Gulftown VS Bloomfield, IPC

[BTRY B 529th FA BN]

Did the ipc, no not insane clown possee, IPC improve from Bloomfield to Gulftown? If so how much?

 

[pantsaregood]

Not really. Gulftown is a variant of Westmere, which is a die shrink of Nehalem. Bloomfield is a variant of Nehalem.

AES instructions were introduced in Westmere, though. In applications that use them, you may see a significant performance boost.

The biggest advantage of Westmere is that it runs cooler and can usually clock higher. Nehalem usually runs pretty hot (had a friend whose i7-920 ran at 55C idle on Hyper 212+ at stock voltage), but Westmere usually runs around 65C under extreme load. That i7 I mentioned would push 90C, even after being reseated several times.

Nehalem usually tops out at 3.8-4.2 GHz. Westmere can often hit 4.2-4.5.

 

[inf64]

Actually it did slightly,but mostly due to 50% larger L3. Core changes were minimal.

 

[Lepton87]

http://www.tomshardware.com/reviews/processor-architecture-benchmark,2974-9.html


It always seemed strange to me that despite 50% larger L3 cache IPC didn't really improve.

Well, it did, but that increase is really diminutive.

 

[Makaveli]

Not really. Gulftown is a variant of Westmere, which is a die shrink of Nehalem. Bloomfield is a variant of Nehalem.

AES instructions were introduced in Westmere, though. In applications that use them, you may see a significant performance boost.

The biggest advantage of Westmere is that it runs cooler and can usually clock higher. Nehalem usually runs pretty hot (had a friend whose i7-920 ran at 55C idle on Hyper 212+ at stock voltage), but Westmere usually runs around 65C under extreme load. That i7 I mentioned would push 90C, even after being reseated several times.

Nehalem usually tops out at 3.8-4.2 GHz. Westmere can often hit 4.2-4.5.

I agreed with this.

i'm hitting the same 4 Ghz overclock with lower voltage on this chip than on my 920 D0. And idle temps are lower full load temps are similar even with the 2 extra cores the 32nm and lower voltage makes a difference.

And as mentioned before if you do anything that uses AES (truecrypt)
its a big boost in performance compared to Bloomfield which has no AES

P.s I always thought those Hyper 212+ were crap don't know why so many people bought them.

And with Summer in canada and temps in the 25c-30c outside this week running linx gives me temps in the 70-75c range if I put HT on its 75-80c at 4Ghz.

TjMax is also higher on this chip Core temp shows 101c think bloomfield was like 99c

 

[pantsaregood]

Hyper 212+ is a great cooler for $20. My 2500K maxes at about 55C full load at 1.4v with it. Usually it runs at around 25C idle.

 

[Makaveli]

Hyper 212+ is a great cooler for $20. My 2500K maxes at about 55C full load at 1.4v with it. Usually it runs at around 25C idle.

Ya for a 95Watt processor or lower I agree not on a 130Watt cpu tho!

The post I was referring to they used it on a 920 which is a 130watt cpu.

 

[pantsaregood]

It shouldn't have much trouble with a 130W CPU. It worked very well on a 125W Phenom X4 9750 at 2.8 GHz and on Core 2 Quads.

 

[mtnd3vil]

(had a friend whose i7-920 ran at 55C idle on Hyper 212+ at stock voltage),

...

That i7 I mentioned would push 90C, even after being reseated several times.

It sounds like it's having a ton of trouble.

 

[Makaveli]

It shouldn't have much trouble with a 130W CPU. It worked very well on a 125W Phenom X4 9750 at 2.8 GHz and on Core 2 Quads.

Were any of those cpus overclocked and not running at stock speed?

Big difference in a 920 overclocked with HT on vs a 2.8 Phenom X4 or C2Q in terms of heat output.

 

[pantsaregood]

The Phenom X4 was overclocked as far as it could go. This was a 65nm Phenom - it put out a pretty decent amount of heat. The stock speed was only 2.4 GHz, but getting it to 2.8 GHz required a voltage bump to around 1.45v. At 1.55v, it was still able to stay under 60C.

That did eventually blow up my motherboard, though.

 

[Makaveli]

That did eventually blow up my motherboard, though.

sorry to hear that but kinda funny still

I would have loved to see how much wattage it was pulling full load on that overclock. The temp isn't as important because AMD and intel have very different Tj max temps I think those amd chips were only rated for a 65c max temp where as all the i7 were rated for the high 90c so apple vs oranges if we look at temps only.

Curious what board were you running it on?

 

[pantsaregood]

Was some Biostar T-series AM2+ board. The CPU was at like 1.7v when it blew the MOSFETs, and it still wasn't hot enough to throttle. Also, the case I was using was an old Antec Sonata II - the airflow was pretty piss poor. At 1.7v, that Phenom was probably pulling some pretty damn high wattage. It was a suicide overclock, really. The CPU survived unscathed.

Also, it wasn't stable above 2.8 GHz at 1.55v or so. That thing was awful at overclocking. It would run 2.8 with voltage barely above stock, then just wouldn't to any further. No memory, CPU-NB, or HT tweaking would fix it.

Also I don't know if AMD officially declares TjMax values or if they compute them like Intel does. Intel CPUs throttle at their tjmax, but AMD CPUs tend to happily heat up considerably more before throttling. I've never damaged a CPU, even when my AMD CPUs were running above 65C.

 

[lamedude]

L3$ latency went up so it could be slower.