely under one year.Quote:
Originally Posted by STaRGaZeR
those results get even prettier if we consider SB had its turbo mode off :D
http://images.anandtech.com/graphs/s...2107/24403.png
i7 880 could be running at 3.6ghz or even 3.73ghz at this test, as it beat the core i7 980X (turbo @ 3.46ghz/3.6ghz)...
i5 2400 was 3.1ghz all the time...
now do the math...
lol, they were trying to compare it to 980x in some benchmarks..
game performance was disappointing so far, hope they iron out that on retail..
You think AMD is going from first silicon on a BRAND NEW microarchitecture on a brand new process to launched parts in *under* 1 year? They would have to be extraordinarily lucky.Quote:
Originally Posted by freeloader
And what makes you think they want to give A0 silicon to a review site??!! These things typically run at low clocks, major functionality can be bugged, etc., etc. It would be a PR disaster.
Using 85% clock scaling, that's about 27% better IPC on Cinebench R10. Think what the AVX version will do. ;)Quote:
Originally Posted by -Sweeper_
Speaking of which, given that the IPC gain *on average* looks to be in the ballpark of 20% once you correct the results for the non-turboing of the sample vs the competition....
where is saaya? :D
You didn't note the lack of turbo mode and clock difference?Quote:
Originally Posted by Loque
Gaming performance is limited by gpus... and it doesn't get better. :p:Quote:
Originally Posted by Loque