For a sample of the processor running at 2.4GHz, the results of the tests are qutie frankly AMAZING.
I really can't wait for this processor, and to see the new dual-cores with hyper-threading.
Amazing CPU. Serious power right there. Intel really are in a field on their own right now.
And to think AMD and Nvidia are still waiting for TSMC to improve their 40nm wields (from like 20%...) to release some new stuff =/
I don't see much amazing going on. It performs identically (within 5%) of the 2.4GHz Bloomfield in all of the tests except Cinebench and the synthetic Sandra suite. Probably because applications can't take advantage of twelve threads rather than eight.
This CPU will only be available as a $1000+ Extreme Edition or a server part anyway; probably less than half a percent of PC owners will have it. That size of market is the reason why AMD's six-core Istanbul isn't getting a desktop part.
All this proves is that there's no IPC increase from Nehalem to Westmere. Power consumption is down I admit.
And to think AMD and Nvidia are still waiting for TSMC to improve their 40nm wields (from like 20%...) to release some new stuff =/
40nm yields are fine, don't believe th rumours. AMD's 4770 is readily availible and their new DX11 stuff comes on September 10. The reason Nvidia have no DX11/40nm parts yet is engineering issues, not TSMC. Those parts are expected December at best.
Nice link SSJ12. Intel is really on the ball. Even though the majority of the benchmarks weren't optimized for six cores it still ate the Bloomfield's lunch. The bandwidth and arithmetic tests were off the charts.
| Darc Requiem said: Nice link SSJ12. Intel is really on the ball. Even though the majority of the benchmarks weren't optimized for six cores it still ate the Bloomfield's lunch. The bandwidth and arithmetic tests were off the charts. |
Yeah, synthetic. That doesn't translate into performance in real-world applications. When Core 2 transitioned from DDR2 to DDR3 bandwidth readings went up by over 50%, and real performance went up by 1-2% on average and 5% at best.
Optimising for six cores in those apps won't happen overnight. It's been 3 years since the first quad-core was released and only a minority of desktop applications and games take advantage of it. It makes sense on workstations and the server, not on the vast majority of PCs.
Plus if the launch clocks are 2.4GHz then it will be owned by the 3.33GHz existing Bloomfield. Any 32nm quads will run at higher clocks than 32nm six-cores and though not faster on paper will outrun them in any desktop app.
Soleron said:
Yeah, synthetic. That doesn't translate into performance in real-world applications. When Core 2 transitioned from DDR2 to DDR3 bandwidth readings went up by over 50%, and real performance went up by 1-2% on average and 5% at best. Optimising for six cores in those apps won't happen overnight. It's been 3 years since the first quad-core was released and only a minority of desktop applications and games take advantage of it. It makes sense on workstations and the server, not on the vast majority of PCs. Plus if the launch clocks are 2.4GHz then it will be owned by the 3.33GHz existing Bloomfield. Any 32nm quads will run at higher clocks than 32nm six-cores and though not faster on paper will outrun them in any desktop app. |
It was just an engineering sample vs engineering sample. The i7 940 beat the original sample by a point or two.
Launch is expected to be 3.2GHz at minimum. What I like is that Intel will keep the earlier i7 models, with 32nm cores, so they will be even more affordable. Plus since the higher end ones still use socket LGA 1366, motherboard costs will drop.
ssj12 said:
It was just an engineering sample vs engineering sample. The i7 940 beat the original sample by a point or two. Launch is expected to be 3.2GHz at minimum. What I like is that Intel will keep the earlier i7 models, with 32nm cores, so they will be even more affordable. Plus since the higher end ones still use socket LGA 1366, motherboard costs will drop. |
Where are you getting your figures from? There are no forecast speeds leaked yet, there will be no 32nm quads (45nm only through 2010), and the figures can't improve from sample to launch this time since there are no core-level changes from Nehalem to Westmere.
Soleron said:
Where are you getting your figures from? There are no forecast speeds leaked yet, there will be no 32nm quads (45nm only through 2010), and the figures can't improve from sample to launch this time since there are no core-level changes from Nehalem to Westmere. |
I can easily forecast speeds because the bloomfield sample they used is 2.4GHz and the Core i7 975 uses a bloomfield core. Also it would be really strange for Intel not have at least one of their flagship processors not be clocked over 3Ghz.
ssj12 said:
I can easily forecast speeds because the bloomfield sample they used is 2.4GHz and the Core i7 975 uses a bloomfield core. Also it would be really strange for Intel not have at least one of their flagship processors not be clocked over 3Ghz. |
Every time Intel or AMD have launched processors with more cores, the launch speeds have been lower than the flagship.
Pentium D launched at 3.2GHz when Pentium 4 was at 3.8GHz
Core 2 Quad launched at 2.66GHz when Core 2 Duo launched at 3.0GHz
AMD's 6-core Istanbul launched at 2.6GHz when their 4-core Shanghai was at 3.1GHz
I'll make a bet with you that the first Gulftowns will launch at 3.0GHz or under.
About Us |
Terms of Use |
Privacy Policy |
Advertise |
Staff |
Contact
Display As Desktop
Display As Mobile
© 2006-2025 VGChartz Ltd. All rights reserved.
Currently Online 
