Can't you just put two 4-gig sticks into SB-E? That way you have 8 gigs and room to expand for the future.
Quad-channel memory costs significantly more than dual-channel and brings with it no real-world performance improvements
Almost no performance improvements
ROFL, I hope you are joking here, because if not then you really have no idea what you are talking about. 4 sticks of 2GB RAM = quad channel.....and those are very cheap now. (Quad channel = 2 Dual channel kits)
ROFL, I hope you are joking here, because if not then you really have no idea what you are talking about. 4 sticks of 2GB RAM = quad channel.....and those are very cheap now. (Quad channel = 2 Dual channel kits)
All your assumptions on dual vs tri channel memory are based on tests done with quad core Nehalam CPUs of 2008. And it is a safe bet that quad core Intel CPUs do not need the tri channel bandwidth. Intel has said before that they estimate 1 channel per 2 cores with HT. So, once 8 cores CPUs start being introduced, then we will see that dual channel memory just does not feed fast enough. Wait and see. Quad chanel memory will be required by the time Haswell comes out.
Update here >> Update: AMD starts shipping Bulldozers, but only server variety. Interesting quote here.....
There were rumours abounding of Bulldozer die issues reaching high clock speeds, but the lower clocked dual die version, where the total TDP needs to be managed, would be spared that problem. Therefore, no wonder this flavour comes out first. Let's hope we see the single die versions with faster clocks later this month as well - AMD needs to act as the Socket 2011 Intel Sandy Bridge E launch is coming closer as well...
ROFL, I hope you are joking here, because if not then you really have no idea what you are talking about. 4 sticks of 2GB RAM = quad channel.....and those are very cheap now. (Quad channel = 2 Dual channel kits)
Yes, nice quote. Pollack's rule at play. To double the cores (and other logic, I/O) inside the same TDP envelope, voltage has to be reduced and frequency scaled down (to roughly 70%).Update here >> Update: AMD starts shipping Bulldozers, but only server variety. Interesting quote here.....
Based on this I think we can safely say that Bulldozer is a failure. IPC can be relatively poor as long as you can ramp the speed up. Or speed can suck if your IPC is great. But poor IPC and poor speeds make a poor CPU. Since most apps (at least those that most people care about) can only take advantage of 3-4 cores, more cores only do so much.
Well, if BD has improved power consumption significantly (perhaps even better than SB) and it offers more cores for servers, it can gain market share in the 2 growing segments: mobile and server space. It may not be a slam dunk for us desktop guys, but those other 2 markets are prob. more important for actually increasing revenues and profitability given the demand in those segments. :awe:
Sep 07, 2011 (MARKETWIRE via COMTEX) -- Today, AMD AMD +6.30% announced revenue shipments of the first processors based on its new x86 "Bulldozer" architecture. Initial production of the world's first 16-core x86 processor, codenamed "Interlagos," began in August and shipping to customers is already underway. Compatible with existing AMD Opteron(TM) 6100 Series platforms and infrastructure, "Interlagos" is expected to launch and be available in partner systems in the fourth quarter of this year. Many of the initial shipments have been earmarked for large custom supercomputer installations that are now underway.
"This is a monumental moment for the industry as this first 'Bulldozer' core represents the beginning of unprecedented performance scaling for x86 CPUs," said Rick Bergman, senior vice president and general manager, AMD Products Group. "The flexible new 'Bulldozer' architecture will give Web and datacenter customers the scalability they need to handle emerging cloud and virtualization workloads." - source
The part "but with something else" and those cache issues, could be related to this (happen to stumble upon this from SI AMD board) >> Looks like Linus isn't happy with a fix......Hans De Vries claims that the Prefetchers are also not enabled on the BD ES chip.
http://semiaccurate.com/forums/showpost.php?p=132211&postcount=788
Looks like Linus isn't happy with a fix AMD proposed for a cache
aliasing problem that hurts Linux performance on Bulldozer
On another forum someone claimed this aliasing bug drops BD performance by 3%
That FX-4170 is kinda worrying, I must say. If AMD wants it to compete with the i5 2500K, it could mean they need 27% higher clock speeds to reach the same performance. In turn, that means only an ~5% IPC improvement from Llano, which is total crap (unless they can get to 5GHz on average, which I doubt seeing how Phenom II X4 vs Core 2 Quad 45nm were in OCing). Or, it could also mean they want to release it at the fastest clock speeds possible so Intel has to lower prices and release new models and they want to take the market again.
In any case, it's highly probable that the Eight-Core CPUs will be the best in multi-threaded.
That FX-4170 is kinda worrying, I must say. If AMD wants it to compete with the i5 2500K, it could mean they need 27% higher clock speeds to reach the same performance.
In any case, it's highly probable that the Eight-Core CPUs will be the best in multi-threaded.
But they are positioning FX-6100 series vs. 2500k. There is no way FX-4170 will be able to compete with 2500k. I am pretty sure FX-4170 will be aimed against lower end i3-2120/2130 CPUs to take over X4 980 positioning. FX-4170 would be a great alternative to 2C/4T i3s.
Earlier....The part "but with something else" and those cache issues, could be related to this (happen to stumble upon this from SI AMD board) >> Looks like Linus isn't happy with a fix......
But they are positioning FX-6100 series vs. 2500k. There is no way FX-4170 will be able to compete with 2500k. I am pretty sure FX-4170 will be aimed against lower end i3-2120/2130 CPUs to take over X4 980 positioning. FX-4170 would be a great alternative to 2C/4T i3s.
They wouldn't need to. The Phenom II X4 965 already competes nicely with the Core i3 2120, and the 970 with the 2130. But then again, if you remember earlier news, AMD will be releasing an A8-3870 Black Edition. The A8-3850 is already a good match for the Core i3 2100, so I can see it being a good fight. AMD is supposed to introduce some IGP-less Llanos to make all their products 32nm at last.
Llano is at least 50% slower in IPC vs. SB. Intel G840 2 cores 2.8ghz is almost as fast as a 4 Core Llano A8-3650 2.6ghz!! 2500k 3.3ghz is 66% faster than A8-3850 2.9ghz. Even if you adjust for clock speeds, SB is again at least 50% faster.
This again. Testing in multi-threaded applications skews the results. Again, 39% higher IPC from Sandy Bridge in comparison to Deneb/Thuban. In comparison to Llano, around 35%.
AMD is definitely well behind in this metric, though. They need all the IPC increase they can get.
8 core vs. 4 core? Obviously. An overclocked X6 1100T can't even beat an overclocked 2500k in rendering, video encoding, encrypting despite having 2 more cores. So if you add 2 more cores on top of the X6, even with the same IPC of Phenom II (worst case scenario), it will easily beat Intel's 4 core SB in multi-threaded apps.
A Phenom II X6 1055T will overclock to 4GHz on average and it costs $150. A Core i5 2500K will overclock to 4.5GHz on average and it costs $220. We know that in encoding, 3D rendering, and content creation it matches or beats slightly the Core i5 both at the same clock speed. Given that, it should be 10-15% slower both OCed to their average on multi-threaded applications. Not bad considering it costs 45% less, if you ask me.
But BD isn't just supposed to be competitive for the next 6 months. This architecture is supposed to be 2-3 years forward looking. So what's going to happen once Intel eventually launches 6 core Sandy/IB processors at $300 in 2-3 years? Is AMD going to give us a 10-12 core CPU then? If they keep neglecting IPC, eventually they'll be relegated to <Sub $200 and lower price levels as IB brings way higher clock speeds. Doesn't sound promising.
They'll continue trading multi-threaded performance for single-threaded, then. I wouldn't call Intel a winner simply because they have higher single-threaded performance. If AMD delivers much higher multi-threaded performance I'd rather go that route, as long as IPC is increased to near Nehalem levels.
I don't care who's Elmer Phud (although he is the Moderator of SI AMD board), just interested in the information he pointed out (in the link).You're quoting and taking seriously Elmer Phud? :thumbsdown::thumbsdown: i forgive you since you obviously dont know who he is.
That FX-4170 is kinda worrying, I must say. If AMD wants it to compete with the i5 2500K, it could mean they need 27% higher clock speeds to reach the same performance. In turn, that means only an ~5% IPC improvement from Llano, which is total crap (unless they can get to 5GHz on average, which I doubt seeing how Phenom II X4 vs Core 2 Quad 45nm were in OCing). Or, it could also mean they want to release it at the fastest clock speeds possible so Intel has to lower prices and release new models and they want to take the market again.
In any case, it's highly probable that the Eight-Core CPUs will be the best in multi-threaded.
Gosh I just wish they would have released the BD desktop so all this just goes away.
You do understand that a 2500K running cinebench scores 7.24 at least mine does . lots of guys here have close to the same setup so they can confirm
A 12core 12 thread AMD scores7.95 @ 2.6ghz that 12 cores 12 threads against 4