What makes more sense, i7 920 or i7 860?

[Mav451]

My concern is if they release a new rev of the i5/i7's to improve the on-die PCIe voltage control it makes buying a 1st rev Lynnfield kind of silly. I'm not trying to play that game.

Combined with far more mature i7 BIOS, I think I'll keep my focus on the i7 920. That said, the cheapest quad-core move for me remains a Q9550 or Q9650, from H users with upgraditis - get upgrading guys

 

[Megalomaniac]

Re read the Anandtech, the PC Perspective or bit-tech articles on the i7 860 vs i7 920

pretty much all recommend the 860 UNLESS you are going to get 4 GPUs! In which case the 920 is a better choice. (or upgrading to the i9)
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634&p=9
http://www.pcper.com/article.php?aid=776&type=expert&pid=11

They also mentioned that 3 cores have to be fully loaded with memory intensive applications in order to fully saturate the 2 DDR buses on the 860, which will rarely be done by anyone.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3634&p=7

overclocking seems to be a bit more up the 920s alley
http://www.bit-tech.net/hardware/cpus/2009/09/08/intel-core-i5-and-i7-lynnfield-cpu-review/3

The 860 is much better at using Turbo mode and will yield much better results due to newer technology. So any single or dual thread applications will benefit more from the 860, then from 920, especially with Win 7 installed.

The one really big bonus that the x58 boards have is the i9 chips.

If you have an extra $200 and plan on upgrading to the i9 for another $300 later, go with 1366 and the 920.

If you want to save on power, have snappier everyday performance (TurboBoost) and save some money, go with the 860.

 

[OpStar]

860 imo. Anand raves about turbo mode, and if you o/c with it enabled you'll end up getting 3.2Ghz on all cores and something like 4.1Ghz for single threaded apps when turbo is applied.

Not to mention a good o/c'ing P55 board like the UD4P is only $170 bucks.

Also, lol at all the people in this thread saying the 860 doesn't have HT.

 

[AMD T-type]

17 920 no questions asked, it is by far the better choice since the 17 9xx series seem to be able to pull 4ghz overclocks in there sleep

17?

 

[lollerskater69]

You have a lot of good sub 200 x58 choices.

 

[Jthm27]

Also, lol at all the people in this thread saying the 860 doesn't have HT.

yea whats up with that?

 

[Spook16]

I think another thing that people are overlooking with the 920 is pure maturity. It is a known good overclocker. Funny thing I'm not seeing on these reviews of the new chips, is actual overclocking. These new chips may be dud's for overclocking, don't be so anxious to go grab one today, or this week. Let someone else do this leg work.

920 on the other hand has been around for almost a year, and pretty good at hitting 4Ghz. Looking a the numbers for these reviews compared to a non-OC'd 920, I think the answer to this question slaps you in the face.... 920 all the way.

 

[vengence]

@ #2. Why do a few people keep acting as if the i7 860/870 don't have HT also? ALL i7s, including the 860 and 870, are HT-enabled. Last time I checked, the 920 was 4 cores/ 8 threads, and the 860 is... 4 cores, 8 threads.

I'm on crack

@ #4. Your statement regarding the work done at "full load", being 1/2, is rather ambiguous. I'm assuming you're basically saying: "with both processors at full load while running an application capable of supporting up to 8 threads, the 750 will be doing 1/2 the work of the 920". Please feel free to correct me if that's an incorrect assumption, because outside of that rather rare scenario for most users, I can't see where the statement that the 750 will only be capable of doing 1/2 the work at full load is justified. As we've already seen with benchmarks that are one threaded, the 750 stays relatively equal with the 920, and even in those situations where the applications are multi-threaded, it's not that far behind. The 920 will definitely be the better multi-threaded application processor, but it seems like it's mostly 920 owners who feel the need to try and diminish what the 750 can accomplish.

This statement was talking in relationship to heat. In most reviews they fully load all threads of a 920 and a 750 and then say hey! the 920 is hotter. Well no shit, it is doing twice the work. In the real world, most people will only load one or two threads, just like you have said. And in those cases, the temps and heat loads generated by the 920 will be the same as or less than the 750. They could be less because of the additional "on chip" "features" of the 750. This wasn't about dimensioning what the 750 can accomplish. It was about the "load temp" comparision between the 750 and the 920.

@ #8. While it's true that a $200+ board will be needed for i5 overclocks, for any users who are not near a MC or can't arrange to get a 920 fairly cheaply, you're still talking a $80-100 cost difference between a i5 OCing setup vs. an i7 OCing setup. Hardly an insignificant difference. That said, you were right before in that the i7 will be the better OC ove rall.

Cost is somewhat relative. 100$ when choosing a car is insignificant. But of course we aren't talking cars we are talking computers. And the difference of a 100$ compared to a 200-300$ CPU is huge (33-50%). In terms of the whole computer at say 800$, you are under 13%. A lot depends on exactly what you need the computer for though.

I'm still having trouble justifying the i5 for someone who JUST wants to game on a budget. I havn't run the numbers lately, but it would seem that a DDR2 C2Q or C2D system would be enough CPU power at a much lower cost. That money they will be much better off putting into a 4890 or a 275 (or a 5870 or 380 or something). At least until they get into using SLI or Crossfire.

 

[ToastyBread]

I'm on crack

This statement was talking in relationship to heat. In most reviews they fully load all threads of a 920 and a 750 and then say hey! the 920 is hotter. Well no shit, it is doing twice the work. In the real world, most people will only load one or two threads, just like you have said. And in those cases, the temps and heat loads generated by the 920 will be the same as or less than the 750. They could be less because of the additional "on chip" "features" of the 750. This wasn't about dimensioning what the 750 can accomplish. It was about the "load temp" comparision between the 750 and the 920.

Are you sure though that in the case of a single application or a couple applications, that the 920 will be cooler? For both idle and operating at full load, Lynnfield appears to usually be 20-30 watts cooler than a 920. Kyle's review had a chart about power consumption for idle, 1 thread and fully loaded, and it seemed to give a pretty good idea: http://hardocp.com/article/2009/09/07/intel_lynnfield_core_i5_i7_processors/6

I guess I just find it hard to believe that Bloomfield could be in any way more power efficient than Lynnfield in non-overclocking situations, given that Intel has had a year to further revise power management for Lynnfield. That's just from my limited understanding of it all though, so maybe there's something I'm missing.

I'm still having trouble justifying the i5 for someone who JUST wants to game on a budget. I havn't run the numbers lately, but it would seem that a DDR2 C2Q or C2D system would be enough CPU power at a much lower cost. That money they will be much better off putting into a 4890 or a 275 (or a 5870 or 380 or something). At least until they get into using SLI or Crossfire.

I think it all just depends on what they're looking for. For people who push their systems heavily, play the most demanding games, and have the money to regularly upgrade, I think the X58/1366 and the 920 is the best route to go.

But if the person is looking to do gaming on a budget, but wants something that will last them fairly well for a few years, and don't plan on upgrading at all, I think a 750 could serve them well. Since they're on a budget, I don't expect them to go quad SLI or anything, and P55 can handle non-quad SLI setups easily, if they even go that way. Otherwise, they'll likely be more GPU-limited than CPU-limited with a 750, so that's probably what I'd recommend.

 

[vengence]

Are you sure though that in the case of a single application or a couple applications, that the 920 will be cooler? For both idle and operating at full load, Lynnfield appears to usually be 20-30 watts cooler than a 920. Kyle's review had a chart about power consumption for idle, 1 thread and fully loaded, and it seemed to give a pretty good idea: http://hardocp.com/article/2009/09/07/intel_lynnfield_core_i5_i7_processors/6

Power at the wall and power to the CPU are different. I was talking about the CPU it's self and not he mobo. Those figures include the difference in mobos (like the NF200, something they even commented on) and the extra RAM. 3 Channel vs 2 Channel, you can't power the extra stick for free!

(Note: If you can power an extra stick of ram for free, please let me know and we'll be rich)

I guess I just find it hard to believe that Bloomfield could be in any way more power efficient than Lynnfield in non-overclocking situations, given that Intel has had a year to further revise power management for Lynnfield. That's just from my limited understanding of it all though, so maybe there's something I'm missing.

There is more on the die now. And as far as cooling (and thermally limited overclocking) how much heat the CPU puts out is imporant. More on the die = more heat to dissipate.

I think it all just depends on what they're looking for. For people who push their systems heavily, play the most demanding games, and have the money to regularly upgrade, I think the X58/1366 and the 920 is the best route to go.

I'll agree whole heartedly. An X58 920 would be what I would get if I was upgrading right now. Frankly, I REALLY wish I could get one of these at work. God knows the FEA I run could take full advantage of it.

But if the person is looking to do gaming on a budget, but wants something that will last them fairly well for a few years, and don't plan on upgrading at all, I think a 750 could serve them well. Since they're on a budget, I don't expect them to go quad SLI or anything, and P55 can handle non-quad SLI setups easily, if they even go that way. Otherwise, they'll likely be more GPU-limited than CPU-limited with a 750, so that's probably what I'd recommend.

Can you justify the price increase over a C2Q? I'm not exactly sure the price difference, but an extra 100$ can buy a LOT more graphics card.

 

[ToastyBread]

Power at the wall and power to the CPU are different. I was talking about the CPU it's self and not he mobo. Those figures include the difference in mobos (like the NF200, something they even commented on) and the extra RAM. 3 Channel vs 2 Channel, you can't power the extra stick for free!

Whether or not they're referencing power to the CPU specifically or power to the system as a whole, there's no denying that Lynnfield is a more power-efficient processor. It's rated at 95W TDP, vs. Bloomfield's 130W TDP. Even if you then factor in chipset power consumption and such, power consumption charts are still viable for use in comparisons, because currently you're limited to the X58 chipset for Socket 1366, and P55 for 1156. Maybe if/when other chipsets appear, you can then remove that aspect for power efficiency values, but people can't just say that 920, at 130W TDP, is more power-efficient than a 95W TDP. A single additional channel of memory isn't responsible for an extra 35W TDP on Bloomfield...

There is more on the die now. And as far as cooling (and thermally limited overclocking) how much heat the CPU puts out is imporant. More on the die = more heat to dissipate.

Even with more on the die, if the core design is consuming less power overall (as Lynnfield is), your thermal output will be down. Any increase in thermal output from the PCIe controller is likely easily overshadowed by the much larger reduction in power consumption of the processor itself. From my experience building systems and viewing heat output and power consumption, and unless someone can show otherwise, thermal output is very much related to power consumption of the chip. As power consumption goes up, thermal output goes up. When power consumption goes down, well...

I'll agree whole heartedly. An X58 920 would be what I would get if I was upgrading right now.

Well, as I said, I'd only recommend a X58 920 to someone who regularly upgrades and has the money to do so (thus, someone who'd likely jump on a i9 / 6-core system when it comes out), who actually needs the extra bandwidth that a third channel brings, or otherwise someone who wants to do quad SLI/Crossfire. Otherwise, if they're on a budget and looking for a basic system that can play most games (so a single GPU, etc.), I'd probably tell them to get a i5 750 and a P55 board. Right now, due to launch inflation, it's harder to recommend it, but in a month or two, when P55 boards are < $100 and the i5 750 is below $200? Oh that'll definitely be the recommendation, easily.

Can you justify the price increase over a C2Q? I'm not exactly sure the price difference, but an extra 100$ can buy a LOT more graphics card.

The only C2Q that can even come close in any way to the i5 750 is the C2Q Q9650, at what is currently a $110 price premium on Newegg over the i5 750. The C2Q's around $200, are no match in any way for it. Given that the cheapest C2Q is $150 on Newegg and gets destroyed by an i5 750, yeah, I can easily recommend the 750 for $50 more

 

[stainremover]

Power at the wall and power to the CPU are different. I was talking about the CPU it's self and not he mobo. Those figures include the difference in mobos (like the NF200, something they even commented on) and the extra RAM. 3 Channel vs 2 Channel, you can't power the extra stick for free!

(Note: If you can power an extra stick of ram for free, please let me know and we'll be rich)

There is more on the die now. And as far as cooling (and thermally limited overclocking) how much heat the CPU puts out is imporant. More on the die = more heat to dissipate.

I'll agree whole heartedly. An X58 920 would be what I would get if I was upgrading right now. Frankly, I REALLY wish I could get one of these at work. God knows the FEA I run could take full advantage of it.

Can you justify the price increase over a C2Q? I'm not exactly sure the price difference, but an extra 100$ can buy a LOT more graphics card.

i would definitely get an i5 over a c2q. i5 runs cooler and is definitely a lot faster. $100 could be the difference between onboard graphics and a 4770 (big) or the difference between a gtx 275 and a 285 (tiny).

 

[Atech]

i7 860 is obviously the better choice, it is faster than the 920 out of the box, the 920 does not overclock well and it gets too much heat


those morons who say 920 is better, maybe they think since 920 is a bigger number than 860, hence it must be better.

Say what?

 

[doz]

Overclock well? Im running 4.2ghz @ 1.25v right now and my daily load temps dont go above 60c w/ 24-25c ambient.

Tell me 920's dont clock well. It took my E8400 C0 1.53v to hit 4.3ghz.

If you are looking to upgrade from C2Q, Id go i7 1366. If you are running an older system and want a budget upgrade, Id look into i5 for sure. In either case, Id pass on 1156 i7, it just doesnt seem logical especially when the prices coming out are higher on the cpu side (and just for 133mhz more, 920 D0 can clock well). You can pickup x58 boards cheaper now for under $200 no problem and the difference in the CPU cost makes it almost negligible. Spend a few more bucks and have 1366 i7 IMO.

 

[Freezebyte]

I knew I was fucked getting socket 775 last year for my new intel build to save money, but damn, this REALLY sucks now. I wish I knew how the Core 2's compare

 

[ToastyBread]

If you are looking to upgrade from C2Q, Id go i7 1366. If you are running an older system and want a budget upgrade, Id look into i5 for sure. In either case, Id pass on 1156 i7, it just doesnt seem logical especially when the prices coming out are higher on the cpu side (and just for 133mhz more, 920 D0 can clock well). You can pickup x58 boards cheaper now for under $200 no problem and the difference in the CPU cost makes it almost negligible. Spend a few more bucks and have 1366 i7 IMO.

I think it depends on the upgrade situation too. For those who prefer SFF systems, Lynnfield i7s could be much more desirable if the system isn't likely to be overclocked. Many of the X58 mATX boards reportedly have very hot IOH's, and between that and the higher power consumption/thermal output of a 920 vs. the lower thermal output of the 860/870 and P55, I think it would potentially work better in such systems.

It all comes down to what the plan of use and build is, ultimately.

 

[Freezebyte]

I think it depends on the upgrade situation too. For those who prefer SFF systems, Lynnfield i7s could be much more desirable if the system isn't likely to be overclocked. Many of the X58 mATX boards reportedly have very hot IOH's, and between that and the higher power consumption/thermal output of a 920 vs. the lower thermal output of the 860/870 and P55, I think it would potentially work better in such systems.

It all comes down to what the plan of use and build is, ultimately.

Meh, I like SFF builds as much as the the next guy, so I don' t worry about heat, I just crank up the fans / mod the case.

Make me wonder if I should swap out my socket 775 setup for a mATX i7 920 setup while I have the chance even though its not even a year old. I hate this hobby sometimes...

 

[vengence]

Whether or not they're referencing power to the CPU specifically or power to the system as a whole, there's no denying that Lynnfield is a more power-efficient processor. It's rated at 95W TDP, vs. Bloomfield's 130W TDP. Even if you then factor in chipset power consumption and such, power consumption charts are still viable for use in comparisons, because currently you're limited to the X58 chipset for Socket 1366, and P55 for 1156. Maybe if/when other chipsets appear, you can then remove that aspect for power efficiency values, but people can't just say that 920, at 130W TDP, is more power-efficient than a 95W TDP. A single additional channel of memory isn't responsible for an extra 35W TDP on Bloomfield...

Between the NF200 which is said to draw 30-40Watts and the extra stick of ram, I think 35W is well within the margin of error. Furthermore, it doesn't bother you that a processor that can in theory do twice the calculations (twice the threads) doesn't have anywhere near twice the TDP, but the 750 is the more efficient one? That doesn't make sense.

Even with more on the die, if the core design is consuming less power overall (as Lynnfield is), your thermal output will be down. Any increase in thermal output from the PCIe controller is likely easily overshadowed by the much larger reduction in power consumption of the processor itself. From my experience building systems and viewing heat output and power consumption, and unless someone can show otherwise, thermal output is very much related to power consumption of the chip. As power consumption goes up, thermal output goes up. When power consumption goes down, well...

TDP is not always equal to the power draw, but if you want to assume it is, then why is the TDP of the 750 95W and the 920 130W. Twice the threads should be 190W shouldn't it?

The only C2Q that can even come close in any way to the i5 750 is the C2Q Q9650, at what is currently a $110 price premium on Newegg over the i5 750. The C2Q's around $200, are no match in any way for it. Given that the cheapest C2Q is $150 on Newegg and gets destroyed by an i5 750, yeah, I can easily recommend the 750 for $50 more

Like I said, I haven't looked it up lately.

 

[skitlets]

I knew I was fucked getting socket 775 last year for my new intel build to save money, but damn, this REALLY sucks now. I wish I knew how the Core 2's compare

If you look at the Anandtech charts, the Q9650 at 3.0ghz isn't very far behind in games and video encoding, the two things I need cpu power for. Too bad I'm limited with my Q9550 to about 3.6ghz due to my motherboard whereas the Nehalems are pushing close to 4ghz. The Yorkfields also lose in the power efficiency department.

 

[kllrnohj]

Furthermore, it doesn't bother you that a processor that can in theory do twice the calculations (twice the threads) doesn't have anywhere near twice the TDP, but the 750 is the more efficient one? That doesn't make sense.

You keep saying that but it is completely false. HT does *not* allow the processor to do twice the work, no where near it. HT allows the CPU to more effectively use its resources (as in thread 1 can fetch crap from memory while thread 2 crunches some numbers). In [H]'s review on multimedia benchmarks, HT gave a 10% boost in performance *at most*. In most tests HT had almost no impact. HT only allows a CPU to work on another thread while a different one is waiting for something. It *does not* allow the CPU to do twice the work. Even synthetic tests only saw gains of 25% in a single case.

Given that you specifically said "twice the calculations", HT would actually provide *ZERO* improvement in that case (assuming threads that never needed to access any memory or cache or anything of that nature and was strictly computations - which of course never happens in the real world, but you did say "in theory").

 

[Joe Average]

HT's true benefit comes through first and foremost on properly multi-threaded apps which makes perfect sense. If you have a single threaded app and you run it on a processor that's HT-enabled, it's not like it's going to magically get higher numbers - the HT aspect would never be brought into play because you're dealing with one thread.

Now, same app, properly multi-threaded (recompiled I suppose, this is just an example), on the same processor - and this entire example is supposed to be a single core processor with HT-enabled like a Pentium 4 w/HT, not a dual/quad core CPU) - and wham, you're gonna get a boost since the HT is now called into action.

No, we all know implicitly it doesn't mean "twice as fast 'cause you've got two threads a-rollin'..." but, it does offer some improvement over single core single threaded non-HT-enabled platforms.

The best example I noted in the current reviews of this aspect, how valuable HT can be, is on Anand's review showing the x264 encoding benchmark. The i5 750 at 2.66 GHz with no HT crunched the second pass at about 21 fps; the i7 870 at 2.93 GHz crunched the second pass at about 27.6 fps - that's a HUGE jump in performance, and it's not because of the 270 MHz difference in clockspeed. The 270 MHz difference works out to about ~10% on the clockspeed but the actual fps difference comes in at ~26%... so one could work out a ~15% difference in performance just because of the HT.

The i7 920 at the same clockspeed came in at 26.7 so... and the i7 920 was tested in a dual channel RAM configuration, just as the i7 870 was so that potential massive benefit from the 920's triple channel capability wasn't even a factor in that test.

That's roughly a 25% difference at nearly the same clockspeed - and it's because of the HT more than anything else...

 

[ToastyBread]

Between the NF200 which is said to draw 30-40Watts and the extra stick of ram, I think 35W is well within the margin of error. Furthermore, it doesn't bother you that a processor that can in theory do twice the calculations (twice the threads) doesn't have anywhere near twice the TDP, but the 750 is the more efficient one? That doesn't make sense.

What I'm trying to get across is that the 35W TDP difference between the two processors is *independent* of any chipsets, etc. When you look at benchmarks showing system power consumption values, yes, you can factor in chipset and other component power consumption. But when we're referring to actual processor TDP, that's specific to the processor itself, and not any associated chipsets or components. Thus, there is a 35W TDP difference between Lynnfield (95W) and Bloomfield (130W), and you can't say that the difference is due to NF200 or a third memory channel.

Edit - I found this nice quote from Tom's HW, which goes along with this logic:

"There is some notable power savings here compared to X58, though. To begin, Lynnfield sports a 95W TDP. Bloomfield is 130W. The X58 Express IOH is a 22W part. That vanishes completely. P55 uses up to 4.7W. And ICH10R consumes up to 4.5W. Add it all up and you&#8217;re down more than 56W right off the bat."

http://www.tomshardware.com/reviews/intel-core-i5,2410-7.html

Factor in the power consumption of an NF200 chip present with a lot of X58 boards, and that 56W TDP difference goes up even more.

In regards to the argument of a processor doing twice the calculations, as others have mentioned, you can't take four logical cores + 4 threads from HT, and say it's doing twice the calculations. Hyperthreading doesn't work that way. That thread isn't a physical core, and thus you can't treat it like it's an 8-core processor. For any application where the 750 or the 920 are relatively equal in performance, the 750 is the most efficient processor. The 920 only becomes more "efficient" at the point in time where it's able to complete a task in such a faster amount of time, that the reduced power consumption of the 750 is lost because it's taking it that much longer to complete the task. At that point and beyond, the 920 is more "efficient" solely because it can complete it so much faster. But for most single-threaded applications, the 750 will remain more efficient.

TDP is not always equal to the power draw, but if you want to assume it is, then why is the TDP of the 750 95W and the 920 130W. Twice the threads should be 190W shouldn't it?

Because HTing doesn't count as a logical core, and thus doesn't consume the power that a true, physical logical core would.

To put that into a better perspective: both the 750 and the 860 are rated at 95W TDP. The 750 has 4 "Threads" via 4 physical cores. The 860 has 8 threads, 4 from its 4 physical cores, and 4 from HT. Thus, using your definition, the 860 has just as many "threads" as the 920, and, since the 860 has 2x the threads of the 750, it should be drawing more power. And yet, it's rated 95W TDP, just like the 750. Because HTing is simply one core capable of processing multiple threads. The OS interprets it as two "virtual" logical cores, but the reality is that it's still just a single physical core that is carrying out the process. And that core is still set at a fixed TDP (for example: when the 860 is running under full load across all four cores, each core is delegated approximately 23.75W TDP). HTing really doesn't effect it much at all. The fact that the 750 and the 860/870 all are rated for the same TDP, shows that you can't use HTing as an argument towards power consumption.

 

[kllrnohj]

HT's true benefit comes through first and foremost on properly multi-threaded apps which makes perfect sense. If you have a single threaded app and you run it on a processor that's HT-enabled, it's not like it's going to magically get higher numbers - the HT aspect would never be brought into play because you're dealing with one thread.

Now, same app, properly multi-threaded (recompiled I suppose, this is just an example), on the same processor - and this entire example is supposed to be a single core processor with HT-enabled like a Pentium 4 w/HT, not a dual/quad core CPU) - and wham, you're gonna get a boost since the HT is now called into action.

No, we all know implicitly it doesn't mean "twice as fast 'cause you've got two threads a-rollin'..." but, it does offer some improvement over single core single threaded non-HT-enabled platforms.

The best example I noted in the current reviews of this aspect, how valuable HT can be, is on Anand's review showing the x264 encoding benchmark. The i5 750 at 2.66 GHz with no HT crunched the second pass at about 21 fps; the i7 870 at 2.93 GHz crunched the second pass at about 27.6 fps - that's a HUGE jump in performance, and it's not because of the 270 MHz difference in clockspeed. The 270 MHz difference works out to about ~10% on the clockspeed but the actual fps difference comes in at ~26%... so one could work out a ~15% difference in performance just because of the HT.

The i7 920 at the same clockspeed came in at 26.7 so... and the i7 920 was tested in a dual channel RAM configuration, just as the i7 870 was so that potential massive benefit from the 920's triple channel capability wasn't even a factor in that test.

That's roughly a 25% difference at nearly the same clockspeed - and it's because of the HT more than anything else...

Oh absolutely, but again, we are talking ~25% boost in performance, which is a far cry from 100% boost in performance

I'm not saying HT is pointless, far from it, just pointing out its actual performance improvements But that is also one of the rarer "best" case scenarios. The first pass showed a whopping 2.6% improvement from HT. Divx was <1%. 3dsmax 14%, Blender 10%. So we have a range of performance improvement due to HT from ~0% to 25% depending on the application. A solid overall improvement, yes, but twice as fast? Not even close.