At the present time I think the AMD Phenom II X6 1055T 95W CPU offers the best value for money:

In terms of crunching performance it matches an i7-930 (which costs much more), the 1055T uses much less power (35W less), and totally outperforms anything Intel has at the same price (such as the i5-750).

Can you link to some testing or project results that prove this out?
I thought the i7-930 was still much better than 1055T for crunching.

Hope it's OK if I jump in here. Performance depends greatly on the project. On many projects The i7 will out-crunch the Phenoms on a core to core basis. On some projects AMD is king, for instance Yoyo where a $1000 i7-980 Extreme 6/12 core can barely outdo my $90 Athlon II 620 that doesn't even have L3 cache:

http://www.rechenkraft.net/yoyo/top_hosts.php

Notice the i7 920 and i7 930 CPUs are running behind a pile of AMDs. In other projects the opposite is true, just depends on the code.

Jumping in is what forums are all about <grin>
~80% of the performance for ~10% of the price ... nice, thanks for the link.

---

Thanks - Steve

Jumping in is what forums are all about <grin>
~80% of the performance for ~10% of the price ... nice, thanks for the link.

Depending on the HardWare you've available. Mobo's are very important, as
they form the 'base' of a good computer.(ChipSet, I prefer one without graphics or sound, the X(38-58) type), QPI, DDR2/3,(Low-Latency)SATA600 & RAID 0;1;5, USB3.0. Low budget doesn't mean low RAC, c.q. valid production.
Just have 2 Q6600's, 1 with 2 ATI cards, EAH4850 & EAH5870,
1 Q6600 + GTS250 (only one with VISTA HP 32BIT) and a QX9650, OC'ed a bit (3.4GHz)+ a GTX470 & GTX480, not to forget my HP LT (T2400 Dual-Core).
All O.S.ses are Windows XP x86, except the above mentioned VISTA and 9650 runs
XP64.
Had a look at some stats and noticed, 12 hosts, haven't been able
to merge them, cause they all have 'overlapping lifespan', ofcoarse, if they're used, permanent, just different O.S. When you switched from Windows
to LINUX, tasks done 'under Windows'still have to be validated, while you're already running LINUX and crunching and switch back........

GPUgrid uses the GPU quite well, average load is 55% on the 480 and 65% on the 470. Einstein still has a too low load, so low, that you easily run 8
CUDA 3.1. tasks on a 480, takes more time cause 0.5CPU was used, instead of 1
core. So no 'real' speed-up, 4 tasks a card works, only results, in[i] messages
like please remove the app_info.xml, in order to receive more tasks.
SETI, also has a high load and 2 of them drives the 480 to 98% load.
So I can run 4 CUDA and 4 on CPU.

The fans of those cards can be quite noisy, due to the high airflow (and this rig uses ~600Watt, +/- 15%).
Compaired to my ATI host, which uses 100-50Watts, less and GPU's throtles back to 15%, if no load, more throtling then NVidia, 30% , if no load.

---

Knight Who Says Ni N!

Can you link to some testing or project results that prove this out?
I thought the i7-930 was still much better than 1055T for crunching.

Hi Steve, Unfortunately I cannot even remember which forum I read it in, but I remember the cruncher said he was surprised to find that his 1055T slightly outperformed an i7-930 when crunching at stock.
No dubt an i7-930 would do more work for some projects and for some the 1055T would do more work, but having read some reviews I think it’s reasonable to say they are roughly the same in performance. However, when you consider the price difference and the running cost difference I would say overall the 1055T would be the CPU to get.

I think an i7-875 has a 95W TDP, and would slightly outperform the 1055T for most crunching projects while using slightly less power, but the purchase cost is far too high (around £100 more expensive). As crunching is always horses for courses, all we can do is be generic and vague when talking about CPU crunching performance, unless we list pages of performance data for various task types.

Here are two review links for anyone that is interested,
http://www.legitreviews.com/article/1289/5/
http://www.hexus.net/content/item.php?item=26122&page=4

No dubt an i7-930 would do more work for some projects and for some the 1055T would do more work, but having read some reviews I think it’s reasonable to say they are roughly the same in performance. However, when you consider the price difference and the running cost difference I would say overall the 1055T would be the CPU to get.

I think an i7-875 has a 95W TDP, and would slightly outperform the 1055T for most crunching projects while using slightly less power, but the purchase cost is far too high (around £100 more expensive). As crunching is always horses for courses, all we can do is be generic and vague when talking about CPU crunching performance, unless we list pages of performance data for various task types.

Dealing with estimates & averages is useful, as it gives a rough picture of what is being used without the need to be promoted or adapted. CUDA is still not used by many even though it has a great potential. So if the 1055T is roughly the same for a lesser price of the i7-875 & the power consumption is also roughly the same. It means that BOINC projects are using the strength of what AMD has to offer over the strength of what Intel has to offer on average.

---

Depending on the HardWare you've available...

Hello Fred. I think the way you configure what you have is also important.

These are two tasks you ran (XPx64, QX9650 @ 3.4GHz):
3098164 1963128 7 Oct 2010 3:10:40 UTC 7 Oct 2010 9:30:33 UTC Completed and validated 9,690.56 1,506.05 4,535.61 6,803.41 ACEMD2: GPU molecular dynamics v6.11 (cuda31)
3096955 1955180 6 Oct 2010 22:57:57 UTC 7 Oct 2010 6:53:23 UTC Completed and validated 13,270.66 793.42 6,016.70 9,025.06 ACEMD2: GPU molecular dynamics v6.11 (cuda31)

And these are two tasks I ran (XPx86, i7-920 @ 2.66GHz Turbo Off, 2 GTX470’s at 715MHz):
3098941 1963560 7 Oct 2010 6:20:54 UTC 7 Oct 2010 11:46:04 UTC Completed and validated 8,381.19 8,389.64 4,535.61 6,803.41 ACEMD2: GPU molecular dynamics v6.11 (cuda31)
3098507 1955896 7 Oct 2010 5:00:13 UTC 7 Oct 2010 11:24:10 UTC Completed and validated 10,311.94 10,319.92 6,016.70 9,025.06 ACEMD2: GPU molecular dynamics v6.11 (cuda31)

Your first task, that ran on your GTX480, is the same type as the first task I ran.
My task ran 15.6% faster on my GTX470.

Your second task that ran on your GTX470 is the same type as the second task I ran.
My task ran 28.7% faster.

While I just overclocked my GPU you OC'd your CPU. But the main reason for the difference is that I leave at least 2 threads free and use swan_sync=0.

Not sure there is much point trying to comparing a 130W CPU, that is overclocked (about 150W @ 3.4GHz) in your dual Fermi rig to either a 95W or 105W Q6600 rig with two different generation ATI cards. The i7-920 however has a TDP of 130W. My i7 rig typically uses just under 500W and my GPU utilization is usually very high. When I have about 83% GPU utilization the power is between 480W and 490W, when it is 98% is can go up to 520W, and when I get the odd slow task it drops to about 470W.

- Agreed Liveonc, for crunching here the GPU is key. While the CPU can make a big difference, any top CPU will do the trick; it matters little if it is AMD or Intel. So the less expensive and more power efficient CPUs are more attractive.