A week ago, we reported on a performance test of Nova Lake desktop CPUs—Intel’s new generation hoped to repair the impression left by Arrow Lake processors two years ago. After the first numbers from the PassMark benchmark, we now have more, and this time they concern gaming performance. That was the main weakness of Arrow Lake processors, and for Nova Lake it may thus be the most important factor.
Last week’s leak presented a rather interesting situation, if it can be trusted: the high-performance P-Cores in Nova Lake processors should, according to it, offer up to 22% better performance per 1 MHz of clock speed than the P-Cores in Arrow Lake processors. That would allow Intel to significantly strengthen the CPU’s crucial single-thread performance, provided clock speeds are decent as well. It is, however, only one specific benchmark, of a more application-oriented type. Moreover, the PassMark database does not reveal what clock speeds Nova Lake will actually reach, so considerable uncertainty remains about whether Intel is heading for a big success or another disappointment.
A better picture might be provided by further reports that appeared afterward. Two different sources have come forward claiming to have information from internal Intel documents that state the performance Intel expects from the final Nova Lake CPUs. The first of these reports appeared on Twitter, but its original source is a comment by Reddit user Suspicious_pasta, who refers to “slides” but without any evidence or image of these materials. It is thus unclear whether it might be just someone’s fabrication.
First Leak: Reddit
If the information is genuine, Intel expects single-thread performance to rise by up to 17% for BFC-series processors with a large gaming cache (also referred to as BLLC). That is a smaller increase than the 22% IPC improvement according to PassMark—either because Nova Lake will have lower clock speeds, or because PassMark is an outlier and the average IPC improvement will be lower. For non-BFC processors without BLLC, however, the increase in single-thread performance is said to be only 8%, which is considerably worse.
Gaming performance is reportedly to rise by up to 20%, but this is probably the figure for gaming BFC processors with BLLC, not for regular Nova Lake. It may also include the effect of the APX extension, which according to Intel can reportedly deliver up to a 15% performance increase (in games, however, its impact will typically be lower).
Second Leak: VideoCardz
Fortunately, another, separate leak then appeared. The VideoCardz website writes that it also obtained some Intel documents on Nova Lake processors and found information in them about the processors’ expected (or promised) performance. Curiously, though, it differs. Whether this means that the earlier, less credible leak is fabricated or otherwise off the mark, we do not yet know.
According to VideoCardz, Intel claims that the Intel Nova Lake processor with 28 cores but without BLLC—which should therefore apparently be the 125W model, the Core Ultra 9 4950K—will have “up to +16%” higher gaming performance than the Arrow Lake-generation Core Ultra 9 285K, according to a metric labeled “Gaming Index.” This is said to come with somewhat higher power consumption. While the Core Ultra 9 285K has a maximum turbo power (PL2) of 250 W, Nova Lake will have 285 W (or at least that was how PL2 was set during the test).
The score of the 125W gaming model with 28 cores and BLLC is also listed—Intel also internally designates this variant “DS,” but according to the leaked data, this processor should be sold as the Core Ultra 9 4970K BFC. This processor is reportedly to have +33% higher gaming performance than the Core Ultra 9 285K. Its PL2 will be slightly higher still, at 296 W.
These projections date from this August, so they are already a few months old. Note that Intel compares performance with the Core Ultra 9 285K, not with the Core Ultra 7 270K Plus of the refresh lineup, in which gaming performance improved somewhat. The Arrow Lake processor is also used in this comparison with DDR5-6400 memory, whereas both Nova Lake models have faster DDR5-8000 (this speed will be supported through XMP profiles).
There is also no comparison with the Core i9-14900K or with AMD processors. The stated percentages unfortunately probably cannot be directly mapped onto other gaming CPU tests to guess whether the 33% improvement will be enough to catch up with processors such as the Ryzen 7 9800X3D. Intel’s position in gaming PCs should certainly improve, though.
Projections May Not Yet Match Final Performance
It should be pointed out, however, that these internal (or partner-oriented) performance projections may not entirely match reality. The nature of these numbers may be similar to officially presented marketing benchmarks—companies may, for example, selectively cite the tests that suit them better or otherwise improve the result. In general, though, these projections should probably be closer to reality, because in its relationship with PC makers, Intel must be less creative than in marketing and promotion aimed at end customers or the media.
According to VideoCardz, the percentages indicating gaming performance increases do not directly state gaming performance measured as frame rates in benchmarks, but are quantified by a kind of “Gaming Index” that Intel defines in some way to express CPU gaming performance. Intel creates the methodology of this Gaming Index itself, so you have to trust that the index reflects performance reasonably realistically (at least it is a comparison of Intel chips with other Intel chips, so there is probably less room for tricks).
Even with these caveats, though, the figures are interesting, at least in the case of the VideoCardz report, where one can probably rely more on the site truly drawing on legitimate documents. A 33% increase in gaming performance could make Nova Lake processors a serious choice even for demanding gaming, one that could finally compete with AMD’s X3D processors. Intel could thus truly manage to halt its slide in gaming PCs and even win back market share.
Sources: Reddit, VideoCardz
English translation and edit by Jozef Dudáš
