The Haswell Review: Intel Core i7-4770K & i5-4670K Tested
by Anand Lal Shimpi on June 1, 2013 10:00 AM ESTThe Launch Lineup: Quad Cores For All
As was the case with the launch of Ivy Bridge last year, Intel is initially launching with their high-end quad core parts, and as the year passes on will progressively rollout dual cores, low voltage parts, and other lower-end parts. That means the bigger notebooks and naturally the performance desktops will arrive first, followed by the ultraportables, Ultrabooks and more affordable desktops. One change however is that Intel will be launching their first BGA (non-socketed) Haswell part right away, the Iris Pro equipped i7-4770R.
| Intel 4th Gen Core i7 Desktop Processors | ||||||
| Model | Core i7-4770K | Core i7-4770 | Core i7-4770S | Core i7-4770T | Core i7-4770R | Core i7-4765T |
| Cores/Threads | 4/8 | 4/8 | 4/8 | 4/8 | 4/8 | 4/8 |
| CPU Base Freq | 3.5 | 3.4 | 3.1 | 2.5 | 3.2 | 2.0 |
| Max Turbo | 3.9 (Unlocked) | 3.9 | 3.9 | 3.7 | 3.9 | 3.0 |
| Test TDP | 84W | 84W | 65W | 45W | 65W | 35W |
| HD Graphics | 4600 | 4600 | 4600 | 4600 | Iris Pro 5200 | 4600 |
| GPU Max Clock | 1250 | 1200 | 1200 | 1200 | 1300 | 1200 |
| L3 Cache | 8MB | 8MB | 8MB | 8MB | 6MB | 8MB |
| DDR3 Support | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 |
| vPro/TXT/VT-d/SIPP | No | Yes | Yes | Yes | No | Yes |
| Package | LGA-1150 | LGA-1150 | LGA-1150 | LGA-1150 | BGA | LGA-1150 |
| Price | $339 | $303 | $303 | $303 | OEM | $303 |
Starting at the top of the product and performance stack, we have the desktop Core i7 parts. All of these CPUs feature Hyper-Threading Technology, so they’re the same quad-core with four virtual cores that we’ve seen since Bloomfield hit the scene. The fastest chip for most purposes remains the K-series 4770K, with its unlocked multiplier and slightly higher base clock speed. Base core clocks as well as maximum Turbo Boost clocks are basically dictated by the TDP, with the 4770S being less likely to maintain maximum turbo most likely, and the 4770T and 4765T giving up quite a bit more in clock speed in order to hit substantially lower power targets.
It’s worth pointing out that the highest “Test TDP” values are up slightly relative to the last generation Ivy Bridge equivalents—84W instead of 77W. Mobile TDPs are a different matter, and as we’ll discuss elsewhere they’re all 2W higher, but that is further offset by the improved idle power consumption Haswell brings.
Nearly all of these are GT2 graphics configurations (20 EUs), so they should be slightly faster than the last generation HD 4000 in graphics workloads. The one exception is the i7-4770R, which is also the only chip that comes in a BGA package. The reasoning here is simple if perhaps flawed: if you want the fastest iGPU configuration (GT3e with 40 EUs and embedded DRAM), you’re probably not going to have a discrete GPU and will most likely be purchasing an OEM desktop. Interestingly, the 4770R also drops the L3 cache down to 6MB, and it’s not clear whether this is due to it having no real benefit (i.e. the eDRAM functions as an even larger L4 cache), or if it’s to reduce power use slightly, or Intel may have a separate die for this particular configuration. Then again, maybe Intel is just busily creating a bit of extra market segmentation.
Not included in the above table are all the common features to the entire Core i7 line: AVX2 instructions, Quick Sync, AES-NI, PCIe 3.0, and Intel Virtualization Technology. As we’ve seen in the past, the K-series parts (and now the R-series as well) omit support for vPro, TXT, VT-d, and SIPP from the list. The 4770K is an enthusiast part with overclocking support, so that makes some sense, but the 4770R doesn’t really have the same qualification. Presumably it’s intended for the consumer market, as businesses are less likely to need the Iris Pro graphics.
| Intel 4th Gen Core i5 Desktop Processors | ||||||
| Model | Core i5-4670K | Core i5-4670 | Core i5-4670S | Core i5-4670T | Core i5-4570 | Core i5-4570S |
| Cores/Threads | 4/4 | 4/4 | 4/4 | 4/4 | 4/4 | 4/4 |
| CPU Base Freq | 3.4 | 3.4 | 3.1 | 2.3 | 3.2 | 2.9 |
| Max Turbo | 3.8 (Unlocked) | 3.8 | 3.8 | 3.3 | 3.6 | 3.6 |
| Test TDP | 84W | 84W | 65W | 45W | 84W | 65W |
| HD Graphics | 4600 | 4600 | 4600 | 4600 | 4600 | 4600 |
| GPU Max Clock | 1200 | 1200 | 1200 | 1200 | 1150 | 1150 |
| L3 Cache | 6MB | 6MB | 6MB | 6MB | 6MB | 6MB |
| DDR3 Support | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 | 1333/1600 |
| vPro/TXT/VT-d/SIPP | No | Yes | Yes | Yes | Yes | Yes |
| Package | LGA-1150 | LGA-1150 | LGA-1150 | LGA-1150 | LGA-1150 | LGA-1150 |
| Price | $242 | $213 | $213 | $213 | $192 | $192 |
The Core i5 lineup basically rehashes the above story, only now without Hyper-Threading. For many users, Core i5 is the sweet spot of price and performance, delivering nearly all the performance of the i7 models at 2/3 the price. There aren’t any Iris or Iris Pro Core i5 desktop parts, at least not yet, and all of the above CPUs are using the GT2 graphics configuration. As above, the K-series part also lacks vPro/TXT/VT-d support but comes with an unlocked multiplier.
Obviously we’re still missing all of the Core i3 parts, which are likely to be dual-core once more, along with some dual-core i5 parts as well. These are probably going to come in another quarter, or at least a month or two out, as there’s no real need for Intel to launch their lower cost parts right now. Similarly, we don’t have any Celeron or Pentium Haswell derivatives launching yet, and judging by the Ivy Bridge rollout I suspect it may be a couple quarters before Intel pushes out ultra-budget Haswell chips. For now, the Ivy Bridge Celeron/Pentium parts are likely as low as Intel wants to go down the food chain for their “big core” architectures.
For those interested in the mobile side of things, we’ve broken out those parts into a separate Pipeline article.
Facebook
Twitter
RSS
210 Comments
View All Comments
Amaranthus - Monday, June 3, 2013 - link
One of the main (and already implemented) uses of TSX is hardware lock elision. I'd guess the hypothesis is that physics code takes locks defensively but rarely actually have contention because they're working on different parts of the world. In this scenario more fine grained locks on sections of the world would let you scale better but that is a lot of work and HLE gives you the same benefit for free.Jaybus - Monday, June 3, 2013 - link
No. HLE (XACQUIRE and XRELEASE) do nothing by themselves. They reuse REPNE/REPE prefixes and on CPUs that do not support TSX are ignored on instructions that would be valid for XACQUIRE/XRELEASE if TSX were available. It is a backward compatibility method. Since all of those instructions may have a LOCK prefix, without TSX capability, a normal lock is used, NOT the optimistic locking provided by TSX that allows other threads to see the lock as already free.Without TSX the code is still (software) lock-free, but there is no possibility of multiple threads accessing the same memory simultaneously (as there is with TSX), so one or more threads will see a pipeline stall due to the LOCK prefix.
bji - Monday, June 3, 2013 - link
I can't imagine that lock elision is that beneficial to very many applications. Lock contention is almost never a significant performance bottleneck; yeah there are poorly designed applications where lock contention can have a more significant effect, but proper multithreaded coding has the contended sections of code reduced to the smallest number of instructions possible, at which point the effects of lock contention are minimized.In order to take advantage of transactional memory and get the full benefits of TSX you have to write such radically different algorithms that I doubt that it's worth it except in the most unusual and specific cases. OK so you can use TSX instructions to make a hashtable or other container class suffer slightly less from lock contention, but that is oh so very rarely a significant aspect to the performance of any program.
klmccaughey - Monday, June 3, 2013 - link
As a programmer, I disagree. This is a very useful feature set that, if it was more widely adopted, would prove very useful for many workaday tasks that the CPU performs.bji - Monday, June 3, 2013 - link
As a programmer, I am pretty sure that the benefits of TSX are limited to a very unusual and uncommon set of problems the performance increase of which will mean very very little to 99.99% of users 99.99% of the time. Also fully transactional memory algorithms require significant rework from their non-transactional counterparts meaning that taking full advantage of TSX takes developer effort which will not be worth it except in very rare circumstances.The HLE instructions may have some very minor benefit because they can be used with algorithms that don't need to be reworked at all (you just get a little bit more parallelism for free), but even then you're going to be avoiding some lock contention; even if you completely eliminated lock contention from most algorithms they would only be fractionally faster in real world usage. Lock contention just isn't that big of a deal in normal circumstances.
klmccaughey - Monday, June 3, 2013 - link
Exactly. It would be the ubiquity of these features that would cause them to be useful - splitting them into segments defeats the adoption and use of said features. Intel are pushing segmentation too hard (too greedily?)bill5 - Saturday, June 1, 2013 - link
kind if a weird, scattered review.loved the q6600 though, since i still have one. and the 8350, since i have my eye on it.
be interesting if this pushed 8350 prices down enough to be more attractive (it's currently only 180 on newegg). if not i'll probably go with i5 4670 (even though i'm getting tired of these faux msrp's, bet money that chip will be 229 on newegg forget 213)
ps, my bill4 account was apparently banned (it kept saying i was posting spam wouldn't allow me to post) i post controversial things that probably get downvoted, but they arent spam. please stop doing that.
bill5 - Saturday, June 1, 2013 - link
this really shows where the 8350 fails, single thread.it looks like clock for clock it's ipc may be similar to my q6600. it only gains in single thread due to the gaudy 4.0 clock speed.
otoh go to multithread and it holds it's own against the other ~200 intel chips.
Nexus-7 - Saturday, June 1, 2013 - link
In for one 4770k --I'm coming from an i5-750 running at 4GHz. I'm thinking this will be a sufficiently large leap forward although I'm tempted to wait for Ivy Bridge-E's 6c12t monsters.
chizow - Saturday, June 1, 2013 - link
Similar boat, but I told myself I wouldn't wait for Intel's E platform anymore. X58 may be the last great E platform, mainly because it actually preceded the rest of the mainstream performance parts. Intel seems to sit on these E platforms for so long now that they almost become irrelevant by the time they launch.