Hyper-Threading performance Test (on vs off)

[MODEL3]

http://ixbtlabs.com/articles3/...chspeed-2009-4-p1.html

However, we expected better results. Having seen a 10.2% performance gain with a quad-core processor, we actually expected at least 20% with a single core. Alas, it did not happen. When you already know test results, it makes no sense to speculate whether it's good or bad -- facts are neutral. We feel sorry for the dream: if HT efficiency had raised significantly when the number of cores is decreased, we might have hoped to see dual-core CPUs with HT support from Intel in the nearest future to be positioned as an alternative to cheap honest quad-core processors from AMD. Considering generally higher performance of the new core from Intel per MHz -- why not?

However, 10-11% advantage, provided by HT, won't be enough for the virtual quad-core processor to compete with a real one. So from the point of view of providing a real performance advantage, Intel has no objective incentive to promote Hyper-Threading to Low-End. HT does not provide a sufficient performance gain to help expensive (relative to dual-core processors from AMD) dual-core processors from Intel to stand up to cheap (relative to quad-core processors from Intel) quad-core processors from AMD. However, pointlessness from the practical point of view does not at all mean that omnipresent marketing specialists won't try to use this resource.

However, there is another conclusion suggested by results of these tests. Indeed, 11 "red cards" for performance drops in HT tests with four physical cores have magically turned into two cards in our tests with a single core. But this technology hasn't changed: because we use the same processor. So our conclusion is quite optimistic: perhaps the problem is not in the hardware optimization of Hyper-Threading. It's reasonable to assume that Windows Vista was so stupefied to detect eight processors that its task scheduler just couldn't cope with such an abyss of new opportunities. At least this explanation to so many HT failures with an 8-processor system (from the OS point of view) seems the most logical to us. So if we take this as a working hypothesis, the problem is in system software only, and the situation can be improved without changing the hardware (that is it's not Intel's fault).

 

[Phynaz]

Link?

 

[MODEL3]

oops, my bad, i will edit.

 

[RussianSensation]

I dont think Intel ever claimed that HT was going to replace extra cores. As is mentioned in the conclusion, you get about 10% of free performance with HT. But things get really interesting when some websites show reduced gaming performance with HT on:

http://www.pcgameshardware.com...CPUs-reviewed/Reviews/
Core i5 750 is actually faster than i7 860 with HT in GTA4 and Race Driver: Grid.

I am not sure how accurate these results are though.

 

[Idontcare]

Its a cool test matrix but they really should have applied the same testing methodology to a 2C/4T (2C w/HT) and a 4C/4T (4C w/o HT...both Nehalem as well as PhII X4) test matrix if they really wanted to generate some meat to any argument they wanted to make regarding the efficacy of Clarkdale competing with "true" quadcores in 2010.

Interesting to see the corner-case where super-linear scaling crop up with Grand Theft Auto 4, no doubt attributable to the aggregate L1/L2 cache of two cores relieving a bottleneck imposed by that of a single core's worth of L1/L2 cache.

 

[imported_Lothar]

What is the relationship between Xbit labs and iXBT labs?
Are they affiliated, or is someone simply trying to do another "MikeRoweSoft" type impression?

 

[aigomorla]

hmmmm... im seeing more then 10% on occasions tho.

 

[Lonyo]

Originally posted by: aigomorla
hmmmm... im seeing more then 10% on occasions tho.

The average was only around 10%, however that's not really to do with HT, and more to do with running single threaded tests where a second core wouldn't do much either.
They should have looked at average increases in multithreaded applications etc and compared that rather than using single threaded apps as well.

The point still stands that HT is no replacement for actual extra cores, but I'm not sure it was ever advertised as such by Intel.
They are planning to use it in the low end to make up for a lack of cores, and will probably market it as a feature which will make up for extra cores, but at the end of the day, it's a nice boost although it won't always be as useful as actual extra cores, and AMD should be able to push against Intel in the low end with their cheap quad cores vs Intels 2C/4T chips when they eventually come out.

 

[Soulkeeper]

i've never been a fan of hyperthreading

 

[IntelUser2000]

Originally posted by: RussianSensation
I dont think Intel ever claimed that HT was going to replace extra cores. As is mentioned in the conclusion, you get about 10% of free performance with HT. But things get really interesting when some websites show reduced gaming performance with HT on:

http://www.pcgameshardware.com...CPUs-reviewed/Reviews/
Core i5 750 is actually faster than i7 860 with HT in GTA4 and Race Driver: Grid.

I am not sure how accurate these results are though.

Mostly the problem with losses in performance is due to the fact that the extra 4 threads aren't used by the games and being a logical thread, it uses up resources if improperly coded.

Interesting to see the corner-case where super-linear scaling crop up with Grand Theft Auto 4, no doubt attributable to the aggregate L1/L2 cache of two cores relieving a bottleneck imposed by that of a single core's worth of L1/L2 cache.

I'm not sure its that simple. Even the Hyperthreading benefits reach 50%, which is quite a lot.

I think 2C/4T vs 4C/4T comparison will be quite simple.

Assume low end Clarkdale at 3.2Ghz(20% clock speed advantage over i7 750). Add 12% with Hyperthreading and it should make up for extra 2 cores with Lynnfield. The difference might be even less as 40% advantage is from 1 to 2 cores, 2 to 4 could be lower.

 

[RussianSensation]

Originally posted by: IntelUser2000

Mostly the problem with losses in performance is due to the fact that the extra 4 threads aren't used by the games and being a logical thread, it uses up resources if improperly coded.

Makes sense. This would explain why a Core i5 750 would outperform a Core i7 860 with HT.

 

[DarkUltra]

This is an old thread, but I thought it proper to add this link. Then we have gone through all the arguments

P4 EE Single core with and without HT, an old tom's hardware video
http://www.youtube.com/watch?v=DliBXd5xvog

 

[SonicIce]

yo lets discuss 386 overclocking

 

[podspi]

This is an old thread, but I thought it proper to add this link. Then we have gone through all the arguments

P4 EE Single core with and without HT, an old tom's hardware video
http://www.youtube.com/watch?v=DliBXd5xvog

That definitely has nothing to do with the Pentium 4, I think you linked to the wrong video. Sweet beat though


I was actually looking at the performance increase/decrease with hyperthreading on a Pentium 4 yesterday. Turns out it was much better to turn HT off for my singlethreaded, (custom) app.

 

[Gundark]

Console emulators like Dolphin and PCSX2 performs slower with HT enabled. But, again, emulators are highly inefficient applications.

 

[VirtualLarry]

I dont think Intel ever claimed that HT was going to replace extra cores.

But they market it as such, they position their dual-cores with HT against AMD's cheap quad-cores. Good to know that the real quad-cores are still faster.

 

[Soleron]

The article spent a lot of time talking about how Intel duals-with-HT will never be priced against AMD quads, yet within a few months Clarkdale was doing just that.

 

[Ben90]

I'm more curious as to how well HTT performance scales with clockspeed.

I've always viewed HTT as a higher level version of OOE. Since stalls take relatively longer the higher the clockspeed of the core, OOE effectively keeps the performance gains more linear when you start pumping clocks in respect to an in order processor.

HTT isn't quite as "free" as OOE however when it comes to burden on the cores. When overclocked 50% above stock, that extra cache HTT needs could become quite a burden, creating more misses. However HTT eases the impact of misses as well.

Does the delta between HTT On and OFF widen with increased clockspeed, or does it shrink?

 

[john3850]

Would the results be different on win 7 .

iirc intel was forced to use ht do to the out of order instruction and or the
cache hit and misses but again I dont remember.

 

[Martimus]

Does anyone else think the concluding paragraph is completely irrational? It sure as heck doesn't jive with the actual results, nor does it have any supporting evidence. It just seems an unprofessional writeup all around to me, and I can say I have never heard of ixbtlabs before. The link makes it look like a spam site since it is so close to xbitlabs.

 

[Rifter]

This is an old thread, but I thought it proper to add this link. Then we have gone through all the arguments

P4 EE Single core with and without HT, an old tom's hardware video
http://www.youtube.com/watch?v=DliBXd5xvog

wat

 

[Voo]

How can you write a whole article about HT comparisons and NOWHERE mention what OS was used? The difference between XP-Vista-Win7 not to mention UNIX OSes in their respective scheduler implementations and how they handle logical cores is rather big (XP ignores the difference completely..), which means one of the most important datapoints is missing.
Couldn't care less about a conclusion if the author obviously isn't savvy enough to notice that he's forgetting one of the most important facts to interpret his data ~

 

[randname]

It just seems an unprofessional writeup all around to me, and I can say I have never heard of ixbtlabs before. The link makes it look like a spam site since it is so close to xbitlabs.

Lol. If they are a spam site, they are a fantastic one. First site I've seen actually benchmark more than media/game apps that people might actually care about on an E-350 vs Atom comparison. (http://ixbtlabs.com/articles3/cpu/amd-zacate-e-350-method-2011-preview-p5.html - bottom of page).

Anandtech's own coverage was mostly Games+Cinebench+x264. That site also benched multiple webbrowsers and office apps. I'd argue webbrowsers matter most for a netbook & the E-350 crushed the Atom, something no other review sites tested that I know of.

 

[AdamK47]

iXBT has been around for a long time. I've been going there for many years. They are better known in Russia. The US site, iXBT Labs has excellent reviews.

 

[classy]

iXBT has been around for a long time. I've been going there for many years. They are better known in Russia. The US site, iXBT Labs has excellent reviews.

Yea they are a real good site. And as you pointed out they have been around a long time.